/* * Copyright (c) 2015 Andrew Kelley * * This file is part of zig, which is MIT licensed. * See http://opensource.org/licenses/MIT */ #include "analyze.hpp" #include "ast_render.hpp" #include "codegen.hpp" #include "config.h" #include "errmsg.hpp" #include "error.hpp" #include "hash_map.hpp" #include "ir.hpp" #include "os.hpp" #include "target.hpp" #include "util.hpp" #include "zig_llvm.h" #include "stage2.h" #include "dump_analysis.hpp" #include "softfloat.hpp" #include #include #include enum ResumeId { ResumeIdManual, ResumeIdReturn, ResumeIdCall, }; static ZigPackage *new_package(const char *root_src_dir, const char *root_src_path, const char *pkg_path) { ZigPackage *entry = heap::c_allocator.create(); entry->package_table.init(4); buf_init_from_str(&entry->root_src_dir, root_src_dir); buf_init_from_str(&entry->root_src_path, root_src_path); buf_init_from_str(&entry->pkg_path, pkg_path); return entry; } ZigPackage *new_anonymous_package() { return new_package("", "", ""); } static const char *symbols_that_llvm_depends_on[] = { "memcpy", "memset", "sqrt", "powi", "sin", "cos", "pow", "exp", "exp2", "log", "log10", "log2", "fma", "fabs", "minnum", "maxnum", "copysign", "floor", "ceil", "trunc", "rint", "nearbyint", "round", // TODO probably all of compiler-rt needs to go here }; void codegen_set_strip(CodeGen *g, bool strip) { g->strip_debug_symbols = strip; if (!target_has_debug_info(g->zig_target)) { g->strip_debug_symbols = true; } } static void render_const_val(CodeGen *g, ZigValue *const_val, const char *name); static void render_const_val_global(CodeGen *g, ZigValue *const_val, const char *name); static LLVMValueRef gen_const_val(CodeGen *g, ZigValue *const_val, const char *name); static void generate_error_name_table(CodeGen *g); static bool value_is_all_undef(CodeGen *g, ZigValue *const_val); static void gen_undef_init(CodeGen *g, ZigType *ptr_type, ZigType *value_type, LLVMValueRef ptr); static LLVMValueRef build_alloca(CodeGen *g, ZigType *type_entry, const char *name, uint32_t alignment); static LLVMValueRef gen_await_early_return(CodeGen *g, IrInstGen *source_instr, LLVMValueRef target_frame_ptr, ZigType *result_type, ZigType *ptr_result_type, LLVMValueRef result_loc, bool non_async); static void addLLVMAttr(LLVMValueRef val, LLVMAttributeIndex attr_index, const char *attr_name) { unsigned kind_id = LLVMGetEnumAttributeKindForName(attr_name, strlen(attr_name)); assert(kind_id != 0); LLVMAttributeRef llvm_attr = LLVMCreateEnumAttribute(LLVMGetGlobalContext(), kind_id, 0); LLVMAddAttributeAtIndex(val, attr_index, llvm_attr); } static void addLLVMAttrStr(LLVMValueRef val, LLVMAttributeIndex attr_index, const char *attr_name, const char *attr_val) { LLVMAttributeRef llvm_attr = LLVMCreateStringAttribute(LLVMGetGlobalContext(), attr_name, (unsigned)strlen(attr_name), attr_val, (unsigned)strlen(attr_val)); LLVMAddAttributeAtIndex(val, attr_index, llvm_attr); } static void addLLVMAttrInt(LLVMValueRef val, LLVMAttributeIndex attr_index, const char *attr_name, uint64_t attr_val) { unsigned kind_id = LLVMGetEnumAttributeKindForName(attr_name, strlen(attr_name)); assert(kind_id != 0); LLVMAttributeRef llvm_attr = LLVMCreateEnumAttribute(LLVMGetGlobalContext(), kind_id, attr_val); LLVMAddAttributeAtIndex(val, attr_index, llvm_attr); } static void addLLVMFnAttr(LLVMValueRef fn_val, const char *attr_name) { return addLLVMAttr(fn_val, -1, attr_name); } static void addLLVMFnAttrStr(LLVMValueRef fn_val, const char *attr_name, const char *attr_val) { return addLLVMAttrStr(fn_val, -1, attr_name, attr_val); } static void addLLVMFnAttrInt(LLVMValueRef fn_val, const char *attr_name, uint64_t attr_val) { return addLLVMAttrInt(fn_val, -1, attr_name, attr_val); } static void addLLVMArgAttr(LLVMValueRef fn_val, unsigned param_index, const char *attr_name) { return addLLVMAttr(fn_val, param_index + 1, attr_name); } static void addLLVMArgAttrInt(LLVMValueRef fn_val, unsigned param_index, const char *attr_name, uint64_t attr_val) { return addLLVMAttrInt(fn_val, param_index + 1, attr_name, attr_val); } static bool is_symbol_available(CodeGen *g, const char *name) { Buf *buf_name = buf_create_from_str(name); bool result = g->exported_symbol_names.maybe_get(buf_name) == nullptr && g->external_symbol_names.maybe_get(buf_name) == nullptr; buf_destroy(buf_name); return result; } static const char *get_mangled_name(CodeGen *g, const char *original_name) { if (is_symbol_available(g, original_name)) return original_name; int n = 0; for (;; n += 1) { const char *new_name = buf_ptr(buf_sprintf("%s.%d", original_name, n)); if (is_symbol_available(g, new_name)) { return new_name; } } } // Sync this with emit_error_unless_callconv_allowed_for_target in analyze.cpp static ZigLLVM_CallingConv get_llvm_cc(CodeGen *g, CallingConvention cc) { switch (cc) { case CallingConventionUnspecified: return ZigLLVM_Fast; case CallingConventionC: return ZigLLVM_C; case CallingConventionNaked: zig_unreachable(); case CallingConventionStdcall: assert(g->zig_target->arch == ZigLLVM_x86); return ZigLLVM_X86_StdCall; case CallingConventionFastcall: assert(g->zig_target->arch == ZigLLVM_x86); return ZigLLVM_X86_FastCall; case CallingConventionVectorcall: if (g->zig_target->arch == ZigLLVM_x86) return ZigLLVM_X86_VectorCall; if (target_is_arm(g->zig_target) && target_arch_pointer_bit_width(g->zig_target->arch) == 64) return ZigLLVM_AArch64_VectorCall; zig_unreachable(); case CallingConventionThiscall: assert(g->zig_target->arch == ZigLLVM_x86); return ZigLLVM_X86_ThisCall; case CallingConventionAsync: return ZigLLVM_Fast; case CallingConventionAPCS: assert(target_is_arm(g->zig_target)); return ZigLLVM_ARM_APCS; case CallingConventionAAPCS: assert(target_is_arm(g->zig_target)); return ZigLLVM_ARM_AAPCS; case CallingConventionAAPCSVFP: assert(target_is_arm(g->zig_target)); return ZigLLVM_ARM_AAPCS_VFP; case CallingConventionInterrupt: if (g->zig_target->arch == ZigLLVM_x86 || g->zig_target->arch == ZigLLVM_x86_64) return ZigLLVM_X86_INTR; if (g->zig_target->arch == ZigLLVM_avr) return ZigLLVM_AVR_INTR; if (g->zig_target->arch == ZigLLVM_msp430) return ZigLLVM_MSP430_INTR; zig_unreachable(); case CallingConventionSignal: assert(g->zig_target->arch == ZigLLVM_avr); return ZigLLVM_AVR_SIGNAL; } zig_unreachable(); } static void add_uwtable_attr(CodeGen *g, LLVMValueRef fn_val) { if (g->zig_target->os == OsWindows) { addLLVMFnAttr(fn_val, "uwtable"); } } static LLVMLinkage to_llvm_linkage(GlobalLinkageId id, bool is_extern) { switch (id) { case GlobalLinkageIdInternal: return LLVMInternalLinkage; case GlobalLinkageIdStrong: return LLVMExternalLinkage; case GlobalLinkageIdWeak: if (is_extern) return LLVMExternalWeakLinkage; return LLVMWeakODRLinkage; case GlobalLinkageIdLinkOnce: return LLVMLinkOnceODRLinkage; } zig_unreachable(); } struct CalcLLVMFieldIndex { uint32_t offset; uint32_t field_index; }; static void calc_llvm_field_index_add(CodeGen *g, CalcLLVMFieldIndex *calc, ZigType *ty) { if (!type_has_bits(g, ty)) return; uint32_t ty_align = get_abi_alignment(g, ty); if (calc->offset % ty_align != 0) { uint32_t llvm_align = LLVMABIAlignmentOfType(g->target_data_ref, get_llvm_type(g, ty)); // Alignment according to Zig. uint32_t adj_offset = calc->offset + (ty_align - (calc->offset % ty_align)); // Alignment according to LLVM. uint32_t adj_llvm_offset = (calc->offset % llvm_align) ? calc->offset + (llvm_align - (calc->offset % llvm_align)) : calc->offset; // Cannot under-align structure fields. assert(adj_offset >= adj_llvm_offset); // Zig will insert an extra padding field here. if (adj_offset != adj_llvm_offset) calc->field_index += 1; calc->offset = adj_offset; } calc->offset += ty->abi_size; calc->field_index += 1; } // label (grep this): [fn_frame_struct_layout] static void frame_index_trace_arg_calc(CodeGen *g, CalcLLVMFieldIndex *calc, ZigType *return_type) { calc_llvm_field_index_add(g, calc, g->builtin_types.entry_usize); // function pointer calc_llvm_field_index_add(g, calc, g->builtin_types.entry_usize); // resume index calc_llvm_field_index_add(g, calc, g->builtin_types.entry_usize); // awaiter index if (type_has_bits(g, return_type)) { calc_llvm_field_index_add(g, calc, g->builtin_types.entry_usize); // *ReturnType (callee's) calc_llvm_field_index_add(g, calc, g->builtin_types.entry_usize); // *ReturnType (awaiter's) calc_llvm_field_index_add(g, calc, return_type); // ReturnType } } static uint32_t frame_index_trace_arg(CodeGen *g, ZigType *return_type) { CalcLLVMFieldIndex calc = {0}; frame_index_trace_arg_calc(g, &calc, return_type); return calc.field_index; } // label (grep this): [fn_frame_struct_layout] static void frame_index_arg_calc(CodeGen *g, CalcLLVMFieldIndex *calc, ZigType *return_type) { frame_index_trace_arg_calc(g, calc, return_type); if (codegen_fn_has_err_ret_tracing_arg(g, return_type)) { calc_llvm_field_index_add(g, calc, g->builtin_types.entry_usize); // *StackTrace (callee's) calc_llvm_field_index_add(g, calc, g->builtin_types.entry_usize); // *StackTrace (awaiter's) } } // label (grep this): [fn_frame_struct_layout] static uint32_t frame_index_trace_stack(CodeGen *g, ZigFn *fn) { size_t field_index = 6; bool have_stack_trace = codegen_fn_has_err_ret_tracing_arg(g, fn->type_entry->data.fn.fn_type_id.return_type); if (have_stack_trace) { field_index += 2; } field_index += fn->type_entry->data.fn.fn_type_id.param_count; ZigType *locals_struct = fn->frame_type->data.frame.locals_struct; TypeStructField *field = locals_struct->data.structure.fields[field_index]; return field->gen_index; } static uint32_t get_err_ret_trace_arg_index(CodeGen *g, ZigFn *fn_table_entry) { if (!g->have_err_ret_tracing) { return UINT32_MAX; } if (fn_is_async(fn_table_entry)) { return UINT32_MAX; } ZigType *fn_type = fn_table_entry->type_entry; if (!fn_type_can_fail(&fn_type->data.fn.fn_type_id)) { return UINT32_MAX; } ZigType *return_type = fn_type->data.fn.fn_type_id.return_type; bool first_arg_ret = type_has_bits(g, return_type) && handle_is_ptr(g, return_type); return first_arg_ret ? 1 : 0; } static void maybe_export_dll(CodeGen *g, LLVMValueRef global_value, GlobalLinkageId linkage) { if (linkage != GlobalLinkageIdInternal && g->zig_target->os == OsWindows && g->dll_export_fns) { LLVMSetDLLStorageClass(global_value, LLVMDLLExportStorageClass); } } static void maybe_import_dll(CodeGen *g, LLVMValueRef global_value, GlobalLinkageId linkage) { if (linkage != GlobalLinkageIdInternal && g->zig_target->os == OsWindows) { // TODO come up with a good explanation/understanding for why we never do // DLLImportStorageClass. Empirically it only causes problems. But let's have // this documented and then clean up the code accordingly. //LLVMSetDLLStorageClass(global_value, LLVMDLLImportStorageClass); } } static bool cc_want_sret_attr(CallingConvention cc) { switch (cc) { case CallingConventionNaked: zig_unreachable(); case CallingConventionC: case CallingConventionInterrupt: case CallingConventionSignal: case CallingConventionStdcall: case CallingConventionFastcall: case CallingConventionVectorcall: case CallingConventionThiscall: case CallingConventionAPCS: case CallingConventionAAPCS: case CallingConventionAAPCSVFP: return true; case CallingConventionAsync: case CallingConventionUnspecified: return false; } zig_unreachable(); } static bool codegen_have_frame_pointer(CodeGen *g) { return g->build_mode == BuildModeDebug; } static LLVMValueRef make_fn_llvm_value(CodeGen *g, ZigFn *fn) { const char *unmangled_name = buf_ptr(&fn->symbol_name); const char *symbol_name; GlobalLinkageId linkage; if (fn->body_node == nullptr) { symbol_name = unmangled_name; linkage = GlobalLinkageIdStrong; } else if (fn->export_list.length == 0) { symbol_name = get_mangled_name(g, unmangled_name); linkage = GlobalLinkageIdInternal; } else { GlobalExport *fn_export = &fn->export_list.items[0]; symbol_name = buf_ptr(&fn_export->name); linkage = fn_export->linkage; } CallingConvention cc = fn->type_entry->data.fn.fn_type_id.cc; bool is_async = fn_is_async(fn); ZigType *fn_type = fn->type_entry; // Make the raw_type_ref populated resolve_llvm_types_fn(g, fn); LLVMTypeRef fn_llvm_type = fn->raw_type_ref; LLVMValueRef llvm_fn = nullptr; if (fn->body_node == nullptr) { assert(fn->proto_node->type == NodeTypeFnProto); AstNodeFnProto *fn_proto = &fn->proto_node->data.fn_proto; const unsigned fn_addrspace = ZigLLVMDataLayoutGetProgramAddressSpace(g->target_data_ref); // The compiler tries to deduplicate extern definitions by looking up // their name, this was introduced to allow the declaration of the same // extern function with differing prototypes. // When Wasm is targeted this check becomes a problem as the user may // declare two (or more) extern functions sharing the same name but // imported from different modules! // To overcome this problem we generate a mangled identifier out of the // import and the function name, this name is only visible within the // compiler as we're telling LLVM (using 'wasm-import-name' and // 'wasm-import-name') what the real function name is and where to find // it. const bool use_mangled_name = target_is_wasm(g->zig_target) && fn_proto->is_extern && fn_proto->lib_name != nullptr; // Pick a weird name to avoid collisions... // This whole function should be burned to the ground. Buf *mangled_symbol_buf = use_mangled_name ? buf_sprintf("%s|%s", unmangled_name, buf_ptr(fn_proto->lib_name)) : nullptr; symbol_name = use_mangled_name ? buf_ptr(mangled_symbol_buf) : unmangled_name; LLVMValueRef existing_llvm_fn = LLVMGetNamedFunction(g->module, symbol_name); if (existing_llvm_fn) { if (mangled_symbol_buf) buf_destroy(mangled_symbol_buf); return LLVMConstBitCast(existing_llvm_fn, LLVMPointerType(fn_llvm_type, fn_addrspace)); } else { Buf *buf_symbol_name = buf_create_from_str(symbol_name); auto entry = g->exported_symbol_names.maybe_get(buf_symbol_name); buf_destroy(buf_symbol_name); if (entry == nullptr) { llvm_fn = LLVMAddFunction(g->module, symbol_name, fn_llvm_type); if (use_mangled_name) { addLLVMFnAttrStr(llvm_fn, "wasm-import-name", unmangled_name); addLLVMFnAttrStr(llvm_fn, "wasm-import-module", buf_ptr(fn_proto->lib_name)); } } else { assert(entry->value->id == TldIdFn); TldFn *tld_fn = reinterpret_cast(entry->value); // Make the raw_type_ref populated resolve_llvm_types_fn(g, tld_fn->fn_entry); tld_fn->fn_entry->llvm_value = LLVMAddFunction(g->module, symbol_name, tld_fn->fn_entry->raw_type_ref); llvm_fn = LLVMConstBitCast(tld_fn->fn_entry->llvm_value, LLVMPointerType(fn_llvm_type, fn_addrspace)); if (mangled_symbol_buf) buf_destroy(mangled_symbol_buf); return llvm_fn; } if (mangled_symbol_buf) buf_destroy(mangled_symbol_buf); } } else { if (llvm_fn == nullptr) { llvm_fn = LLVMAddFunction(g->module, symbol_name, fn_llvm_type); } for (size_t i = 1; i < fn->export_list.length; i += 1) { GlobalExport *fn_export = &fn->export_list.items[i]; LLVMAddAlias(g->module, LLVMTypeOf(llvm_fn), llvm_fn, buf_ptr(&fn_export->name)); } } switch (fn->fn_inline) { case FnInlineAlways: addLLVMFnAttr(llvm_fn, "alwaysinline"); g->inline_fns.append(fn); break; case FnInlineNever: addLLVMFnAttr(llvm_fn, "noinline"); break; case FnInlineAuto: if (fn->alignstack_value != 0) { addLLVMFnAttr(llvm_fn, "noinline"); } break; } if (cc == CallingConventionNaked) { addLLVMFnAttr(llvm_fn, "naked"); } else { ZigLLVMFunctionSetCallingConv(llvm_fn, get_llvm_cc(g, cc)); } bool want_cold = fn->is_cold; if (want_cold) { ZigLLVMAddFunctionAttrCold(llvm_fn); } LLVMSetLinkage(llvm_fn, to_llvm_linkage(linkage, fn->body_node == nullptr)); if (linkage == GlobalLinkageIdInternal) { LLVMSetUnnamedAddr(llvm_fn, true); } ZigType *return_type = fn_type->data.fn.fn_type_id.return_type; if (return_type->id == ZigTypeIdUnreachable) { addLLVMFnAttr(llvm_fn, "noreturn"); } if (fn->body_node != nullptr) { maybe_export_dll(g, llvm_fn, linkage); bool want_fn_safety = g->build_mode != BuildModeFastRelease && g->build_mode != BuildModeSmallRelease && !fn->def_scope->safety_off; if (want_fn_safety) { if (g->link_libc) { addLLVMFnAttr(llvm_fn, "sspstrong"); addLLVMFnAttrStr(llvm_fn, "stack-protector-buffer-size", "4"); } } if (g->have_stack_probing && !fn->def_scope->safety_off) { addLLVMFnAttrStr(llvm_fn, "probe-stack", "__zig_probe_stack"); } else if (g->zig_target->os == OsUefi) { addLLVMFnAttrStr(llvm_fn, "no-stack-arg-probe", ""); } } else { maybe_import_dll(g, llvm_fn, linkage); } if (fn->alignstack_value != 0) { addLLVMFnAttrInt(llvm_fn, "alignstack", fn->alignstack_value); } if (g->build_mode == BuildModeSmallRelease) { // Optimize for small code size. addLLVMFnAttr(llvm_fn, "minsize"); addLLVMFnAttr(llvm_fn, "optsize"); } addLLVMFnAttr(llvm_fn, "nounwind"); add_uwtable_attr(g, llvm_fn); addLLVMFnAttr(llvm_fn, "nobuiltin"); if (codegen_have_frame_pointer(g) && fn->fn_inline != FnInlineAlways) { ZigLLVMAddFunctionAttr(llvm_fn, "frame-pointer", "all"); } if (fn->section_name) { LLVMSetSection(llvm_fn, buf_ptr(fn->section_name)); } if (fn->align_bytes > 0) { LLVMSetAlignment(llvm_fn, (unsigned)fn->align_bytes); } else { // We'd like to set the best alignment for the function here, but on Darwin LLVM gives // "Cannot getTypeInfo() on a type that is unsized!" assertion failure when calling // any of the functions for getting alignment. Not specifying the alignment should // use the ABI alignment, which is fine. } if (is_async) { addLLVMArgAttr(llvm_fn, 0, "nonnull"); } else { unsigned init_gen_i = 0; if (!type_has_bits(g, return_type)) { // nothing to do } else if (type_is_nonnull_ptr(g, return_type)) { addLLVMAttr(llvm_fn, 0, "nonnull"); } else if (want_first_arg_sret(g, &fn_type->data.fn.fn_type_id)) { // Sret pointers must not be address 0 addLLVMArgAttr(llvm_fn, 0, "nonnull"); addLLVMArgAttr(llvm_fn, 0, "sret"); if (cc_want_sret_attr(cc)) { addLLVMArgAttr(llvm_fn, 0, "noalias"); } init_gen_i = 1; } // set parameter attributes FnWalk fn_walk = {}; fn_walk.id = FnWalkIdAttrs; fn_walk.data.attrs.fn = fn; fn_walk.data.attrs.llvm_fn = llvm_fn; fn_walk.data.attrs.gen_i = init_gen_i; walk_function_params(g, fn_type, &fn_walk); uint32_t err_ret_trace_arg_index = get_err_ret_trace_arg_index(g, fn); if (err_ret_trace_arg_index != UINT32_MAX) { // Error return trace memory is in the stack, which is impossible to be at address 0 // on any architecture. addLLVMArgAttr(llvm_fn, (unsigned)err_ret_trace_arg_index, "nonnull"); } } return llvm_fn; } static LLVMValueRef fn_llvm_value(CodeGen *g, ZigFn *fn) { if (fn->llvm_value) return fn->llvm_value; fn->llvm_value = make_fn_llvm_value(g, fn); fn->llvm_name = strdup(LLVMGetValueName(fn->llvm_value)); return fn->llvm_value; } static ZigLLVMDIScope *get_di_scope(CodeGen *g, Scope *scope) { if (scope->di_scope) return scope->di_scope; ZigType *import = get_scope_import(scope); switch (scope->id) { case ScopeIdCImport: zig_unreachable(); case ScopeIdFnDef: { assert(scope->parent); ScopeFnDef *fn_scope = (ScopeFnDef *)scope; ZigFn *fn_table_entry = fn_scope->fn_entry; if (!fn_table_entry->proto_node) return get_di_scope(g, scope->parent); unsigned line_number = (unsigned)(fn_table_entry->proto_node->line == 0) ? 0 : (fn_table_entry->proto_node->line + 1); unsigned scope_line = line_number; bool is_definition = fn_table_entry->body_node != nullptr; bool is_optimized = g->build_mode != BuildModeDebug; bool is_internal_linkage = (fn_table_entry->body_node != nullptr && fn_table_entry->export_list.length == 0); unsigned flags = ZigLLVM_DIFlags_StaticMember; ZigLLVMDIScope *fn_di_scope = get_di_scope(g, scope->parent); assert(fn_di_scope != nullptr); assert(fn_table_entry->raw_di_type != nullptr); ZigLLVMDISubprogram *subprogram = ZigLLVMCreateFunction(g->dbuilder, fn_di_scope, buf_ptr(&fn_table_entry->symbol_name), "", import->data.structure.root_struct->di_file, line_number, fn_table_entry->raw_di_type, is_internal_linkage, is_definition, scope_line, flags, is_optimized, nullptr); scope->di_scope = ZigLLVMSubprogramToScope(subprogram); if (!g->strip_debug_symbols) { ZigLLVMFnSetSubprogram(fn_llvm_value(g, fn_table_entry), subprogram); } return scope->di_scope; } case ScopeIdDecls: if (scope->parent) { ScopeDecls *decls_scope = (ScopeDecls *)scope; assert(decls_scope->container_type); scope->di_scope = ZigLLVMTypeToScope(get_llvm_di_type(g, decls_scope->container_type)); } else { scope->di_scope = ZigLLVMFileToScope(import->data.structure.root_struct->di_file); } return scope->di_scope; case ScopeIdBlock: case ScopeIdDefer: { assert(scope->parent); ZigLLVMDILexicalBlock *di_block = ZigLLVMCreateLexicalBlock(g->dbuilder, get_di_scope(g, scope->parent), import->data.structure.root_struct->di_file, (unsigned)scope->source_node->line + 1, (unsigned)scope->source_node->column + 1); scope->di_scope = ZigLLVMLexicalBlockToScope(di_block); return scope->di_scope; } case ScopeIdVarDecl: case ScopeIdDeferExpr: case ScopeIdLoop: case ScopeIdSuspend: case ScopeIdCompTime: case ScopeIdNoSuspend: case ScopeIdRuntime: case ScopeIdTypeOf: case ScopeIdExpr: return get_di_scope(g, scope->parent); } zig_unreachable(); } static void clear_debug_source_node(CodeGen *g) { ZigLLVMClearCurrentDebugLocation(g->builder); } static LLVMValueRef get_arithmetic_overflow_fn(CodeGen *g, ZigType *operand_type, const char *signed_name, const char *unsigned_name) { ZigType *int_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; char fn_name[64]; assert(int_type->id == ZigTypeIdInt); const char *signed_str = int_type->data.integral.is_signed ? signed_name : unsigned_name; LLVMTypeRef param_types[] = { get_llvm_type(g, operand_type), get_llvm_type(g, operand_type), }; if (operand_type->id == ZigTypeIdVector) { sprintf(fn_name, "llvm.%s.with.overflow.v%" PRIu64 "i%" PRIu32, signed_str, operand_type->data.vector.len, int_type->data.integral.bit_count); LLVMTypeRef return_elem_types[] = { get_llvm_type(g, operand_type), LLVMVectorType(LLVMInt1Type(), operand_type->data.vector.len), }; LLVMTypeRef return_struct_type = LLVMStructType(return_elem_types, 2, false); LLVMTypeRef fn_type = LLVMFunctionType(return_struct_type, param_types, 2, false); LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type); assert(LLVMGetIntrinsicID(fn_val)); return fn_val; } else { sprintf(fn_name, "llvm.%s.with.overflow.i%" PRIu32, signed_str, int_type->data.integral.bit_count); LLVMTypeRef return_elem_types[] = { get_llvm_type(g, operand_type), LLVMInt1Type(), }; LLVMTypeRef return_struct_type = LLVMStructType(return_elem_types, 2, false); LLVMTypeRef fn_type = LLVMFunctionType(return_struct_type, param_types, 2, false); LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type); assert(LLVMGetIntrinsicID(fn_val)); return fn_val; } } static LLVMValueRef get_int_overflow_fn(CodeGen *g, ZigType *operand_type, AddSubMul add_sub_mul) { ZigType *int_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; assert(int_type->id == ZigTypeIdInt); ZigLLVMFnKey key = {}; key.id = ZigLLVMFnIdOverflowArithmetic; key.data.overflow_arithmetic.is_signed = int_type->data.integral.is_signed; key.data.overflow_arithmetic.add_sub_mul = add_sub_mul; key.data.overflow_arithmetic.bit_count = (uint32_t)int_type->data.integral.bit_count; key.data.overflow_arithmetic.vector_len = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.len : 0; auto existing_entry = g->llvm_fn_table.maybe_get(key); if (existing_entry) return existing_entry->value; LLVMValueRef fn_val; switch (add_sub_mul) { case AddSubMulAdd: fn_val = get_arithmetic_overflow_fn(g, operand_type, "sadd", "uadd"); break; case AddSubMulSub: fn_val = get_arithmetic_overflow_fn(g, operand_type, "ssub", "usub"); break; case AddSubMulMul: fn_val = get_arithmetic_overflow_fn(g, operand_type, "smul", "umul"); break; } g->llvm_fn_table.put(key, fn_val); return fn_val; } static LLVMValueRef get_float_fn(CodeGen *g, ZigType *type_entry, ZigLLVMFnId fn_id, BuiltinFnId op) { assert(type_entry->id == ZigTypeIdFloat || type_entry->id == ZigTypeIdVector); bool is_vector = (type_entry->id == ZigTypeIdVector); ZigType *float_type = is_vector ? type_entry->data.vector.elem_type : type_entry; ZigLLVMFnKey key = {}; key.id = fn_id; key.data.floating.bit_count = (uint32_t)float_type->data.floating.bit_count; key.data.floating.vector_len = is_vector ? (uint32_t)type_entry->data.vector.len : 0; key.data.floating.op = op; auto existing_entry = g->llvm_fn_table.maybe_get(key); if (existing_entry) return existing_entry->value; const char *name; uint32_t num_args; if (fn_id == ZigLLVMFnIdFMA) { name = "fma"; num_args = 3; } else if (fn_id == ZigLLVMFnIdFloatOp) { name = float_op_to_name(op); num_args = 1; } else { zig_unreachable(); } char fn_name[64]; if (is_vector) sprintf(fn_name, "llvm.%s.v%" PRIu32 "f%" PRIu32, name, key.data.floating.vector_len, key.data.floating.bit_count); else sprintf(fn_name, "llvm.%s.f%" PRIu32, name, key.data.floating.bit_count); LLVMTypeRef float_type_ref = get_llvm_type(g, type_entry); LLVMTypeRef return_elem_types[3] = { float_type_ref, float_type_ref, float_type_ref, }; LLVMTypeRef fn_type = LLVMFunctionType(float_type_ref, return_elem_types, num_args, false); LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type); assert(LLVMGetIntrinsicID(fn_val)); g->llvm_fn_table.put(key, fn_val); return fn_val; } static LLVMValueRef gen_store_untyped(CodeGen *g, LLVMValueRef value, LLVMValueRef ptr, uint32_t alignment, bool is_volatile) { LLVMValueRef instruction = LLVMBuildStore(g->builder, value, ptr); if (is_volatile) LLVMSetVolatile(instruction, true); if (alignment != 0) { LLVMSetAlignment(instruction, alignment); } return instruction; } static LLVMValueRef gen_store(CodeGen *g, LLVMValueRef value, LLVMValueRef ptr, ZigType *ptr_type) { assert(ptr_type->id == ZigTypeIdPointer); uint32_t alignment = get_ptr_align(g, ptr_type); return gen_store_untyped(g, value, ptr, alignment, ptr_type->data.pointer.is_volatile); } static LLVMValueRef gen_load_untyped(CodeGen *g, LLVMValueRef ptr, uint32_t alignment, bool is_volatile, const char *name) { LLVMValueRef result = LLVMBuildLoad(g->builder, ptr, name); if (is_volatile) LLVMSetVolatile(result, true); if (alignment == 0) { LLVMSetAlignment(result, LLVMABIAlignmentOfType(g->target_data_ref, LLVMGetElementType(LLVMTypeOf(ptr)))); } else { LLVMSetAlignment(result, alignment); } return result; } static LLVMValueRef gen_load(CodeGen *g, LLVMValueRef ptr, ZigType *ptr_type, const char *name) { assert(ptr_type->id == ZigTypeIdPointer); uint32_t alignment = get_ptr_align(g, ptr_type); return gen_load_untyped(g, ptr, alignment, ptr_type->data.pointer.is_volatile, name); } static LLVMValueRef get_handle_value(CodeGen *g, LLVMValueRef ptr, ZigType *type, ZigType *ptr_type) { if (type_has_bits(g, type)) { if (handle_is_ptr(g, type)) { return ptr; } else { assert(ptr_type->id == ZigTypeIdPointer); return gen_load(g, ptr, ptr_type, ""); } } else { return nullptr; } } static void ir_assert_impl(bool ok, IrInstGen *source_instruction, const char *file, unsigned int line) { if (ok) return; src_assert_impl(ok, source_instruction->base.source_node, file, line); } #define ir_assert(OK, SOURCE_INSTRUCTION) ir_assert_impl((OK), (SOURCE_INSTRUCTION), __FILE__, __LINE__) static bool ir_want_fast_math(CodeGen *g, IrInstGen *instruction) { // TODO memoize Scope *scope = instruction->base.scope; while (scope) { if (scope->id == ScopeIdBlock) { ScopeBlock *block_scope = (ScopeBlock *)scope; if (block_scope->fast_math_set_node) return block_scope->fast_math_on; } else if (scope->id == ScopeIdDecls) { ScopeDecls *decls_scope = (ScopeDecls *)scope; if (decls_scope->fast_math_set_node) return decls_scope->fast_math_on; } scope = scope->parent; } return false; } static bool ir_want_runtime_safety_scope(CodeGen *g, Scope *scope) { // TODO memoize while (scope) { if (scope->id == ScopeIdBlock) { ScopeBlock *block_scope = (ScopeBlock *)scope; if (block_scope->safety_set_node) return !block_scope->safety_off; } else if (scope->id == ScopeIdDecls) { ScopeDecls *decls_scope = (ScopeDecls *)scope; if (decls_scope->safety_set_node) return !decls_scope->safety_off; } scope = scope->parent; } return (g->build_mode != BuildModeFastRelease && g->build_mode != BuildModeSmallRelease); } static bool ir_want_runtime_safety(CodeGen *g, IrInstGen *instruction) { return ir_want_runtime_safety_scope(g, instruction->base.scope); } static Buf *panic_msg_buf(PanicMsgId msg_id) { switch (msg_id) { case PanicMsgIdCount: zig_unreachable(); case PanicMsgIdBoundsCheckFailure: return buf_create_from_str("index out of bounds"); case PanicMsgIdCastNegativeToUnsigned: return buf_create_from_str("attempt to cast negative value to unsigned integer"); case PanicMsgIdCastTruncatedData: return buf_create_from_str("integer cast truncated bits"); case PanicMsgIdIntegerOverflow: return buf_create_from_str("integer overflow"); case PanicMsgIdShlOverflowedBits: return buf_create_from_str("left shift overflowed bits"); case PanicMsgIdShrOverflowedBits: return buf_create_from_str("right shift overflowed bits"); case PanicMsgIdDivisionByZero: return buf_create_from_str("division by zero"); case PanicMsgIdRemainderDivisionByZero: return buf_create_from_str("remainder division by zero or negative value"); case PanicMsgIdExactDivisionRemainder: return buf_create_from_str("exact division produced remainder"); case PanicMsgIdUnwrapOptionalFail: return buf_create_from_str("attempt to use null value"); case PanicMsgIdUnreachable: return buf_create_from_str("reached unreachable code"); case PanicMsgIdInvalidErrorCode: return buf_create_from_str("invalid error code"); case PanicMsgIdIncorrectAlignment: return buf_create_from_str("incorrect alignment"); case PanicMsgIdBadUnionField: return buf_create_from_str("access of inactive union field"); case PanicMsgIdBadEnumValue: return buf_create_from_str("invalid enum value"); case PanicMsgIdFloatToInt: return buf_create_from_str("integer part of floating point value out of bounds"); case PanicMsgIdPtrCastNull: return buf_create_from_str("cast causes pointer to be null"); case PanicMsgIdBadResume: return buf_create_from_str("resumed an async function which already returned"); case PanicMsgIdBadAwait: return buf_create_from_str("async function awaited twice"); case PanicMsgIdBadReturn: return buf_create_from_str("async function returned twice"); case PanicMsgIdResumedAnAwaitingFn: return buf_create_from_str("awaiting function resumed"); case PanicMsgIdFrameTooSmall: return buf_create_from_str("frame too small"); case PanicMsgIdResumedFnPendingAwait: return buf_create_from_str("resumed an async function which can only be awaited"); case PanicMsgIdBadNoSuspendCall: return buf_create_from_str("async function called in nosuspend scope suspended"); case PanicMsgIdResumeNotSuspendedFn: return buf_create_from_str("resumed a non-suspended function"); case PanicMsgIdBadSentinel: return buf_create_from_str("sentinel mismatch"); case PanicMsgIdShxTooBigRhs: return buf_create_from_str("shift amount is greater than the type size"); } zig_unreachable(); } static LLVMValueRef get_panic_msg_ptr_val(CodeGen *g, PanicMsgId msg_id) { ZigValue *val = &g->panic_msg_vals[msg_id]; if (!val->llvm_global) { Buf *buf_msg = panic_msg_buf(msg_id); ZigValue *array_val = create_const_str_lit(g, buf_msg)->data.x_ptr.data.ref.pointee; init_const_slice(g, val, array_val, 0, buf_len(buf_msg), true); render_const_val(g, val, ""); render_const_val_global(g, val, ""); assert(val->llvm_global); } ZigType *u8_ptr_type = get_pointer_to_type_extra(g, g->builtin_types.entry_u8, true, false, PtrLenUnknown, get_abi_alignment(g, g->builtin_types.entry_u8), 0, 0, false); ZigType *str_type = get_slice_type(g, u8_ptr_type); return LLVMConstBitCast(val->llvm_global, LLVMPointerType(get_llvm_type(g, str_type), 0)); } static ZigType *ptr_to_stack_trace_type(CodeGen *g) { return get_pointer_to_type(g, get_stack_trace_type(g), false); } static void gen_panic(CodeGen *g, LLVMValueRef msg_arg, LLVMValueRef stack_trace_arg, bool stack_trace_is_llvm_alloca) { assert(g->panic_fn != nullptr); LLVMValueRef fn_val = fn_llvm_value(g, g->panic_fn); ZigLLVM_CallingConv llvm_cc = get_llvm_cc(g, g->panic_fn->type_entry->data.fn.fn_type_id.cc); if (stack_trace_arg == nullptr) { stack_trace_arg = LLVMConstNull(get_llvm_type(g, ptr_to_stack_trace_type(g))); } LLVMValueRef args[] = { msg_arg, stack_trace_arg, }; ZigLLVMBuildCall(g->builder, fn_val, args, 2, llvm_cc, ZigLLVM_CallAttrAuto, ""); if (!stack_trace_is_llvm_alloca) { // The stack trace argument is not in the stack of the caller, so // we'd like to set tail call here, but because slices (the type of msg_arg) are // still passed as pointers (see https://github.com/ziglang/zig/issues/561) we still // cannot make this a tail call. //LLVMSetTailCall(call_instruction, true); } LLVMBuildUnreachable(g->builder); } // TODO update most callsites to call gen_assertion instead of this static void gen_safety_crash(CodeGen *g, PanicMsgId msg_id) { gen_panic(g, get_panic_msg_ptr_val(g, msg_id), nullptr, false); } static void gen_assertion_scope(CodeGen *g, PanicMsgId msg_id, Scope *source_scope) { if (ir_want_runtime_safety_scope(g, source_scope)) { gen_safety_crash(g, msg_id); } else { LLVMBuildUnreachable(g->builder); } } static void gen_assertion(CodeGen *g, PanicMsgId msg_id, IrInstGen *source_instruction) { return gen_assertion_scope(g, msg_id, source_instruction->base.scope); } static LLVMValueRef gen_wasm_memory_size(CodeGen *g) { if (g->wasm_memory_size) return g->wasm_memory_size; // TODO adjust for wasm64 as well // declare i32 @llvm.wasm.memory.size.i32(i32) nounwind readonly LLVMTypeRef param_type = LLVMInt32Type(); LLVMTypeRef fn_type = LLVMFunctionType(LLVMInt32Type(), ¶m_type, 1, false); g->wasm_memory_size = LLVMAddFunction(g->module, "llvm.wasm.memory.size.i32", fn_type); assert(LLVMGetIntrinsicID(g->wasm_memory_size)); return g->wasm_memory_size; } static LLVMValueRef gen_wasm_memory_grow(CodeGen *g) { if (g->wasm_memory_grow) return g->wasm_memory_grow; // TODO adjust for wasm64 as well // declare i32 @llvm.wasm.memory.grow.i32(i32, i32) nounwind LLVMTypeRef param_types[] = { LLVMInt32Type(), LLVMInt32Type(), }; LLVMTypeRef fn_type = LLVMFunctionType(LLVMInt32Type(), param_types, 2, false); g->wasm_memory_grow = LLVMAddFunction(g->module, "llvm.wasm.memory.grow.i32", fn_type); assert(LLVMGetIntrinsicID(g->wasm_memory_grow)); return g->wasm_memory_grow; } static LLVMValueRef get_stacksave_fn_val(CodeGen *g) { if (g->stacksave_fn_val) return g->stacksave_fn_val; // declare i8* @llvm.stacksave() LLVMTypeRef fn_type = LLVMFunctionType(LLVMPointerType(LLVMInt8Type(), 0), nullptr, 0, false); g->stacksave_fn_val = LLVMAddFunction(g->module, "llvm.stacksave", fn_type); assert(LLVMGetIntrinsicID(g->stacksave_fn_val)); return g->stacksave_fn_val; } static LLVMValueRef get_stackrestore_fn_val(CodeGen *g) { if (g->stackrestore_fn_val) return g->stackrestore_fn_val; // declare void @llvm.stackrestore(i8* %ptr) LLVMTypeRef param_type = LLVMPointerType(LLVMInt8Type(), 0); LLVMTypeRef fn_type = LLVMFunctionType(LLVMVoidType(), ¶m_type, 1, false); g->stackrestore_fn_val = LLVMAddFunction(g->module, "llvm.stackrestore", fn_type); assert(LLVMGetIntrinsicID(g->stackrestore_fn_val)); return g->stackrestore_fn_val; } static LLVMValueRef get_write_register_fn_val(CodeGen *g) { if (g->write_register_fn_val) return g->write_register_fn_val; // declare void @llvm.write_register.i64(metadata, i64 @value) // !0 = !{!"sp\00"} LLVMTypeRef param_types[] = { LLVMMetadataTypeInContext(LLVMGetGlobalContext()), LLVMIntType(g->pointer_size_bytes * 8), }; LLVMTypeRef fn_type = LLVMFunctionType(LLVMVoidType(), param_types, 2, false); Buf *name = buf_sprintf("llvm.write_register.i%d", g->pointer_size_bytes * 8); g->write_register_fn_val = LLVMAddFunction(g->module, buf_ptr(name), fn_type); assert(LLVMGetIntrinsicID(g->write_register_fn_val)); return g->write_register_fn_val; } static LLVMValueRef get_return_address_fn_val(CodeGen *g) { if (g->return_address_fn_val) return g->return_address_fn_val; ZigType *return_type = get_pointer_to_type(g, g->builtin_types.entry_u8, true); LLVMTypeRef fn_type = LLVMFunctionType(get_llvm_type(g, return_type), &g->builtin_types.entry_i32->llvm_type, 1, false); g->return_address_fn_val = LLVMAddFunction(g->module, "llvm.returnaddress", fn_type); assert(LLVMGetIntrinsicID(g->return_address_fn_val)); return g->return_address_fn_val; } static LLVMValueRef get_add_error_return_trace_addr_fn(CodeGen *g) { if (g->add_error_return_trace_addr_fn_val != nullptr) return g->add_error_return_trace_addr_fn_val; LLVMTypeRef arg_types[] = { get_llvm_type(g, ptr_to_stack_trace_type(g)), g->builtin_types.entry_usize->llvm_type, }; LLVMTypeRef fn_type_ref = LLVMFunctionType(LLVMVoidType(), arg_types, 2, false); const char *fn_name = get_mangled_name(g, "__zig_add_err_ret_trace_addr"); LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type_ref); addLLVMFnAttr(fn_val, "alwaysinline"); LLVMSetLinkage(fn_val, LLVMInternalLinkage); ZigLLVMFunctionSetCallingConv(fn_val, get_llvm_cc(g, CallingConventionUnspecified)); addLLVMFnAttr(fn_val, "nounwind"); add_uwtable_attr(g, fn_val); // Error return trace memory is in the stack, which is impossible to be at address 0 // on any architecture. addLLVMArgAttr(fn_val, (unsigned)0, "nonnull"); if (codegen_have_frame_pointer(g)) { ZigLLVMAddFunctionAttr(fn_val, "frame-pointer", "all"); } LLVMBasicBlockRef entry_block = LLVMAppendBasicBlock(fn_val, "Entry"); LLVMBasicBlockRef prev_block = LLVMGetInsertBlock(g->builder); LLVMValueRef prev_debug_location = LLVMGetCurrentDebugLocation(g->builder); LLVMPositionBuilderAtEnd(g->builder, entry_block); ZigLLVMClearCurrentDebugLocation(g->builder); LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; // stack_trace.instruction_addresses[stack_trace.index & (stack_trace.instruction_addresses.len - 1)] = return_address; LLVMValueRef err_ret_trace_ptr = LLVMGetParam(fn_val, 0); LLVMValueRef address_value = LLVMGetParam(fn_val, 1); size_t index_field_index = g->stack_trace_type->data.structure.fields[0]->gen_index; LLVMValueRef index_field_ptr = LLVMBuildStructGEP(g->builder, err_ret_trace_ptr, (unsigned)index_field_index, ""); size_t addresses_field_index = g->stack_trace_type->data.structure.fields[1]->gen_index; LLVMValueRef addresses_field_ptr = LLVMBuildStructGEP(g->builder, err_ret_trace_ptr, (unsigned)addresses_field_index, ""); ZigType *slice_type = g->stack_trace_type->data.structure.fields[1]->type_entry; size_t ptr_field_index = slice_type->data.structure.fields[slice_ptr_index]->gen_index; LLVMValueRef ptr_field_ptr = LLVMBuildStructGEP(g->builder, addresses_field_ptr, (unsigned)ptr_field_index, ""); size_t len_field_index = slice_type->data.structure.fields[slice_len_index]->gen_index; LLVMValueRef len_field_ptr = LLVMBuildStructGEP(g->builder, addresses_field_ptr, (unsigned)len_field_index, ""); LLVMValueRef len_value = gen_load_untyped(g, len_field_ptr, 0, false, ""); LLVMValueRef index_val = gen_load_untyped(g, index_field_ptr, 0, false, ""); LLVMValueRef len_val_minus_one = LLVMBuildSub(g->builder, len_value, LLVMConstInt(usize_type_ref, 1, false), ""); LLVMValueRef masked_val = LLVMBuildAnd(g->builder, index_val, len_val_minus_one, ""); LLVMValueRef address_indices[] = { masked_val, }; LLVMValueRef ptr_value = gen_load_untyped(g, ptr_field_ptr, 0, false, ""); LLVMValueRef address_slot = LLVMBuildInBoundsGEP(g->builder, ptr_value, address_indices, 1, ""); gen_store_untyped(g, address_value, address_slot, 0, false); // stack_trace.index += 1; LLVMValueRef index_plus_one_val = LLVMBuildNUWAdd(g->builder, index_val, LLVMConstInt(usize_type_ref, 1, false), ""); gen_store_untyped(g, index_plus_one_val, index_field_ptr, 0, false); // return; LLVMBuildRetVoid(g->builder); LLVMPositionBuilderAtEnd(g->builder, prev_block); if (!g->strip_debug_symbols) { LLVMSetCurrentDebugLocation(g->builder, prev_debug_location); } g->add_error_return_trace_addr_fn_val = fn_val; return fn_val; } static LLVMValueRef get_return_err_fn(CodeGen *g) { if (g->return_err_fn != nullptr) return g->return_err_fn; assert(g->err_tag_type != nullptr); LLVMTypeRef arg_types[] = { // error return trace pointer get_llvm_type(g, ptr_to_stack_trace_type(g)), }; LLVMTypeRef fn_type_ref = LLVMFunctionType(LLVMVoidType(), arg_types, 1, false); const char *fn_name = get_mangled_name(g, "__zig_return_error"); LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type_ref); addLLVMFnAttr(fn_val, "noinline"); // so that we can look at return address addLLVMFnAttr(fn_val, "cold"); LLVMSetLinkage(fn_val, LLVMInternalLinkage); ZigLLVMFunctionSetCallingConv(fn_val, get_llvm_cc(g, CallingConventionUnspecified)); addLLVMFnAttr(fn_val, "nounwind"); add_uwtable_attr(g, fn_val); if (codegen_have_frame_pointer(g)) { ZigLLVMAddFunctionAttr(fn_val, "frame-pointer", "all"); } // this is above the ZigLLVMClearCurrentDebugLocation LLVMValueRef add_error_return_trace_addr_fn_val = get_add_error_return_trace_addr_fn(g); LLVMBasicBlockRef entry_block = LLVMAppendBasicBlock(fn_val, "Entry"); LLVMBasicBlockRef prev_block = LLVMGetInsertBlock(g->builder); LLVMValueRef prev_debug_location = LLVMGetCurrentDebugLocation(g->builder); LLVMPositionBuilderAtEnd(g->builder, entry_block); ZigLLVMClearCurrentDebugLocation(g->builder); LLVMValueRef err_ret_trace_ptr = LLVMGetParam(fn_val, 0); LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, g->builtin_types.entry_i32)); LLVMValueRef return_address_ptr = LLVMBuildCall(g->builder, get_return_address_fn_val(g), &zero, 1, ""); LLVMValueRef return_address = LLVMBuildPtrToInt(g->builder, return_address_ptr, usize_type_ref, ""); LLVMBasicBlockRef return_block = LLVMAppendBasicBlock(fn_val, "Return"); LLVMBasicBlockRef dest_non_null_block = LLVMAppendBasicBlock(fn_val, "DestNonNull"); LLVMValueRef null_dest_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, err_ret_trace_ptr, LLVMConstNull(LLVMTypeOf(err_ret_trace_ptr)), ""); LLVMBuildCondBr(g->builder, null_dest_bit, return_block, dest_non_null_block); LLVMPositionBuilderAtEnd(g->builder, return_block); LLVMBuildRetVoid(g->builder); LLVMPositionBuilderAtEnd(g->builder, dest_non_null_block); LLVMValueRef args[] = { err_ret_trace_ptr, return_address }; ZigLLVMBuildCall(g->builder, add_error_return_trace_addr_fn_val, args, 2, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAlwaysInline, ""); LLVMBuildRetVoid(g->builder); LLVMPositionBuilderAtEnd(g->builder, prev_block); if (!g->strip_debug_symbols) { LLVMSetCurrentDebugLocation(g->builder, prev_debug_location); } g->return_err_fn = fn_val; return fn_val; } static LLVMValueRef get_safety_crash_err_fn(CodeGen *g) { if (g->safety_crash_err_fn != nullptr) return g->safety_crash_err_fn; static const char *unwrap_err_msg_text = "attempt to unwrap error: "; g->generate_error_name_table = true; generate_error_name_table(g); assert(g->err_name_table != nullptr); // Generate the constant part of the error message LLVMValueRef msg_prefix_init = LLVMConstString(unwrap_err_msg_text, strlen(unwrap_err_msg_text), 1); LLVMValueRef msg_prefix = LLVMAddGlobal(g->module, LLVMTypeOf(msg_prefix_init), ""); LLVMSetInitializer(msg_prefix, msg_prefix_init); LLVMSetLinkage(msg_prefix, LLVMPrivateLinkage); LLVMSetGlobalConstant(msg_prefix, true); const char *fn_name = get_mangled_name(g, "__zig_fail_unwrap"); LLVMTypeRef fn_type_ref; if (g->have_err_ret_tracing) { LLVMTypeRef arg_types[] = { get_llvm_type(g, get_pointer_to_type(g, get_stack_trace_type(g), false)), get_llvm_type(g, g->err_tag_type), }; fn_type_ref = LLVMFunctionType(LLVMVoidType(), arg_types, 2, false); } else { LLVMTypeRef arg_types[] = { get_llvm_type(g, g->err_tag_type), }; fn_type_ref = LLVMFunctionType(LLVMVoidType(), arg_types, 1, false); } LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type_ref); addLLVMFnAttr(fn_val, "noreturn"); addLLVMFnAttr(fn_val, "cold"); LLVMSetLinkage(fn_val, LLVMInternalLinkage); ZigLLVMFunctionSetCallingConv(fn_val, get_llvm_cc(g, CallingConventionUnspecified)); addLLVMFnAttr(fn_val, "nounwind"); add_uwtable_attr(g, fn_val); if (codegen_have_frame_pointer(g)) { ZigLLVMAddFunctionAttr(fn_val, "frame-pointer", "all"); } // Not setting alignment here. See the comment above about // "Cannot getTypeInfo() on a type that is unsized!" // assertion failure on Darwin. LLVMBasicBlockRef entry_block = LLVMAppendBasicBlock(fn_val, "Entry"); LLVMBasicBlockRef prev_block = LLVMGetInsertBlock(g->builder); LLVMValueRef prev_debug_location = LLVMGetCurrentDebugLocation(g->builder); LLVMPositionBuilderAtEnd(g->builder, entry_block); ZigLLVMClearCurrentDebugLocation(g->builder); ZigType *usize_ty = g->builtin_types.entry_usize; ZigType *u8_ptr_type = get_pointer_to_type_extra(g, g->builtin_types.entry_u8, true, false, PtrLenUnknown, get_abi_alignment(g, g->builtin_types.entry_u8), 0, 0, false); ZigType *str_type = get_slice_type(g, u8_ptr_type); // Allocate a buffer to hold the fully-formatted error message const size_t err_buf_len = strlen(unwrap_err_msg_text) + g->largest_err_name_len; LLVMValueRef max_msg_len = LLVMConstInt(usize_ty->llvm_type, err_buf_len, 0); LLVMValueRef msg_buffer = LLVMBuildArrayAlloca(g->builder, LLVMInt8Type(), max_msg_len, "msg_buffer"); // Allocate a []u8 slice for the message LLVMValueRef msg_slice = build_alloca(g, str_type, "msg_slice", 0); LLVMValueRef err_ret_trace_arg; LLVMValueRef err_val; if (g->have_err_ret_tracing) { err_ret_trace_arg = LLVMGetParam(fn_val, 0); err_val = LLVMGetParam(fn_val, 1); } else { err_ret_trace_arg = nullptr; err_val = LLVMGetParam(fn_val, 0); } // Fetch the error name from the global table LLVMValueRef err_table_indices[] = { LLVMConstNull(usize_ty->llvm_type), err_val, }; LLVMValueRef err_name_val = LLVMBuildInBoundsGEP(g->builder, g->err_name_table, err_table_indices, 2, ""); LLVMValueRef ptr_field_ptr = LLVMBuildStructGEP(g->builder, err_name_val, slice_ptr_index, ""); LLVMValueRef err_name_ptr = gen_load_untyped(g, ptr_field_ptr, 0, false, ""); LLVMValueRef len_field_ptr = LLVMBuildStructGEP(g->builder, err_name_val, slice_len_index, ""); LLVMValueRef err_name_len = gen_load_untyped(g, len_field_ptr, 0, false, ""); LLVMValueRef msg_prefix_len = LLVMConstInt(usize_ty->llvm_type, strlen(unwrap_err_msg_text), false); // Points to the beginning of msg_buffer LLVMValueRef msg_buffer_ptr_indices[] = { LLVMConstNull(usize_ty->llvm_type), }; LLVMValueRef msg_buffer_ptr = LLVMBuildInBoundsGEP(g->builder, msg_buffer, msg_buffer_ptr_indices, 1, ""); // Points to the beginning of the constant prefix message LLVMValueRef msg_prefix_ptr_indices[] = { LLVMConstNull(usize_ty->llvm_type), }; LLVMValueRef msg_prefix_ptr = LLVMConstInBoundsGEP(msg_prefix, msg_prefix_ptr_indices, 1); // Build the message using the prefix... ZigLLVMBuildMemCpy(g->builder, msg_buffer_ptr, 1, msg_prefix_ptr, 1, msg_prefix_len, false); // ..and append the error name LLVMValueRef msg_buffer_ptr_after_indices[] = { msg_prefix_len, }; LLVMValueRef msg_buffer_ptr_after = LLVMBuildInBoundsGEP(g->builder, msg_buffer, msg_buffer_ptr_after_indices, 1, ""); ZigLLVMBuildMemCpy(g->builder, msg_buffer_ptr_after, 1, err_name_ptr, 1, err_name_len, false); // Set the slice pointer LLVMValueRef msg_slice_ptr_field_ptr = LLVMBuildStructGEP(g->builder, msg_slice, slice_ptr_index, ""); gen_store_untyped(g, msg_buffer_ptr, msg_slice_ptr_field_ptr, 0, false); // Set the slice length LLVMValueRef slice_len = LLVMBuildNUWAdd(g->builder, msg_prefix_len, err_name_len, ""); LLVMValueRef msg_slice_len_field_ptr = LLVMBuildStructGEP(g->builder, msg_slice, slice_len_index, ""); gen_store_untyped(g, slice_len, msg_slice_len_field_ptr, 0, false); // Call panic() gen_panic(g, msg_slice, err_ret_trace_arg, false); LLVMPositionBuilderAtEnd(g->builder, prev_block); if (!g->strip_debug_symbols) { LLVMSetCurrentDebugLocation(g->builder, prev_debug_location); } g->safety_crash_err_fn = fn_val; return fn_val; } static LLVMValueRef get_cur_err_ret_trace_val(CodeGen *g, Scope *scope, bool *is_llvm_alloca) { if (!g->have_err_ret_tracing) { *is_llvm_alloca = false; return nullptr; } if (g->cur_err_ret_trace_val_stack != nullptr) { *is_llvm_alloca = !fn_is_async(g->cur_fn); return g->cur_err_ret_trace_val_stack; } *is_llvm_alloca = false; return g->cur_err_ret_trace_val_arg; } static void gen_safety_crash_for_err(CodeGen *g, LLVMValueRef err_val, Scope *scope) { LLVMValueRef safety_crash_err_fn = get_safety_crash_err_fn(g); LLVMValueRef call_instruction; bool is_llvm_alloca = false; if (g->have_err_ret_tracing) { LLVMValueRef err_ret_trace_val = get_cur_err_ret_trace_val(g, scope, &is_llvm_alloca); if (err_ret_trace_val == nullptr) { err_ret_trace_val = LLVMConstNull(get_llvm_type(g, ptr_to_stack_trace_type(g))); } LLVMValueRef args[] = { err_ret_trace_val, err_val, }; call_instruction = ZigLLVMBuildCall(g->builder, safety_crash_err_fn, args, 2, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAuto, ""); } else { LLVMValueRef args[] = { err_val, }; call_instruction = ZigLLVMBuildCall(g->builder, safety_crash_err_fn, args, 1, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAuto, ""); } if (!is_llvm_alloca) { LLVMSetTailCall(call_instruction, true); } LLVMBuildUnreachable(g->builder); } static void add_bounds_check(CodeGen *g, LLVMValueRef target_val, LLVMIntPredicate lower_pred, LLVMValueRef lower_value, LLVMIntPredicate upper_pred, LLVMValueRef upper_value) { if (!lower_value && !upper_value) { return; } if (upper_value && !lower_value) { lower_value = upper_value; lower_pred = upper_pred; upper_value = nullptr; } LLVMBasicBlockRef bounds_check_fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "BoundsCheckFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "BoundsCheckOk"); LLVMBasicBlockRef lower_ok_block = upper_value ? LLVMAppendBasicBlock(g->cur_fn_val, "FirstBoundsCheckOk") : ok_block; LLVMValueRef lower_ok_val = LLVMBuildICmp(g->builder, lower_pred, target_val, lower_value, ""); LLVMBuildCondBr(g->builder, lower_ok_val, lower_ok_block, bounds_check_fail_block); LLVMPositionBuilderAtEnd(g->builder, bounds_check_fail_block); gen_safety_crash(g, PanicMsgIdBoundsCheckFailure); if (upper_value) { LLVMPositionBuilderAtEnd(g->builder, lower_ok_block); LLVMValueRef upper_ok_val = LLVMBuildICmp(g->builder, upper_pred, target_val, upper_value, ""); LLVMBuildCondBr(g->builder, upper_ok_val, ok_block, bounds_check_fail_block); } LLVMPositionBuilderAtEnd(g->builder, ok_block); } static void add_sentinel_check(CodeGen *g, LLVMValueRef sentinel_elem_ptr, ZigValue *sentinel) { LLVMValueRef expected_sentinel = gen_const_val(g, sentinel, ""); LLVMValueRef actual_sentinel = gen_load_untyped(g, sentinel_elem_ptr, 0, false, ""); LLVMValueRef ok_bit; if (sentinel->type->id == ZigTypeIdFloat) { ok_bit = LLVMBuildFCmp(g->builder, LLVMRealOEQ, actual_sentinel, expected_sentinel, ""); } else { ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, actual_sentinel, expected_sentinel, ""); } LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "SentinelFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "SentinelOk"); LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdBadSentinel); LLVMPositionBuilderAtEnd(g->builder, ok_block); } static LLVMValueRef gen_assert_zero(CodeGen *g, LLVMValueRef expr_val, ZigType *int_type) { LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, int_type)); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, expr_val, zero, ""); if (int_type->id == ZigTypeIdVector) { ok_bit = ZigLLVMBuildAndReduce(g->builder, ok_bit); } LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "CastShortenOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "CastShortenFail"); LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdCastTruncatedData); LLVMPositionBuilderAtEnd(g->builder, ok_block); return nullptr; } static LLVMValueRef gen_widen_or_shorten(CodeGen *g, bool want_runtime_safety, ZigType *actual_type, ZigType *wanted_type, LLVMValueRef expr_val) { assert(actual_type->id == wanted_type->id); assert(expr_val != nullptr); ZigType *scalar_actual_type = (actual_type->id == ZigTypeIdVector) ? actual_type->data.vector.elem_type : actual_type; ZigType *scalar_wanted_type = (wanted_type->id == ZigTypeIdVector) ? wanted_type->data.vector.elem_type : wanted_type; uint64_t actual_bits; uint64_t wanted_bits; if (scalar_actual_type->id == ZigTypeIdFloat) { actual_bits = scalar_actual_type->data.floating.bit_count; wanted_bits = scalar_wanted_type->data.floating.bit_count; } else if (scalar_actual_type->id == ZigTypeIdInt) { actual_bits = scalar_actual_type->data.integral.bit_count; wanted_bits = scalar_wanted_type->data.integral.bit_count; } else { zig_unreachable(); } if (scalar_actual_type->id == ZigTypeIdInt && want_runtime_safety && ( // negative to unsigned (!scalar_wanted_type->data.integral.is_signed && scalar_actual_type->data.integral.is_signed) || // unsigned would become negative (scalar_wanted_type->data.integral.is_signed && !scalar_actual_type->data.integral.is_signed && actual_bits == wanted_bits))) { LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, actual_type)); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntSGE, expr_val, zero, ""); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "SignCastOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "SignCastFail"); if (actual_type->id == ZigTypeIdVector) { ok_bit = ZigLLVMBuildAndReduce(g->builder, ok_bit); } LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, scalar_actual_type->data.integral.is_signed ? PanicMsgIdCastNegativeToUnsigned : PanicMsgIdCastTruncatedData); LLVMPositionBuilderAtEnd(g->builder, ok_block); } if (actual_bits == wanted_bits) { return expr_val; } else if (actual_bits < wanted_bits) { if (scalar_actual_type->id == ZigTypeIdFloat) { return LLVMBuildFPExt(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } else if (scalar_actual_type->id == ZigTypeIdInt) { if (scalar_actual_type->data.integral.is_signed) { return LLVMBuildSExt(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } else { return LLVMBuildZExt(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } } else { zig_unreachable(); } } else if (actual_bits > wanted_bits) { if (scalar_actual_type->id == ZigTypeIdFloat) { return LLVMBuildFPTrunc(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } else if (scalar_actual_type->id == ZigTypeIdInt) { if (wanted_bits == 0) { if (!want_runtime_safety) return nullptr; return gen_assert_zero(g, expr_val, actual_type); } LLVMValueRef trunc_val = LLVMBuildTrunc(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); if (!want_runtime_safety) { return trunc_val; } LLVMValueRef orig_val; if (scalar_wanted_type->data.integral.is_signed) { orig_val = LLVMBuildSExt(g->builder, trunc_val, get_llvm_type(g, actual_type), ""); } else { orig_val = LLVMBuildZExt(g->builder, trunc_val, get_llvm_type(g, actual_type), ""); } LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, expr_val, orig_val, ""); if (actual_type->id == ZigTypeIdVector) { ok_bit = ZigLLVMBuildAndReduce(g->builder, ok_bit); } LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "CastShortenOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "CastShortenFail"); LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdCastTruncatedData); LLVMPositionBuilderAtEnd(g->builder, ok_block); return trunc_val; } else { zig_unreachable(); } } else { zig_unreachable(); } } typedef LLVMValueRef (*BuildBinOpFunc)(LLVMBuilderRef, LLVMValueRef, LLVMValueRef, const char *); // These are lookup table using the AddSubMul enum as the lookup. // If AddSubMul ever changes, then these tables will be out of // date. static const BuildBinOpFunc float_op[3] = { LLVMBuildFAdd, LLVMBuildFSub, LLVMBuildFMul }; static const BuildBinOpFunc wrap_op[3] = { LLVMBuildAdd, LLVMBuildSub, LLVMBuildMul }; static const BuildBinOpFunc signed_op[3] = { LLVMBuildNSWAdd, LLVMBuildNSWSub, LLVMBuildNSWMul }; static const BuildBinOpFunc unsigned_op[3] = { LLVMBuildNUWAdd, LLVMBuildNUWSub, LLVMBuildNUWMul }; static LLVMValueRef gen_overflow_op(CodeGen *g, ZigType *operand_type, AddSubMul op, LLVMValueRef val1, LLVMValueRef val2) { LLVMValueRef overflow_bit; LLVMValueRef result; if (operand_type->id == ZigTypeIdVector) { ZigType *int_type = operand_type->data.vector.elem_type; assert(int_type->id == ZigTypeIdInt); LLVMTypeRef one_more_bit_int = LLVMIntType(int_type->data.integral.bit_count + 1); LLVMTypeRef one_more_bit_int_vector = LLVMVectorType(one_more_bit_int, operand_type->data.vector.len); const auto buildExtFn = int_type->data.integral.is_signed ? LLVMBuildSExt : LLVMBuildZExt; LLVMValueRef extended1 = buildExtFn(g->builder, val1, one_more_bit_int_vector, ""); LLVMValueRef extended2 = buildExtFn(g->builder, val2, one_more_bit_int_vector, ""); LLVMValueRef extended_result = wrap_op[op](g->builder, extended1, extended2, ""); result = LLVMBuildTrunc(g->builder, extended_result, get_llvm_type(g, operand_type), ""); LLVMValueRef re_extended_result = buildExtFn(g->builder, result, one_more_bit_int_vector, ""); LLVMValueRef overflow_vector = LLVMBuildICmp(g->builder, LLVMIntNE, extended_result, re_extended_result, ""); LLVMTypeRef bitcast_int_type = LLVMIntType(operand_type->data.vector.len); LLVMValueRef bitcasted_overflow = LLVMBuildBitCast(g->builder, overflow_vector, bitcast_int_type, ""); LLVMValueRef zero = LLVMConstNull(bitcast_int_type); overflow_bit = LLVMBuildICmp(g->builder, LLVMIntNE, bitcasted_overflow, zero, ""); } else { LLVMValueRef fn_val = get_int_overflow_fn(g, operand_type, op); LLVMValueRef params[] = { val1, val2, }; LLVMValueRef result_struct = LLVMBuildCall(g->builder, fn_val, params, 2, ""); result = LLVMBuildExtractValue(g->builder, result_struct, 0, ""); overflow_bit = LLVMBuildExtractValue(g->builder, result_struct, 1, ""); } LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "OverflowFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "OverflowOk"); LLVMBuildCondBr(g->builder, overflow_bit, fail_block, ok_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdIntegerOverflow); LLVMPositionBuilderAtEnd(g->builder, ok_block); return result; } static LLVMIntPredicate cmp_op_to_int_predicate(IrBinOp cmp_op, bool is_signed) { switch (cmp_op) { case IrBinOpCmpEq: return LLVMIntEQ; case IrBinOpCmpNotEq: return LLVMIntNE; case IrBinOpCmpLessThan: return is_signed ? LLVMIntSLT : LLVMIntULT; case IrBinOpCmpGreaterThan: return is_signed ? LLVMIntSGT : LLVMIntUGT; case IrBinOpCmpLessOrEq: return is_signed ? LLVMIntSLE : LLVMIntULE; case IrBinOpCmpGreaterOrEq: return is_signed ? LLVMIntSGE : LLVMIntUGE; default: zig_unreachable(); } } static LLVMRealPredicate cmp_op_to_real_predicate(IrBinOp cmp_op) { switch (cmp_op) { case IrBinOpCmpEq: return LLVMRealOEQ; case IrBinOpCmpNotEq: return LLVMRealUNE; case IrBinOpCmpLessThan: return LLVMRealOLT; case IrBinOpCmpGreaterThan: return LLVMRealOGT; case IrBinOpCmpLessOrEq: return LLVMRealOLE; case IrBinOpCmpGreaterOrEq: return LLVMRealOGE; default: zig_unreachable(); } } static void gen_assign_raw(CodeGen *g, LLVMValueRef ptr, ZigType *ptr_type, LLVMValueRef value) { assert(ptr_type->id == ZigTypeIdPointer); ZigType *child_type = ptr_type->data.pointer.child_type; if (!type_has_bits(g, child_type)) return; if (handle_is_ptr(g, child_type)) { assert(LLVMGetTypeKind(LLVMTypeOf(value)) == LLVMPointerTypeKind); assert(LLVMGetTypeKind(LLVMTypeOf(ptr)) == LLVMPointerTypeKind); LLVMTypeRef ptr_u8 = LLVMPointerType(LLVMInt8Type(), 0); LLVMValueRef src_ptr = LLVMBuildBitCast(g->builder, value, ptr_u8, ""); LLVMValueRef dest_ptr = LLVMBuildBitCast(g->builder, ptr, ptr_u8, ""); ZigType *usize = g->builtin_types.entry_usize; uint64_t size_bytes = LLVMStoreSizeOfType(g->target_data_ref, get_llvm_type(g, child_type)); uint64_t src_align_bytes = get_abi_alignment(g, child_type); uint64_t dest_align_bytes = get_ptr_align(g, ptr_type); assert(size_bytes > 0); assert(src_align_bytes > 0); assert(dest_align_bytes > 0); ZigLLVMBuildMemCpy(g->builder, dest_ptr, dest_align_bytes, src_ptr, src_align_bytes, LLVMConstInt(usize->llvm_type, size_bytes, false), ptr_type->data.pointer.is_volatile); return; } assert(ptr_type->data.pointer.vector_index != VECTOR_INDEX_RUNTIME); if (ptr_type->data.pointer.vector_index != VECTOR_INDEX_NONE) { LLVMValueRef index_val = LLVMConstInt(LLVMInt32Type(), ptr_type->data.pointer.vector_index, false); LLVMValueRef loaded_vector = gen_load(g, ptr, ptr_type, ""); LLVMValueRef new_vector = LLVMBuildInsertElement(g->builder, loaded_vector, value, index_val, ""); gen_store(g, new_vector, ptr, ptr_type); return; } uint32_t host_int_bytes = ptr_type->data.pointer.host_int_bytes; if (host_int_bytes == 0) { gen_store(g, value, ptr, ptr_type); return; } bool big_endian = g->is_big_endian; LLVMTypeRef int_ptr_ty = LLVMPointerType(LLVMIntType(host_int_bytes * 8), 0); LLVMValueRef int_ptr = LLVMBuildBitCast(g->builder, ptr, int_ptr_ty, ""); LLVMValueRef containing_int = gen_load(g, int_ptr, ptr_type, ""); uint32_t host_bit_count = LLVMGetIntTypeWidth(LLVMTypeOf(containing_int)); assert(host_bit_count == host_int_bytes * 8); uint32_t size_in_bits = type_size_bits(g, child_type); uint32_t bit_offset = ptr_type->data.pointer.bit_offset_in_host; uint32_t shift_amt = big_endian ? host_bit_count - bit_offset - size_in_bits : bit_offset; LLVMValueRef shift_amt_val = LLVMConstInt(LLVMTypeOf(containing_int), shift_amt, false); // Convert to equally-sized integer type in order to perform the bit // operations on the value to store LLVMTypeRef value_bits_type = LLVMIntType(size_in_bits); LLVMValueRef value_bits = LLVMBuildBitCast(g->builder, value, value_bits_type, ""); LLVMValueRef mask_val = LLVMConstAllOnes(value_bits_type); mask_val = LLVMConstZExt(mask_val, LLVMTypeOf(containing_int)); mask_val = LLVMConstShl(mask_val, shift_amt_val); mask_val = LLVMConstNot(mask_val); LLVMValueRef anded_containing_int = LLVMBuildAnd(g->builder, containing_int, mask_val, ""); LLVMValueRef extended_value = LLVMBuildZExt(g->builder, value_bits, LLVMTypeOf(containing_int), ""); LLVMValueRef shifted_value = LLVMBuildShl(g->builder, extended_value, shift_amt_val, ""); LLVMValueRef ored_value = LLVMBuildOr(g->builder, shifted_value, anded_containing_int, ""); gen_store(g, ored_value, int_ptr, ptr_type); } static void gen_var_debug_decl(CodeGen *g, ZigVar *var) { if (g->strip_debug_symbols) return; assert(var->di_loc_var != nullptr); AstNode *source_node = var->decl_node; ZigLLVMDILocation *debug_loc = ZigLLVMGetDebugLoc((unsigned)source_node->line + 1, (unsigned)source_node->column + 1, get_di_scope(g, var->parent_scope)); ZigLLVMInsertDeclareAtEnd(g->dbuilder, var->value_ref, var->di_loc_var, debug_loc, LLVMGetInsertBlock(g->builder)); } static LLVMValueRef ir_llvm_value(CodeGen *g, IrInstGen *instruction) { Error err; bool value_has_bits; if ((err = type_has_bits2(g, instruction->value->type, &value_has_bits))) codegen_report_errors_and_exit(g); if (!value_has_bits) return nullptr; if (!instruction->llvm_value) { if (instruction->id == IrInstGenIdAwait) { IrInstGenAwait *await = reinterpret_cast(instruction); if (await->result_loc != nullptr) { return get_handle_value(g, ir_llvm_value(g, await->result_loc), await->result_loc->value->type->data.pointer.child_type, await->result_loc->value->type); } } if (instruction->spill != nullptr) { ZigType *ptr_type = instruction->spill->value->type; ir_assert(ptr_type->id == ZigTypeIdPointer, instruction); return get_handle_value(g, ir_llvm_value(g, instruction->spill), ptr_type->data.pointer.child_type, instruction->spill->value->type); } ir_assert(instruction->value->special != ConstValSpecialRuntime, instruction); assert(instruction->value->type); render_const_val(g, instruction->value, ""); // we might have to do some pointer casting here due to the way union // values are rendered with a type other than the one we expect if (handle_is_ptr(g, instruction->value->type)) { render_const_val_global(g, instruction->value, ""); ZigType *ptr_type = get_pointer_to_type(g, instruction->value->type, true); instruction->llvm_value = LLVMBuildBitCast(g->builder, instruction->value->llvm_global, get_llvm_type(g, ptr_type), ""); } else { instruction->llvm_value = LLVMBuildBitCast(g->builder, instruction->value->llvm_value, get_llvm_type(g, instruction->value->type), ""); } assert(instruction->llvm_value); } return instruction->llvm_value; } void codegen_report_errors_and_exit(CodeGen *g) { // Clear progress indicator before printing errors if (g->sub_progress_node != nullptr) { stage2_progress_end(g->sub_progress_node); g->sub_progress_node = nullptr; } if (g->main_progress_node != nullptr) { stage2_progress_end(g->main_progress_node); g->main_progress_node = nullptr; } assert(g->errors.length != 0); for (size_t i = 0; i < g->errors.length; i += 1) { ErrorMsg *err = g->errors.at(i); print_err_msg(err, g->err_color); } exit(1); } static void report_errors_and_maybe_exit(CodeGen *g) { if (g->errors.length != 0) { codegen_report_errors_and_exit(g); } } ATTRIBUTE_NORETURN static void give_up_with_c_abi_error(CodeGen *g, AstNode *source_node) { ErrorMsg *msg = add_node_error(g, source_node, buf_sprintf("TODO: support C ABI for more targets. https://github.com/ziglang/zig/issues/1481")); add_error_note(g, msg, source_node, buf_sprintf("pointers, integers, floats, bools, and enums work on all targets")); codegen_report_errors_and_exit(g); } static LLVMValueRef build_alloca(CodeGen *g, ZigType *type_entry, const char *name, uint32_t alignment) { LLVMValueRef result = LLVMBuildAlloca(g->builder, get_llvm_type(g, type_entry), name); LLVMSetAlignment(result, (alignment == 0) ? get_abi_alignment(g, type_entry) : alignment); return result; } static bool iter_function_params_c_abi(CodeGen *g, ZigType *fn_type, FnWalk *fn_walk, size_t src_i) { // Initialized from the type for some walks, but because of C var args, // initialized based on callsite instructions for that one. FnTypeParamInfo *param_info = nullptr; ZigType *ty; ZigType *dest_ty = nullptr; AstNode *source_node = nullptr; LLVMValueRef val; LLVMValueRef llvm_fn; unsigned di_arg_index; ZigVar *var; switch (fn_walk->id) { case FnWalkIdAttrs: if (src_i >= fn_type->data.fn.fn_type_id.param_count) return false; param_info = &fn_type->data.fn.fn_type_id.param_info[src_i]; ty = param_info->type; source_node = fn_walk->data.attrs.fn->proto_node; llvm_fn = fn_walk->data.attrs.llvm_fn; break; case FnWalkIdCall: { if (src_i >= fn_walk->data.call.inst->arg_count) return false; IrInstGen *arg = fn_walk->data.call.inst->args[src_i]; ty = arg->value->type; source_node = arg->base.source_node; val = ir_llvm_value(g, arg); break; } case FnWalkIdTypes: if (src_i >= fn_type->data.fn.fn_type_id.param_count) return false; param_info = &fn_type->data.fn.fn_type_id.param_info[src_i]; ty = param_info->type; break; case FnWalkIdVars: assert(src_i < fn_type->data.fn.fn_type_id.param_count); param_info = &fn_type->data.fn.fn_type_id.param_info[src_i]; ty = param_info->type; var = fn_walk->data.vars.var; source_node = var->decl_node; llvm_fn = fn_walk->data.vars.llvm_fn; break; case FnWalkIdInits: if (src_i >= fn_type->data.fn.fn_type_id.param_count) return false; param_info = &fn_type->data.fn.fn_type_id.param_info[src_i]; ty = param_info->type; var = fn_walk->data.inits.fn->variable_list.at(src_i); source_node = fn_walk->data.inits.fn->proto_node; llvm_fn = fn_walk->data.inits.llvm_fn; break; } if (type_is_c_abi_int_bail(g, ty) || ty->id == ZigTypeIdFloat || ty->id == ZigTypeIdVector || ty->id == ZigTypeIdInt // TODO investigate if we need to change this ) { switch (fn_walk->id) { case FnWalkIdAttrs: { ZigType *ptr_type = get_codegen_ptr_type_bail(g, ty); if (ptr_type != nullptr) { if (type_is_nonnull_ptr(g, ty)) { addLLVMArgAttr(llvm_fn, fn_walk->data.attrs.gen_i, "nonnull"); } if (ptr_type->id == ZigTypeIdPointer && ptr_type->data.pointer.is_const) { addLLVMArgAttr(llvm_fn, fn_walk->data.attrs.gen_i, "readonly"); } if (param_info->is_noalias) { addLLVMArgAttr(llvm_fn, fn_walk->data.attrs.gen_i, "noalias"); } } fn_walk->data.attrs.gen_i += 1; break; } case FnWalkIdCall: fn_walk->data.call.gen_param_values->append(val); break; case FnWalkIdTypes: fn_walk->data.types.gen_param_types->append(get_llvm_type(g, ty)); fn_walk->data.types.param_di_types->append(get_llvm_di_type(g, ty)); break; case FnWalkIdVars: { var->value_ref = build_alloca(g, ty, var->name, var->align_bytes); di_arg_index = fn_walk->data.vars.gen_i; fn_walk->data.vars.gen_i += 1; dest_ty = ty; goto var_ok; } case FnWalkIdInits: clear_debug_source_node(g); gen_store_untyped(g, LLVMGetParam(llvm_fn, fn_walk->data.inits.gen_i), var->value_ref, var->align_bytes, false); if (var->decl_node) { gen_var_debug_decl(g, var); } fn_walk->data.inits.gen_i += 1; break; } return true; } { // Arrays are just pointers if (ty->id == ZigTypeIdArray) { assert(handle_is_ptr(g, ty)); switch (fn_walk->id) { case FnWalkIdAttrs: // arrays passed to C ABI functions may not be at address 0 addLLVMArgAttr(llvm_fn, fn_walk->data.attrs.gen_i, "nonnull"); addLLVMArgAttrInt(llvm_fn, fn_walk->data.attrs.gen_i, "align", get_abi_alignment(g, ty)); fn_walk->data.attrs.gen_i += 1; break; case FnWalkIdCall: fn_walk->data.call.gen_param_values->append(val); break; case FnWalkIdTypes: { ZigType *gen_type = get_pointer_to_type(g, ty, true); fn_walk->data.types.gen_param_types->append(get_llvm_type(g, gen_type)); fn_walk->data.types.param_di_types->append(get_llvm_di_type(g, gen_type)); break; } case FnWalkIdVars: { var->value_ref = LLVMGetParam(llvm_fn, fn_walk->data.vars.gen_i); di_arg_index = fn_walk->data.vars.gen_i; dest_ty = get_pointer_to_type(g, ty, false); fn_walk->data.vars.gen_i += 1; goto var_ok; } case FnWalkIdInits: if (var->decl_node) { gen_var_debug_decl(g, var); } fn_walk->data.inits.gen_i += 1; break; } return true; } X64CABIClass abi_class = type_c_abi_x86_64_class(g, ty); size_t ty_size = type_size(g, ty); if (abi_class == X64CABIClass_MEMORY || abi_class == X64CABIClass_MEMORY_nobyval) { assert(handle_is_ptr(g, ty)); switch (fn_walk->id) { case FnWalkIdAttrs: if (abi_class != X64CABIClass_MEMORY_nobyval) { ZigLLVMAddByValAttr(llvm_fn, fn_walk->data.attrs.gen_i + 1, get_llvm_type(g, ty)); addLLVMArgAttrInt(llvm_fn, fn_walk->data.attrs.gen_i, "align", get_abi_alignment(g, ty)); } else if (g->zig_target->arch == ZigLLVM_aarch64 || g->zig_target->arch == ZigLLVM_aarch64_be) { // no attrs needed } else { if (source_node != nullptr) { give_up_with_c_abi_error(g, source_node); } // otherwise allow codegen code to report a compile error return false; } // Byvalue parameters must not have address 0 addLLVMArgAttr(llvm_fn, fn_walk->data.attrs.gen_i, "nonnull"); fn_walk->data.attrs.gen_i += 1; break; case FnWalkIdCall: fn_walk->data.call.gen_param_values->append(val); break; case FnWalkIdTypes: { ZigType *gen_type = get_pointer_to_type(g, ty, true); fn_walk->data.types.gen_param_types->append(get_llvm_type(g, gen_type)); fn_walk->data.types.param_di_types->append(get_llvm_di_type(g, gen_type)); break; } case FnWalkIdVars: { di_arg_index = fn_walk->data.vars.gen_i; var->value_ref = LLVMGetParam(llvm_fn, fn_walk->data.vars.gen_i); dest_ty = get_pointer_to_type(g, ty, false); fn_walk->data.vars.gen_i += 1; goto var_ok; } case FnWalkIdInits: if (var->decl_node) { gen_var_debug_decl(g, var); } fn_walk->data.inits.gen_i += 1; break; } return true; } else if (abi_class == X64CABIClass_INTEGER) { switch (fn_walk->id) { case FnWalkIdAttrs: fn_walk->data.attrs.gen_i += 1; break; case FnWalkIdCall: { LLVMTypeRef ptr_to_int_type_ref = LLVMPointerType(LLVMIntType((unsigned)ty_size * 8), 0); LLVMValueRef bitcasted = LLVMBuildBitCast(g->builder, val, ptr_to_int_type_ref, ""); LLVMValueRef loaded = LLVMBuildLoad(g->builder, bitcasted, ""); fn_walk->data.call.gen_param_values->append(loaded); break; } case FnWalkIdTypes: { ZigType *gen_type = get_int_type(g, false, ty_size * 8); fn_walk->data.types.gen_param_types->append(get_llvm_type(g, gen_type)); fn_walk->data.types.param_di_types->append(get_llvm_di_type(g, gen_type)); break; } case FnWalkIdVars: { di_arg_index = fn_walk->data.vars.gen_i; var->value_ref = build_alloca(g, ty, var->name, var->align_bytes); fn_walk->data.vars.gen_i += 1; dest_ty = ty; goto var_ok; } case FnWalkIdInits: { clear_debug_source_node(g); if (!fn_is_async(fn_walk->data.inits.fn)) { LLVMValueRef arg = LLVMGetParam(llvm_fn, fn_walk->data.inits.gen_i); LLVMTypeRef ptr_to_int_type_ref = LLVMPointerType(LLVMIntType((unsigned)ty_size * 8), 0); LLVMValueRef bitcasted = LLVMBuildBitCast(g->builder, var->value_ref, ptr_to_int_type_ref, ""); gen_store_untyped(g, arg, bitcasted, var->align_bytes, false); } if (var->decl_node) { gen_var_debug_decl(g, var); } fn_walk->data.inits.gen_i += 1; break; } } return true; } else if (abi_class == X64CABIClass_SSE) { // For now only handle structs with only floats/doubles in it. if (ty->id != ZigTypeIdStruct) { if (source_node != nullptr) { give_up_with_c_abi_error(g, source_node); } // otherwise allow codegen code to report a compile error return false; } for (uint32_t i = 0; i < ty->data.structure.src_field_count; i += 1) { if (ty->data.structure.fields[i]->type_entry->id != ZigTypeIdFloat) { if (source_node != nullptr) { give_up_with_c_abi_error(g, source_node); } // otherwise allow codegen code to report a compile error return false; } } // The SystemV ABI says that we have to setup 1 FP register per f64. // So two f32 can be passed in one f64, but 3 f32 have to be passed in 2 FP registers. // To achieve this with LLVM API, we pass multiple f64 parameters to the LLVM function if // the type is bigger than 8 bytes. // Example: // extern struct { // x: f32, // y: f32, // z: f32, // }; // const ptr = (*f64)*Struct; // Register 1: ptr.* // Register 2: (ptr + 1).* // One floating point register per f64 or 2 f32's size_t number_of_fp_regs = (size_t)ceilf((float)ty_size / (float)8); switch (fn_walk->id) { case FnWalkIdAttrs: { fn_walk->data.attrs.gen_i += 1; break; } case FnWalkIdCall: { LLVMValueRef f64_ptr_to_struct = LLVMBuildBitCast(g->builder, val, LLVMPointerType(LLVMDoubleType(), 0), ""); for (uint32_t i = 0; i < number_of_fp_regs; i += 1) { LLVMValueRef index = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, i, false); LLVMValueRef indices[] = { index }; LLVMValueRef adjusted_ptr_to_struct = LLVMBuildInBoundsGEP(g->builder, f64_ptr_to_struct, indices, 1, ""); LLVMValueRef loaded = LLVMBuildLoad(g->builder, adjusted_ptr_to_struct, ""); fn_walk->data.call.gen_param_values->append(loaded); } break; } case FnWalkIdTypes: { for (uint32_t i = 0; i < number_of_fp_regs; i += 1) { fn_walk->data.types.gen_param_types->append(get_llvm_type(g, g->builtin_types.entry_f64)); fn_walk->data.types.param_di_types->append(get_llvm_di_type(g, g->builtin_types.entry_f64)); } break; } case FnWalkIdVars: case FnWalkIdInits: { // TODO: Handle exporting functions if (source_node != nullptr) { give_up_with_c_abi_error(g, source_node); } // otherwise allow codegen code to report a compile error return false; } } return true; } } if (source_node != nullptr) { give_up_with_c_abi_error(g, source_node); } // otherwise allow codegen code to report a compile error return false; var_ok: if (dest_ty != nullptr && var->decl_node) { // arg index + 1 because the 0 index is return value var->di_loc_var = ZigLLVMCreateParameterVariable(g->dbuilder, get_di_scope(g, var->parent_scope), var->name, fn_walk->data.vars.import->data.structure.root_struct->di_file, (unsigned)(var->decl_node->line + 1), get_llvm_di_type(g, dest_ty), !g->strip_debug_symbols, 0, di_arg_index + 1); } return true; } void walk_function_params(CodeGen *g, ZigType *fn_type, FnWalk *fn_walk) { CallingConvention cc = fn_type->data.fn.fn_type_id.cc; if (!calling_convention_allows_zig_types(cc)) { size_t src_i = 0; for (;;) { if (!iter_function_params_c_abi(g, fn_type, fn_walk, src_i)) break; src_i += 1; } return; } if (fn_walk->id == FnWalkIdCall) { IrInstGenCall *instruction = fn_walk->data.call.inst; bool is_var_args = fn_walk->data.call.is_var_args; for (size_t call_i = 0; call_i < instruction->arg_count; call_i += 1) { IrInstGen *param_instruction = instruction->args[call_i]; ZigType *param_type = param_instruction->value->type; if (is_var_args || type_has_bits(g, param_type)) { LLVMValueRef param_value = ir_llvm_value(g, param_instruction); assert(param_value); fn_walk->data.call.gen_param_values->append(param_value); fn_walk->data.call.gen_param_types->append(param_type); } } return; } size_t next_var_i = 0; for (size_t param_i = 0; param_i < fn_type->data.fn.fn_type_id.param_count; param_i += 1) { FnGenParamInfo *gen_info = &fn_type->data.fn.gen_param_info[param_i]; size_t gen_index = gen_info->gen_index; if (gen_index == SIZE_MAX) { continue; } switch (fn_walk->id) { case FnWalkIdAttrs: { LLVMValueRef llvm_fn = fn_walk->data.attrs.llvm_fn; bool is_byval = gen_info->is_byval; FnTypeParamInfo *param_info = &fn_type->data.fn.fn_type_id.param_info[param_i]; ZigType *param_type = gen_info->type; if (param_info->is_noalias) { addLLVMArgAttr(llvm_fn, (unsigned)gen_index, "noalias"); } if ((param_type->id == ZigTypeIdPointer && param_type->data.pointer.is_const) || is_byval) { addLLVMArgAttr(llvm_fn, (unsigned)gen_index, "readonly"); } if (get_codegen_ptr_type_bail(g, param_type) != nullptr) { addLLVMArgAttrInt(llvm_fn, (unsigned)gen_index, "align", get_ptr_align(g, param_type)); } if (type_is_nonnull_ptr(g, param_type)) { addLLVMArgAttr(llvm_fn, (unsigned)gen_index, "nonnull"); } break; } case FnWalkIdInits: { ZigFn *fn_table_entry = fn_walk->data.inits.fn; LLVMValueRef llvm_fn = fn_table_entry->llvm_value; ZigVar *variable = fn_table_entry->variable_list.at(next_var_i); assert(variable->src_arg_index != SIZE_MAX); next_var_i += 1; assert(variable); assert(variable->value_ref); if (!handle_is_ptr(g, variable->var_type) && !fn_is_async(fn_walk->data.inits.fn)) { clear_debug_source_node(g); ZigType *fn_type = fn_table_entry->type_entry; unsigned gen_arg_index = fn_type->data.fn.gen_param_info[variable->src_arg_index].gen_index; gen_store_untyped(g, LLVMGetParam(llvm_fn, gen_arg_index), variable->value_ref, variable->align_bytes, false); } if (variable->decl_node) { gen_var_debug_decl(g, variable); } break; } case FnWalkIdCall: // handled before for loop zig_unreachable(); case FnWalkIdTypes: // Not called for non-c-abi zig_unreachable(); case FnWalkIdVars: // iter_function_params_c_abi is called directly for this one zig_unreachable(); } } } static LLVMValueRef get_merge_err_ret_traces_fn_val(CodeGen *g) { if (g->merge_err_ret_traces_fn_val) return g->merge_err_ret_traces_fn_val; assert(g->stack_trace_type != nullptr); LLVMTypeRef param_types[] = { get_llvm_type(g, ptr_to_stack_trace_type(g)), get_llvm_type(g, ptr_to_stack_trace_type(g)), }; LLVMTypeRef fn_type_ref = LLVMFunctionType(LLVMVoidType(), param_types, 2, false); const char *fn_name = get_mangled_name(g, "__zig_merge_error_return_traces"); LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type_ref); LLVMSetLinkage(fn_val, LLVMInternalLinkage); ZigLLVMFunctionSetCallingConv(fn_val, get_llvm_cc(g, CallingConventionUnspecified)); addLLVMFnAttr(fn_val, "nounwind"); add_uwtable_attr(g, fn_val); addLLVMArgAttr(fn_val, (unsigned)0, "noalias"); addLLVMArgAttr(fn_val, (unsigned)0, "writeonly"); addLLVMArgAttr(fn_val, (unsigned)1, "noalias"); addLLVMArgAttr(fn_val, (unsigned)1, "readonly"); if (codegen_have_frame_pointer(g)) { ZigLLVMAddFunctionAttr(fn_val, "frame-pointer", "all"); } // this is above the ZigLLVMClearCurrentDebugLocation LLVMValueRef add_error_return_trace_addr_fn_val = get_add_error_return_trace_addr_fn(g); LLVMBasicBlockRef entry_block = LLVMAppendBasicBlock(fn_val, "Entry"); LLVMBasicBlockRef prev_block = LLVMGetInsertBlock(g->builder); LLVMValueRef prev_debug_location = LLVMGetCurrentDebugLocation(g->builder); LLVMPositionBuilderAtEnd(g->builder, entry_block); ZigLLVMClearCurrentDebugLocation(g->builder); // if (dest_stack_trace == null or src_stack_trace == null) return; // var frame_index: usize = undefined; // var frames_left: usize = undefined; // if (src_stack_trace.index < src_stack_trace.instruction_addresses.len) { // frame_index = 0; // frames_left = src_stack_trace.index; // if (frames_left == 0) return; // } else { // frame_index = (src_stack_trace.index + 1) % src_stack_trace.instruction_addresses.len; // frames_left = src_stack_trace.instruction_addresses.len; // } // while (true) { // __zig_add_err_ret_trace_addr(dest_stack_trace, src_stack_trace.instruction_addresses[frame_index]); // frames_left -= 1; // if (frames_left == 0) return; // frame_index = (frame_index + 1) % src_stack_trace.instruction_addresses.len; // } LLVMBasicBlockRef return_block = LLVMAppendBasicBlock(fn_val, "Return"); LLVMBasicBlockRef non_null_block = LLVMAppendBasicBlock(fn_val, "NonNull"); LLVMValueRef frame_index_ptr = LLVMBuildAlloca(g->builder, g->builtin_types.entry_usize->llvm_type, "frame_index"); LLVMValueRef frames_left_ptr = LLVMBuildAlloca(g->builder, g->builtin_types.entry_usize->llvm_type, "frames_left"); LLVMValueRef dest_stack_trace_ptr = LLVMGetParam(fn_val, 0); LLVMValueRef src_stack_trace_ptr = LLVMGetParam(fn_val, 1); LLVMValueRef null_dest_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, dest_stack_trace_ptr, LLVMConstNull(LLVMTypeOf(dest_stack_trace_ptr)), ""); LLVMValueRef null_src_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, src_stack_trace_ptr, LLVMConstNull(LLVMTypeOf(src_stack_trace_ptr)), ""); LLVMValueRef null_bit = LLVMBuildOr(g->builder, null_dest_bit, null_src_bit, ""); LLVMBuildCondBr(g->builder, null_bit, return_block, non_null_block); LLVMPositionBuilderAtEnd(g->builder, non_null_block); size_t src_index_field_index = g->stack_trace_type->data.structure.fields[0]->gen_index; size_t src_addresses_field_index = g->stack_trace_type->data.structure.fields[1]->gen_index; LLVMValueRef src_index_field_ptr = LLVMBuildStructGEP(g->builder, src_stack_trace_ptr, (unsigned)src_index_field_index, ""); LLVMValueRef src_addresses_field_ptr = LLVMBuildStructGEP(g->builder, src_stack_trace_ptr, (unsigned)src_addresses_field_index, ""); ZigType *slice_type = g->stack_trace_type->data.structure.fields[1]->type_entry; size_t ptr_field_index = slice_type->data.structure.fields[slice_ptr_index]->gen_index; LLVMValueRef src_ptr_field_ptr = LLVMBuildStructGEP(g->builder, src_addresses_field_ptr, (unsigned)ptr_field_index, ""); size_t len_field_index = slice_type->data.structure.fields[slice_len_index]->gen_index; LLVMValueRef src_len_field_ptr = LLVMBuildStructGEP(g->builder, src_addresses_field_ptr, (unsigned)len_field_index, ""); LLVMValueRef src_index_val = LLVMBuildLoad(g->builder, src_index_field_ptr, ""); LLVMValueRef src_ptr_val = LLVMBuildLoad(g->builder, src_ptr_field_ptr, ""); LLVMValueRef src_len_val = LLVMBuildLoad(g->builder, src_len_field_ptr, ""); LLVMValueRef no_wrap_bit = LLVMBuildICmp(g->builder, LLVMIntULT, src_index_val, src_len_val, ""); LLVMBasicBlockRef no_wrap_block = LLVMAppendBasicBlock(fn_val, "NoWrap"); LLVMBasicBlockRef yes_wrap_block = LLVMAppendBasicBlock(fn_val, "YesWrap"); LLVMBasicBlockRef loop_block = LLVMAppendBasicBlock(fn_val, "Loop"); LLVMBuildCondBr(g->builder, no_wrap_bit, no_wrap_block, yes_wrap_block); LLVMPositionBuilderAtEnd(g->builder, no_wrap_block); LLVMValueRef usize_zero = LLVMConstNull(g->builtin_types.entry_usize->llvm_type); LLVMBuildStore(g->builder, usize_zero, frame_index_ptr); LLVMBuildStore(g->builder, src_index_val, frames_left_ptr); LLVMValueRef frames_left_eq_zero_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, src_index_val, usize_zero, ""); LLVMBuildCondBr(g->builder, frames_left_eq_zero_bit, return_block, loop_block); LLVMPositionBuilderAtEnd(g->builder, yes_wrap_block); LLVMValueRef usize_one = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, 1, false); LLVMValueRef plus_one = LLVMBuildNUWAdd(g->builder, src_index_val, usize_one, ""); LLVMValueRef mod_len = LLVMBuildURem(g->builder, plus_one, src_len_val, ""); LLVMBuildStore(g->builder, mod_len, frame_index_ptr); LLVMBuildStore(g->builder, src_len_val, frames_left_ptr); LLVMBuildBr(g->builder, loop_block); LLVMPositionBuilderAtEnd(g->builder, loop_block); LLVMValueRef ptr_index = LLVMBuildLoad(g->builder, frame_index_ptr, ""); LLVMValueRef addr_ptr = LLVMBuildInBoundsGEP(g->builder, src_ptr_val, &ptr_index, 1, ""); LLVMValueRef this_addr_val = LLVMBuildLoad(g->builder, addr_ptr, ""); LLVMValueRef args[] = {dest_stack_trace_ptr, this_addr_val}; ZigLLVMBuildCall(g->builder, add_error_return_trace_addr_fn_val, args, 2, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAlwaysInline, ""); LLVMValueRef prev_frames_left = LLVMBuildLoad(g->builder, frames_left_ptr, ""); LLVMValueRef new_frames_left = LLVMBuildNUWSub(g->builder, prev_frames_left, usize_one, ""); LLVMValueRef done_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, new_frames_left, usize_zero, ""); LLVMBasicBlockRef continue_block = LLVMAppendBasicBlock(fn_val, "Continue"); LLVMBuildCondBr(g->builder, done_bit, return_block, continue_block); LLVMPositionBuilderAtEnd(g->builder, return_block); LLVMBuildRetVoid(g->builder); LLVMPositionBuilderAtEnd(g->builder, continue_block); LLVMBuildStore(g->builder, new_frames_left, frames_left_ptr); LLVMValueRef prev_index = LLVMBuildLoad(g->builder, frame_index_ptr, ""); LLVMValueRef index_plus_one = LLVMBuildNUWAdd(g->builder, prev_index, usize_one, ""); LLVMValueRef index_mod_len = LLVMBuildURem(g->builder, index_plus_one, src_len_val, ""); LLVMBuildStore(g->builder, index_mod_len, frame_index_ptr); LLVMBuildBr(g->builder, loop_block); LLVMPositionBuilderAtEnd(g->builder, prev_block); if (!g->strip_debug_symbols) { LLVMSetCurrentDebugLocation(g->builder, prev_debug_location); } g->merge_err_ret_traces_fn_val = fn_val; return fn_val; } static LLVMValueRef ir_render_save_err_ret_addr(CodeGen *g, IrExecutableGen *executable, IrInstGenSaveErrRetAddr *save_err_ret_addr_instruction) { assert(g->have_err_ret_tracing); LLVMValueRef return_err_fn = get_return_err_fn(g); bool is_llvm_alloca; LLVMValueRef my_err_trace_val = get_cur_err_ret_trace_val(g, save_err_ret_addr_instruction->base.base.scope, &is_llvm_alloca); ZigLLVMBuildCall(g->builder, return_err_fn, &my_err_trace_val, 1, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAuto, ""); ZigType *ret_type = g->cur_fn->type_entry->data.fn.fn_type_id.return_type; if (fn_is_async(g->cur_fn) && codegen_fn_has_err_ret_tracing_arg(g, ret_type)) { LLVMValueRef trace_ptr_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, frame_index_trace_arg(g, ret_type), ""); LLVMBuildStore(g->builder, my_err_trace_val, trace_ptr_ptr); } return nullptr; } static void gen_assert_resume_id(CodeGen *g, IrInstGen *source_instr, ResumeId resume_id, PanicMsgId msg_id, LLVMBasicBlockRef end_bb) { LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; if (ir_want_runtime_safety(g, source_instr)) { // Write a value to the resume index which indicates the function was resumed while not suspended. LLVMBuildStore(g->builder, g->cur_bad_not_suspended_index, g->cur_async_resume_index_ptr); } LLVMBasicBlockRef bad_resume_block = LLVMAppendBasicBlock(g->cur_fn_val, "BadResume"); if (end_bb == nullptr) end_bb = LLVMAppendBasicBlock(g->cur_fn_val, "OkResume"); LLVMValueRef expected_value = LLVMConstSub(LLVMConstAllOnes(usize_type_ref), LLVMConstInt(usize_type_ref, resume_id, false)); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, LLVMGetParam(g->cur_fn_val, 1), expected_value, ""); LLVMBuildCondBr(g->builder, ok_bit, end_bb, bad_resume_block); LLVMPositionBuilderAtEnd(g->builder, bad_resume_block); gen_assertion(g, msg_id, source_instr); LLVMPositionBuilderAtEnd(g->builder, end_bb); } static LLVMValueRef gen_resume(CodeGen *g, LLVMValueRef fn_val, LLVMValueRef target_frame_ptr, ResumeId resume_id) { LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; if (fn_val == nullptr) { LLVMValueRef fn_ptr_ptr = LLVMBuildStructGEP(g->builder, target_frame_ptr, frame_fn_ptr_index, ""); fn_val = LLVMBuildLoad(g->builder, fn_ptr_ptr, ""); } LLVMValueRef arg_val = LLVMConstSub(LLVMConstAllOnes(usize_type_ref), LLVMConstInt(usize_type_ref, resume_id, false)); LLVMValueRef args[] = {target_frame_ptr, arg_val}; return ZigLLVMBuildCall(g->builder, fn_val, args, 2, ZigLLVM_Fast, ZigLLVM_CallAttrAuto, ""); } static LLVMBasicBlockRef gen_suspend_begin(CodeGen *g, const char *name_hint) { LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMBasicBlockRef resume_bb = LLVMAppendBasicBlock(g->cur_fn_val, name_hint); size_t new_block_index = g->cur_resume_block_count; g->cur_resume_block_count += 1; LLVMValueRef new_block_index_val = LLVMConstInt(usize_type_ref, new_block_index, false); LLVMAddCase(g->cur_async_switch_instr, new_block_index_val, resume_bb); LLVMBuildStore(g->builder, new_block_index_val, g->cur_async_resume_index_ptr); return resume_bb; } // Be careful setting tail call. According to LLVM lang ref, // tail and musttail imply that the callee does not access allocas from the caller. // This works for async functions since the locals are spilled. // http://llvm.org/docs/LangRef.html#id320 static void set_tail_call_if_appropriate(CodeGen *g, LLVMValueRef call_inst) { LLVMSetTailCall(call_inst, true); } static LLVMValueRef gen_maybe_atomic_op(CodeGen *g, LLVMAtomicRMWBinOp op, LLVMValueRef ptr, LLVMValueRef val, LLVMAtomicOrdering order) { if (g->is_single_threaded) { LLVMValueRef loaded = LLVMBuildLoad(g->builder, ptr, ""); LLVMValueRef modified; switch (op) { case LLVMAtomicRMWBinOpXchg: modified = val; break; case LLVMAtomicRMWBinOpXor: modified = LLVMBuildXor(g->builder, loaded, val, ""); break; default: zig_unreachable(); } LLVMBuildStore(g->builder, modified, ptr); return loaded; } else { return LLVMBuildAtomicRMW(g->builder, op, ptr, val, order, false); } } static void gen_async_return(CodeGen *g, IrInstGenReturn *instruction) { LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; ZigType *operand_type = (instruction->operand != nullptr) ? instruction->operand->value->type : nullptr; bool operand_has_bits = (operand_type != nullptr) && type_has_bits(g, operand_type); ZigType *ret_type = g->cur_fn->type_entry->data.fn.fn_type_id.return_type; bool ret_type_has_bits = type_has_bits(g, ret_type); if (operand_has_bits && instruction->operand != nullptr) { bool need_store = instruction->operand->value->special != ConstValSpecialRuntime || !handle_is_ptr(g, ret_type); if (need_store) { // It didn't get written to the result ptr. We do that now. ZigType *ret_ptr_type = get_pointer_to_type(g, ret_type, true); gen_assign_raw(g, g->cur_ret_ptr, ret_ptr_type, ir_llvm_value(g, instruction->operand)); } } // Whether we tail resume the awaiter, or do an early return, we are done and will not be resumed. if (ir_want_runtime_safety(g, &instruction->base)) { LLVMValueRef new_resume_index = LLVMConstAllOnes(usize_type_ref); LLVMBuildStore(g->builder, new_resume_index, g->cur_async_resume_index_ptr); } LLVMValueRef zero = LLVMConstNull(usize_type_ref); LLVMValueRef all_ones = LLVMConstAllOnes(usize_type_ref); LLVMValueRef prev_val = gen_maybe_atomic_op(g, LLVMAtomicRMWBinOpXor, g->cur_async_awaiter_ptr, all_ones, LLVMAtomicOrderingAcquire); LLVMBasicBlockRef bad_return_block = LLVMAppendBasicBlock(g->cur_fn_val, "BadReturn"); LLVMBasicBlockRef early_return_block = LLVMAppendBasicBlock(g->cur_fn_val, "EarlyReturn"); LLVMBasicBlockRef resume_them_block = LLVMAppendBasicBlock(g->cur_fn_val, "ResumeThem"); LLVMValueRef switch_instr = LLVMBuildSwitch(g->builder, prev_val, resume_them_block, 2); LLVMAddCase(switch_instr, zero, early_return_block); LLVMAddCase(switch_instr, all_ones, bad_return_block); // Something has gone horribly wrong, and this is an invalid second return. LLVMPositionBuilderAtEnd(g->builder, bad_return_block); gen_assertion(g, PanicMsgIdBadReturn, &instruction->base); // There is no awaiter yet, but we're completely done. LLVMPositionBuilderAtEnd(g->builder, early_return_block); LLVMBuildRetVoid(g->builder); // We need to resume the caller by tail calling them, // but first write through the result pointer and possibly // error return trace pointer. LLVMPositionBuilderAtEnd(g->builder, resume_them_block); if (ret_type_has_bits) { // If the awaiter result pointer is non-null, we need to copy the result to there. LLVMBasicBlockRef copy_block = LLVMAppendBasicBlock(g->cur_fn_val, "CopyResult"); LLVMBasicBlockRef copy_end_block = LLVMAppendBasicBlock(g->cur_fn_val, "CopyResultEnd"); LLVMValueRef awaiter_ret_ptr_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, frame_ret_start + 1, ""); LLVMValueRef awaiter_ret_ptr = LLVMBuildLoad(g->builder, awaiter_ret_ptr_ptr, ""); LLVMValueRef zero_ptr = LLVMConstNull(LLVMTypeOf(awaiter_ret_ptr)); LLVMValueRef need_copy_bit = LLVMBuildICmp(g->builder, LLVMIntNE, awaiter_ret_ptr, zero_ptr, ""); LLVMBuildCondBr(g->builder, need_copy_bit, copy_block, copy_end_block); LLVMPositionBuilderAtEnd(g->builder, copy_block); LLVMTypeRef ptr_u8 = LLVMPointerType(LLVMInt8Type(), 0); LLVMValueRef dest_ptr_casted = LLVMBuildBitCast(g->builder, awaiter_ret_ptr, ptr_u8, ""); LLVMValueRef src_ptr_casted = LLVMBuildBitCast(g->builder, g->cur_ret_ptr, ptr_u8, ""); bool is_volatile = false; uint32_t abi_align = get_abi_alignment(g, ret_type); LLVMValueRef byte_count_val = LLVMConstInt(usize_type_ref, type_size(g, ret_type), false); ZigLLVMBuildMemCpy(g->builder, dest_ptr_casted, abi_align, src_ptr_casted, abi_align, byte_count_val, is_volatile); LLVMBuildBr(g->builder, copy_end_block); LLVMPositionBuilderAtEnd(g->builder, copy_end_block); if (codegen_fn_has_err_ret_tracing_arg(g, ret_type)) { LLVMValueRef awaiter_trace_ptr_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, frame_index_trace_arg(g, ret_type) + 1, ""); LLVMValueRef dest_trace_ptr = LLVMBuildLoad(g->builder, awaiter_trace_ptr_ptr, ""); bool is_llvm_alloca; LLVMValueRef my_err_trace_val = get_cur_err_ret_trace_val(g, instruction->base.base.scope, &is_llvm_alloca); LLVMValueRef args[] = { dest_trace_ptr, my_err_trace_val }; ZigLLVMBuildCall(g->builder, get_merge_err_ret_traces_fn_val(g), args, 2, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAuto, ""); } } // Resume the caller by tail calling them. ZigType *any_frame_type = get_any_frame_type(g, ret_type); LLVMValueRef their_frame_ptr = LLVMBuildIntToPtr(g->builder, prev_val, get_llvm_type(g, any_frame_type), ""); LLVMValueRef call_inst = gen_resume(g, nullptr, their_frame_ptr, ResumeIdReturn); set_tail_call_if_appropriate(g, call_inst); LLVMBuildRetVoid(g->builder); } static LLVMValueRef ir_render_return(CodeGen *g, IrExecutableGen *executable, IrInstGenReturn *instruction) { if (fn_is_async(g->cur_fn)) { gen_async_return(g, instruction); return nullptr; } if (want_first_arg_sret(g, &g->cur_fn->type_entry->data.fn.fn_type_id)) { if (instruction->operand == nullptr) { LLVMBuildRetVoid(g->builder); return nullptr; } assert(g->cur_ret_ptr); ir_assert(instruction->operand->value->special != ConstValSpecialRuntime, &instruction->base); LLVMValueRef value = ir_llvm_value(g, instruction->operand); ZigType *return_type = instruction->operand->value->type; gen_assign_raw(g, g->cur_ret_ptr, get_pointer_to_type(g, return_type, false), value); LLVMBuildRetVoid(g->builder); } else if (g->cur_fn->type_entry->data.fn.fn_type_id.cc != CallingConventionAsync && handle_is_ptr(g, g->cur_fn->type_entry->data.fn.fn_type_id.return_type)) { if (instruction->operand == nullptr) { LLVMValueRef by_val_value = gen_load_untyped(g, g->cur_ret_ptr, 0, false, ""); LLVMBuildRet(g->builder, by_val_value); } else { LLVMValueRef value = ir_llvm_value(g, instruction->operand); LLVMValueRef by_val_value = gen_load_untyped(g, value, 0, false, ""); LLVMBuildRet(g->builder, by_val_value); } } else if (instruction->operand == nullptr) { if (g->cur_ret_ptr == nullptr) { LLVMBuildRetVoid(g->builder); } else { LLVMValueRef by_val_value = gen_load_untyped(g, g->cur_ret_ptr, 0, false, ""); LLVMBuildRet(g->builder, by_val_value); } } else { LLVMValueRef value = ir_llvm_value(g, instruction->operand); LLVMBuildRet(g->builder, value); } return nullptr; } static LLVMValueRef gen_overflow_shl_op(CodeGen *g, ZigType *operand_type, LLVMValueRef val1, LLVMValueRef val2) { // for unsigned left shifting, we do the lossy shift, then logically shift // right the same number of bits // if the values don't match, we have an overflow // for signed left shifting we do the same except arithmetic shift right ZigType *scalar_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; assert(scalar_type->id == ZigTypeIdInt); LLVMValueRef result = LLVMBuildShl(g->builder, val1, val2, ""); LLVMValueRef orig_val; if (scalar_type->data.integral.is_signed) { orig_val = LLVMBuildAShr(g->builder, result, val2, ""); } else { orig_val = LLVMBuildLShr(g->builder, result, val2, ""); } LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, val1, orig_val, ""); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "OverflowOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "OverflowFail"); if (operand_type->id == ZigTypeIdVector) { ok_bit = ZigLLVMBuildAndReduce(g->builder, ok_bit); } LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdShlOverflowedBits); LLVMPositionBuilderAtEnd(g->builder, ok_block); return result; } static LLVMValueRef gen_overflow_shr_op(CodeGen *g, ZigType *operand_type, LLVMValueRef val1, LLVMValueRef val2) { ZigType *scalar_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; assert(scalar_type->id == ZigTypeIdInt); LLVMValueRef result; if (scalar_type->data.integral.is_signed) { result = LLVMBuildAShr(g->builder, val1, val2, ""); } else { result = LLVMBuildLShr(g->builder, val1, val2, ""); } LLVMValueRef orig_val = LLVMBuildShl(g->builder, result, val2, ""); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, val1, orig_val, ""); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "OverflowOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "OverflowFail"); if (operand_type->id == ZigTypeIdVector) { ok_bit = ZigLLVMBuildAndReduce(g->builder, ok_bit); } LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdShrOverflowedBits); LLVMPositionBuilderAtEnd(g->builder, ok_block); return result; } static LLVMValueRef gen_float_op(CodeGen *g, LLVMValueRef val, ZigType *type_entry, BuiltinFnId op) { assert(type_entry->id == ZigTypeIdFloat || type_entry->id == ZigTypeIdVector); LLVMValueRef floor_fn = get_float_fn(g, type_entry, ZigLLVMFnIdFloatOp, op); return LLVMBuildCall(g->builder, floor_fn, &val, 1, ""); } enum DivKind { DivKindFloat, DivKindTrunc, DivKindFloor, DivKindExact, }; static LLVMValueRef bigint_to_llvm_const(LLVMTypeRef type_ref, BigInt *bigint) { if (bigint->digit_count == 0) { return LLVMConstNull(type_ref); } if (LLVMGetTypeKind(type_ref) == LLVMVectorTypeKind) { const unsigned vector_len = LLVMGetVectorSize(type_ref); LLVMTypeRef elem_type = LLVMGetElementType(type_ref); LLVMValueRef *values = heap::c_allocator.allocate_nonzero(vector_len); // Create a vector with all the elements having the same value for (unsigned i = 0; i < vector_len; i++) { values[i] = bigint_to_llvm_const(elem_type, bigint); } LLVMValueRef result = LLVMConstVector(values, vector_len); heap::c_allocator.deallocate(values, vector_len); return result; } LLVMValueRef unsigned_val; if (bigint->digit_count == 1) { unsigned_val = LLVMConstInt(type_ref, bigint_ptr(bigint)[0], false); } else { unsigned_val = LLVMConstIntOfArbitraryPrecision(type_ref, bigint->digit_count, bigint_ptr(bigint)); } if (bigint->is_negative) { return LLVMConstNeg(unsigned_val); } else { return unsigned_val; } } static LLVMValueRef gen_div(CodeGen *g, bool want_runtime_safety, bool want_fast_math, LLVMValueRef val1, LLVMValueRef val2, ZigType *operand_type, DivKind div_kind) { ZigType *scalar_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; ZigLLVMSetFastMath(g->builder, want_fast_math); LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, operand_type)); if (want_runtime_safety && (want_fast_math || scalar_type->id != ZigTypeIdFloat)) { // Safety check: divisor != 0 LLVMValueRef is_zero_bit; if (scalar_type->id == ZigTypeIdInt) { is_zero_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, val2, zero, ""); } else if (scalar_type->id == ZigTypeIdFloat) { is_zero_bit = LLVMBuildFCmp(g->builder, LLVMRealOEQ, val2, zero, ""); } else { zig_unreachable(); } if (operand_type->id == ZigTypeIdVector) { is_zero_bit = ZigLLVMBuildOrReduce(g->builder, is_zero_bit); } LLVMBasicBlockRef div_zero_fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivZeroFail"); LLVMBasicBlockRef div_zero_ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivZeroOk"); LLVMBuildCondBr(g->builder, is_zero_bit, div_zero_fail_block, div_zero_ok_block); LLVMPositionBuilderAtEnd(g->builder, div_zero_fail_block); gen_safety_crash(g, PanicMsgIdDivisionByZero); LLVMPositionBuilderAtEnd(g->builder, div_zero_ok_block); // Safety check: check for overflow (dividend = minInt and divisor = -1) if (scalar_type->id == ZigTypeIdInt && scalar_type->data.integral.is_signed) { LLVMValueRef neg_1_value = LLVMConstAllOnes(get_llvm_type(g, operand_type)); BigInt int_min_bi = {0}; eval_min_max_value_int(g, scalar_type, &int_min_bi, false); LLVMValueRef int_min_value = bigint_to_llvm_const(get_llvm_type(g, operand_type), &int_min_bi); LLVMBasicBlockRef overflow_fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivOverflowFail"); LLVMBasicBlockRef overflow_ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivOverflowOk"); LLVMValueRef num_is_int_min = LLVMBuildICmp(g->builder, LLVMIntEQ, val1, int_min_value, ""); LLVMValueRef den_is_neg_1 = LLVMBuildICmp(g->builder, LLVMIntEQ, val2, neg_1_value, ""); LLVMValueRef overflow_fail_bit = LLVMBuildAnd(g->builder, num_is_int_min, den_is_neg_1, ""); if (operand_type->id == ZigTypeIdVector) { overflow_fail_bit = ZigLLVMBuildOrReduce(g->builder, overflow_fail_bit); } LLVMBuildCondBr(g->builder, overflow_fail_bit, overflow_fail_block, overflow_ok_block); LLVMPositionBuilderAtEnd(g->builder, overflow_fail_block); gen_safety_crash(g, PanicMsgIdIntegerOverflow); LLVMPositionBuilderAtEnd(g->builder, overflow_ok_block); } } if (scalar_type->id == ZigTypeIdFloat) { LLVMValueRef result = LLVMBuildFDiv(g->builder, val1, val2, ""); switch (div_kind) { case DivKindFloat: return result; case DivKindExact: if (want_runtime_safety) { // Safety check: a / b == floor(a / b) LLVMValueRef floored = gen_float_op(g, result, operand_type, BuiltinFnIdFloor); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivExactOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivExactFail"); LLVMValueRef ok_bit = LLVMBuildFCmp(g->builder, LLVMRealOEQ, floored, result, ""); if (operand_type->id == ZigTypeIdVector) { ok_bit = ZigLLVMBuildAndReduce(g->builder, ok_bit); } LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdExactDivisionRemainder); LLVMPositionBuilderAtEnd(g->builder, ok_block); } return result; case DivKindTrunc: { LLVMBasicBlockRef ltz_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivTruncLTZero"); LLVMBasicBlockRef gez_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivTruncGEZero"); LLVMBasicBlockRef end_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivTruncEnd"); LLVMValueRef ltz = LLVMBuildFCmp(g->builder, LLVMRealOLT, val1, zero, ""); if (operand_type->id == ZigTypeIdVector) { ltz = ZigLLVMBuildOrReduce(g->builder, ltz); } LLVMBuildCondBr(g->builder, ltz, ltz_block, gez_block); LLVMPositionBuilderAtEnd(g->builder, ltz_block); LLVMValueRef ceiled = gen_float_op(g, result, operand_type, BuiltinFnIdCeil); LLVMBasicBlockRef ceiled_end_block = LLVMGetInsertBlock(g->builder); LLVMBuildBr(g->builder, end_block); LLVMPositionBuilderAtEnd(g->builder, gez_block); LLVMValueRef floored = gen_float_op(g, result, operand_type, BuiltinFnIdFloor); LLVMBasicBlockRef floored_end_block = LLVMGetInsertBlock(g->builder); LLVMBuildBr(g->builder, end_block); LLVMPositionBuilderAtEnd(g->builder, end_block); LLVMValueRef phi = LLVMBuildPhi(g->builder, get_llvm_type(g, operand_type), ""); LLVMValueRef incoming_values[] = { ceiled, floored }; LLVMBasicBlockRef incoming_blocks[] = { ceiled_end_block, floored_end_block }; LLVMAddIncoming(phi, incoming_values, incoming_blocks, 2); return phi; } case DivKindFloor: return gen_float_op(g, result, operand_type, BuiltinFnIdFloor); } zig_unreachable(); } assert(scalar_type->id == ZigTypeIdInt); switch (div_kind) { case DivKindFloat: zig_unreachable(); case DivKindTrunc: if (scalar_type->data.integral.is_signed) { return LLVMBuildSDiv(g->builder, val1, val2, ""); } else { return LLVMBuildUDiv(g->builder, val1, val2, ""); } case DivKindExact: if (want_runtime_safety) { // Safety check: a % b == 0 LLVMValueRef remainder_val; if (scalar_type->data.integral.is_signed) { remainder_val = LLVMBuildSRem(g->builder, val1, val2, ""); } else { remainder_val = LLVMBuildURem(g->builder, val1, val2, ""); } LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivExactOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "DivExactFail"); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, remainder_val, zero, ""); if (operand_type->id == ZigTypeIdVector) { ok_bit = ZigLLVMBuildAndReduce(g->builder, ok_bit); } LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdExactDivisionRemainder); LLVMPositionBuilderAtEnd(g->builder, ok_block); } if (scalar_type->data.integral.is_signed) { return LLVMBuildExactSDiv(g->builder, val1, val2, ""); } else { return LLVMBuildExactUDiv(g->builder, val1, val2, ""); } case DivKindFloor: { if (!scalar_type->data.integral.is_signed) { return LLVMBuildUDiv(g->builder, val1, val2, ""); } // const d = @divTrunc(a, b); // const r = @rem(a, b); // return if (r == 0) d else d - ((a < 0) ^ (b < 0)); LLVMValueRef div_trunc = LLVMBuildSDiv(g->builder, val1, val2, ""); LLVMValueRef rem = LLVMBuildSRem(g->builder, val1, val2, ""); LLVMValueRef rem_eq_0 = LLVMBuildICmp(g->builder, LLVMIntEQ, rem, zero, ""); LLVMValueRef a_lt_0 = LLVMBuildICmp(g->builder, LLVMIntSLT, val1, zero, ""); LLVMValueRef b_lt_0 = LLVMBuildICmp(g->builder, LLVMIntSLT, val2, zero, ""); LLVMValueRef a_b_xor = LLVMBuildXor(g->builder, a_lt_0, b_lt_0, ""); LLVMValueRef a_b_xor_ext = LLVMBuildZExt(g->builder, a_b_xor, LLVMTypeOf(div_trunc), ""); LLVMValueRef d_sub_xor = LLVMBuildSub(g->builder, div_trunc, a_b_xor_ext, ""); return LLVMBuildSelect(g->builder, rem_eq_0, div_trunc, d_sub_xor, ""); } } zig_unreachable(); } enum RemKind { RemKindRem, RemKindMod, }; static LLVMValueRef gen_rem(CodeGen *g, bool want_runtime_safety, bool want_fast_math, LLVMValueRef val1, LLVMValueRef val2, ZigType *operand_type, RemKind rem_kind) { ZigType *scalar_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; ZigLLVMSetFastMath(g->builder, want_fast_math); LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, operand_type)); if (want_runtime_safety) { // Safety check: divisor != 0 LLVMValueRef is_zero_bit; if (scalar_type->id == ZigTypeIdInt) { LLVMIntPredicate pred = scalar_type->data.integral.is_signed ? LLVMIntSLE : LLVMIntEQ; is_zero_bit = LLVMBuildICmp(g->builder, pred, val2, zero, ""); } else if (scalar_type->id == ZigTypeIdFloat) { is_zero_bit = LLVMBuildFCmp(g->builder, LLVMRealOEQ, val2, zero, ""); } else { zig_unreachable(); } if (operand_type->id == ZigTypeIdVector) { is_zero_bit = ZigLLVMBuildOrReduce(g->builder, is_zero_bit); } LLVMBasicBlockRef rem_zero_ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "RemZeroOk"); LLVMBasicBlockRef rem_zero_fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "RemZeroFail"); LLVMBuildCondBr(g->builder, is_zero_bit, rem_zero_fail_block, rem_zero_ok_block); LLVMPositionBuilderAtEnd(g->builder, rem_zero_fail_block); gen_safety_crash(g, PanicMsgIdRemainderDivisionByZero); LLVMPositionBuilderAtEnd(g->builder, rem_zero_ok_block); } if (scalar_type->id == ZigTypeIdFloat) { if (rem_kind == RemKindRem) { return LLVMBuildFRem(g->builder, val1, val2, ""); } else { LLVMValueRef a = LLVMBuildFRem(g->builder, val1, val2, ""); LLVMValueRef b = LLVMBuildFAdd(g->builder, a, val2, ""); LLVMValueRef c = LLVMBuildFRem(g->builder, b, val2, ""); LLVMValueRef ltz = LLVMBuildFCmp(g->builder, LLVMRealOLT, val1, zero, ""); return LLVMBuildSelect(g->builder, ltz, c, a, ""); } } else { assert(scalar_type->id == ZigTypeIdInt); if (scalar_type->data.integral.is_signed) { if (rem_kind == RemKindRem) { return LLVMBuildSRem(g->builder, val1, val2, ""); } else { LLVMValueRef a = LLVMBuildSRem(g->builder, val1, val2, ""); LLVMValueRef b = LLVMBuildNSWAdd(g->builder, a, val2, ""); LLVMValueRef c = LLVMBuildSRem(g->builder, b, val2, ""); LLVMValueRef ltz = LLVMBuildICmp(g->builder, LLVMIntSLT, val1, zero, ""); return LLVMBuildSelect(g->builder, ltz, c, a, ""); } } else { return LLVMBuildURem(g->builder, val1, val2, ""); } } } static void gen_shift_rhs_check(CodeGen *g, ZigType *lhs_type, ZigType *rhs_type, LLVMValueRef value) { // We only check if the rhs value of the shift expression is greater or // equal to the number of bits of the lhs if it's not a power of two, // otherwise the check is useful as the allowed values are limited by the // operand type itself if (!is_power_of_2(lhs_type->data.integral.bit_count)) { BigInt bit_count_bi = {0}; bigint_init_unsigned(&bit_count_bi, lhs_type->data.integral.bit_count); LLVMValueRef bit_count_value = bigint_to_llvm_const(get_llvm_type(g, rhs_type), &bit_count_bi); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "CheckFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "CheckOk"); LLVMValueRef less_than_bit = LLVMBuildICmp(g->builder, LLVMIntULT, value, bit_count_value, ""); if (rhs_type->id == ZigTypeIdVector) { less_than_bit = ZigLLVMBuildOrReduce(g->builder, less_than_bit); } LLVMBuildCondBr(g->builder, less_than_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdShxTooBigRhs); LLVMPositionBuilderAtEnd(g->builder, ok_block); } } static LLVMValueRef ir_render_bin_op(CodeGen *g, IrExecutableGen *executable, IrInstGenBinOp *bin_op_instruction) { IrBinOp op_id = bin_op_instruction->op_id; IrInstGen *op1 = bin_op_instruction->op1; IrInstGen *op2 = bin_op_instruction->op2; ZigType *operand_type = op1->value->type; ZigType *scalar_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; bool want_runtime_safety = bin_op_instruction->safety_check_on && ir_want_runtime_safety(g, &bin_op_instruction->base); LLVMValueRef op1_value = ir_llvm_value(g, op1); LLVMValueRef op2_value = ir_llvm_value(g, op2); switch (op_id) { case IrBinOpInvalid: case IrBinOpArrayCat: case IrBinOpArrayMult: case IrBinOpRemUnspecified: zig_unreachable(); case IrBinOpBoolOr: return LLVMBuildOr(g->builder, op1_value, op2_value, ""); case IrBinOpBoolAnd: return LLVMBuildAnd(g->builder, op1_value, op2_value, ""); case IrBinOpCmpEq: case IrBinOpCmpNotEq: case IrBinOpCmpLessThan: case IrBinOpCmpGreaterThan: case IrBinOpCmpLessOrEq: case IrBinOpCmpGreaterOrEq: if (scalar_type->id == ZigTypeIdFloat) { ZigLLVMSetFastMath(g->builder, ir_want_fast_math(g, &bin_op_instruction->base)); LLVMRealPredicate pred = cmp_op_to_real_predicate(op_id); return LLVMBuildFCmp(g->builder, pred, op1_value, op2_value, ""); } else if (scalar_type->id == ZigTypeIdInt) { LLVMIntPredicate pred = cmp_op_to_int_predicate(op_id, scalar_type->data.integral.is_signed); return LLVMBuildICmp(g->builder, pred, op1_value, op2_value, ""); } else if (scalar_type->id == ZigTypeIdEnum || scalar_type->id == ZigTypeIdErrorSet || scalar_type->id == ZigTypeIdBool || get_codegen_ptr_type_bail(g, scalar_type) != nullptr) { LLVMIntPredicate pred = cmp_op_to_int_predicate(op_id, false); return LLVMBuildICmp(g->builder, pred, op1_value, op2_value, ""); } else { zig_unreachable(); } case IrBinOpMult: case IrBinOpMultWrap: case IrBinOpAdd: case IrBinOpAddWrap: case IrBinOpSub: case IrBinOpSubWrap: { bool is_wrapping = (op_id == IrBinOpSubWrap || op_id == IrBinOpAddWrap || op_id == IrBinOpMultWrap); AddSubMul add_sub_mul = op_id == IrBinOpAdd || op_id == IrBinOpAddWrap ? AddSubMulAdd : op_id == IrBinOpSub || op_id == IrBinOpSubWrap ? AddSubMulSub : AddSubMulMul; if (scalar_type->id == ZigTypeIdPointer) { LLVMValueRef subscript_value; if (operand_type->id == ZigTypeIdVector) zig_panic("TODO: Implement vector operations on pointers."); switch (add_sub_mul) { case AddSubMulAdd: subscript_value = op2_value; break; case AddSubMulSub: subscript_value = LLVMBuildNeg(g->builder, op2_value, ""); break; case AddSubMulMul: zig_unreachable(); } // TODO runtime safety return LLVMBuildInBoundsGEP(g->builder, op1_value, &subscript_value, 1, ""); } else if (scalar_type->id == ZigTypeIdFloat) { ZigLLVMSetFastMath(g->builder, ir_want_fast_math(g, &bin_op_instruction->base)); return float_op[add_sub_mul](g->builder, op1_value, op2_value, ""); } else if (scalar_type->id == ZigTypeIdInt) { if (is_wrapping) { return wrap_op[add_sub_mul](g->builder, op1_value, op2_value, ""); } else if (want_runtime_safety) { return gen_overflow_op(g, operand_type, add_sub_mul, op1_value, op2_value); } else if (scalar_type->data.integral.is_signed) { return signed_op[add_sub_mul](g->builder, op1_value, op2_value, ""); } else { return unsigned_op[add_sub_mul](g->builder, op1_value, op2_value, ""); } } else { zig_unreachable(); } } case IrBinOpBinOr: return LLVMBuildOr(g->builder, op1_value, op2_value, ""); case IrBinOpBinXor: return LLVMBuildXor(g->builder, op1_value, op2_value, ""); case IrBinOpBinAnd: return LLVMBuildAnd(g->builder, op1_value, op2_value, ""); case IrBinOpBitShiftLeftLossy: case IrBinOpBitShiftLeftExact: { assert(scalar_type->id == ZigTypeIdInt); LLVMValueRef op2_casted = LLVMBuildZExt(g->builder, op2_value, LLVMTypeOf(op1_value), ""); if (want_runtime_safety) { gen_shift_rhs_check(g, scalar_type, op2->value->type, op2_value); } bool is_sloppy = (op_id == IrBinOpBitShiftLeftLossy); if (is_sloppy) { return LLVMBuildShl(g->builder, op1_value, op2_casted, ""); } else if (want_runtime_safety) { return gen_overflow_shl_op(g, operand_type, op1_value, op2_casted); } else if (scalar_type->data.integral.is_signed) { return ZigLLVMBuildNSWShl(g->builder, op1_value, op2_casted, ""); } else { return ZigLLVMBuildNUWShl(g->builder, op1_value, op2_casted, ""); } } case IrBinOpBitShiftRightLossy: case IrBinOpBitShiftRightExact: { assert(scalar_type->id == ZigTypeIdInt); LLVMValueRef op2_casted = LLVMBuildZExt(g->builder, op2_value, LLVMTypeOf(op1_value), ""); if (want_runtime_safety) { gen_shift_rhs_check(g, scalar_type, op2->value->type, op2_value); } bool is_sloppy = (op_id == IrBinOpBitShiftRightLossy); if (is_sloppy) { if (scalar_type->data.integral.is_signed) { return LLVMBuildAShr(g->builder, op1_value, op2_casted, ""); } else { return LLVMBuildLShr(g->builder, op1_value, op2_casted, ""); } } else if (want_runtime_safety) { return gen_overflow_shr_op(g, operand_type, op1_value, op2_casted); } else if (scalar_type->data.integral.is_signed) { return ZigLLVMBuildAShrExact(g->builder, op1_value, op2_casted, ""); } else { return ZigLLVMBuildLShrExact(g->builder, op1_value, op2_casted, ""); } } case IrBinOpDivUnspecified: return gen_div(g, want_runtime_safety, ir_want_fast_math(g, &bin_op_instruction->base), op1_value, op2_value, operand_type, DivKindFloat); case IrBinOpDivExact: return gen_div(g, want_runtime_safety, ir_want_fast_math(g, &bin_op_instruction->base), op1_value, op2_value, operand_type, DivKindExact); case IrBinOpDivTrunc: return gen_div(g, want_runtime_safety, ir_want_fast_math(g, &bin_op_instruction->base), op1_value, op2_value, operand_type, DivKindTrunc); case IrBinOpDivFloor: return gen_div(g, want_runtime_safety, ir_want_fast_math(g, &bin_op_instruction->base), op1_value, op2_value, operand_type, DivKindFloor); case IrBinOpRemRem: return gen_rem(g, want_runtime_safety, ir_want_fast_math(g, &bin_op_instruction->base), op1_value, op2_value, operand_type, RemKindRem); case IrBinOpRemMod: return gen_rem(g, want_runtime_safety, ir_want_fast_math(g, &bin_op_instruction->base), op1_value, op2_value, operand_type, RemKindMod); } zig_unreachable(); } static void add_error_range_check(CodeGen *g, ZigType *err_set_type, ZigType *int_type, LLVMValueRef target_val) { assert(err_set_type->id == ZigTypeIdErrorSet); if (type_is_global_error_set(err_set_type)) { LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, int_type)); LLVMValueRef neq_zero_bit = LLVMBuildICmp(g->builder, LLVMIntNE, target_val, zero, ""); LLVMValueRef ok_bit; BigInt biggest_possible_err_val = {0}; eval_min_max_value_int(g, int_type, &biggest_possible_err_val, true); if (bigint_fits_in_bits(&biggest_possible_err_val, 64, false) && bigint_as_usize(&biggest_possible_err_val) < g->errors_by_index.length) { ok_bit = neq_zero_bit; } else { LLVMValueRef error_value_count = LLVMConstInt(get_llvm_type(g, int_type), g->errors_by_index.length, false); LLVMValueRef in_bounds_bit = LLVMBuildICmp(g->builder, LLVMIntULT, target_val, error_value_count, ""); ok_bit = LLVMBuildAnd(g->builder, neq_zero_bit, in_bounds_bit, ""); } LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "IntToErrOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "IntToErrFail"); LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdInvalidErrorCode); LLVMPositionBuilderAtEnd(g->builder, ok_block); } else { LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "IntToErrOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "IntToErrFail"); uint32_t err_count = err_set_type->data.error_set.err_count; LLVMValueRef switch_instr = LLVMBuildSwitch(g->builder, target_val, fail_block, err_count); for (uint32_t i = 0; i < err_count; i += 1) { LLVMValueRef case_value = LLVMConstInt(get_llvm_type(g, g->err_tag_type), err_set_type->data.error_set.errors[i]->value, false); LLVMAddCase(switch_instr, case_value, ok_block); } LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdInvalidErrorCode); LLVMPositionBuilderAtEnd(g->builder, ok_block); } } static LLVMValueRef ir_render_cast(CodeGen *g, IrExecutableGen *executable, IrInstGenCast *cast_instruction) { Error err; ZigType *actual_type = cast_instruction->value->value->type; ZigType *wanted_type = cast_instruction->base.value->type; bool wanted_type_has_bits; if ((err = type_has_bits2(g, wanted_type, &wanted_type_has_bits))) codegen_report_errors_and_exit(g); if (!wanted_type_has_bits) return nullptr; LLVMValueRef expr_val = ir_llvm_value(g, cast_instruction->value); ir_assert(expr_val, &cast_instruction->base); switch (cast_instruction->cast_op) { case CastOpNoCast: case CastOpNumLitToConcrete: zig_unreachable(); case CastOpNoop: if (actual_type->id == ZigTypeIdPointer && wanted_type->id == ZigTypeIdPointer && actual_type->data.pointer.child_type->id == ZigTypeIdArray && wanted_type->data.pointer.child_type->id == ZigTypeIdArray) { return LLVMBuildBitCast(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } else { return expr_val; } case CastOpIntToFloat: assert(actual_type->id == ZigTypeIdInt); if (actual_type->data.integral.is_signed) { return LLVMBuildSIToFP(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } else { return LLVMBuildUIToFP(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } case CastOpFloatToInt: { assert(wanted_type->id == ZigTypeIdInt); ZigLLVMSetFastMath(g->builder, ir_want_fast_math(g, &cast_instruction->base)); bool want_safety = ir_want_runtime_safety(g, &cast_instruction->base); LLVMValueRef result; if (wanted_type->data.integral.is_signed) { result = LLVMBuildFPToSI(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } else { result = LLVMBuildFPToUI(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } if (want_safety) { LLVMValueRef back_to_float; if (wanted_type->data.integral.is_signed) { back_to_float = LLVMBuildSIToFP(g->builder, result, LLVMTypeOf(expr_val), ""); } else { back_to_float = LLVMBuildUIToFP(g->builder, result, LLVMTypeOf(expr_val), ""); } LLVMValueRef difference = LLVMBuildFSub(g->builder, expr_val, back_to_float, ""); LLVMValueRef one_pos = LLVMConstReal(LLVMTypeOf(expr_val), 1.0f); LLVMValueRef one_neg = LLVMConstReal(LLVMTypeOf(expr_val), -1.0f); LLVMValueRef ok_bit_pos = LLVMBuildFCmp(g->builder, LLVMRealOLT, difference, one_pos, ""); LLVMValueRef ok_bit_neg = LLVMBuildFCmp(g->builder, LLVMRealOGT, difference, one_neg, ""); LLVMValueRef ok_bit = LLVMBuildAnd(g->builder, ok_bit_pos, ok_bit_neg, ""); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "FloatCheckOk"); LLVMBasicBlockRef bad_block = LLVMAppendBasicBlock(g->cur_fn_val, "FloatCheckFail"); LLVMBuildCondBr(g->builder, ok_bit, ok_block, bad_block); LLVMPositionBuilderAtEnd(g->builder, bad_block); gen_safety_crash(g, PanicMsgIdFloatToInt); LLVMPositionBuilderAtEnd(g->builder, ok_block); } return result; } case CastOpBoolToInt: assert(wanted_type->id == ZigTypeIdInt); assert(actual_type->id == ZigTypeIdBool); return LLVMBuildZExt(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); case CastOpErrSet: if (ir_want_runtime_safety(g, &cast_instruction->base)) { add_error_range_check(g, wanted_type, g->err_tag_type, expr_val); } return expr_val; case CastOpBitCast: return LLVMBuildBitCast(g->builder, expr_val, get_llvm_type(g, wanted_type), ""); } zig_unreachable(); } static LLVMValueRef ir_render_ptr_of_array_to_slice(CodeGen *g, IrExecutableGen *executable, IrInstGenPtrOfArrayToSlice *instruction) { ZigType *actual_type = instruction->operand->value->type; ZigType *slice_type = instruction->base.value->type; ZigType *slice_ptr_type = slice_type->data.structure.fields[slice_ptr_index]->type_entry; size_t ptr_index = slice_type->data.structure.fields[slice_ptr_index]->gen_index; size_t len_index = slice_type->data.structure.fields[slice_len_index]->gen_index; LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); assert(actual_type->id == ZigTypeIdPointer); ZigType *array_type = actual_type->data.pointer.child_type; assert(array_type->id == ZigTypeIdArray); if (type_has_bits(g, actual_type)) { LLVMValueRef ptr_field_ptr = LLVMBuildStructGEP(g->builder, result_loc, ptr_index, ""); LLVMValueRef indices[] = { LLVMConstNull(g->builtin_types.entry_usize->llvm_type), LLVMConstInt(g->builtin_types.entry_usize->llvm_type, 0, false), }; LLVMValueRef expr_val = ir_llvm_value(g, instruction->operand); LLVMValueRef slice_start_ptr = LLVMBuildInBoundsGEP(g->builder, expr_val, indices, 2, ""); gen_store_untyped(g, slice_start_ptr, ptr_field_ptr, 0, false); } else if (ir_want_runtime_safety(g, &instruction->base)) { LLVMValueRef ptr_field_ptr = LLVMBuildStructGEP(g->builder, result_loc, ptr_index, ""); gen_undef_init(g, slice_ptr_type, slice_ptr_type, ptr_field_ptr); } LLVMValueRef len_field_ptr = LLVMBuildStructGEP(g->builder, result_loc, len_index, ""); LLVMValueRef len_value = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, array_type->data.array.len, false); gen_store_untyped(g, len_value, len_field_ptr, 0, false); return result_loc; } static LLVMValueRef ir_render_ptr_cast(CodeGen *g, IrExecutableGen *executable, IrInstGenPtrCast *instruction) { ZigType *wanted_type = instruction->base.value->type; if (!type_has_bits(g, wanted_type)) { return nullptr; } LLVMValueRef ptr = ir_llvm_value(g, instruction->ptr); LLVMValueRef result_ptr = LLVMBuildBitCast(g->builder, ptr, get_llvm_type(g, wanted_type), ""); bool want_safety_check = instruction->safety_check_on && ir_want_runtime_safety(g, &instruction->base); if (!want_safety_check || ptr_allows_addr_zero(wanted_type)) return result_ptr; LLVMValueRef zero = LLVMConstNull(LLVMTypeOf(result_ptr)); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntNE, result_ptr, zero, ""); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "PtrCastFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "PtrCastOk"); LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdPtrCastNull); LLVMPositionBuilderAtEnd(g->builder, ok_block); return result_ptr; } static LLVMValueRef ir_render_bit_cast(CodeGen *g, IrExecutableGen *executable, IrInstGenBitCast *instruction) { ZigType *wanted_type = instruction->base.value->type; ZigType *actual_type = instruction->operand->value->type; LLVMValueRef value = ir_llvm_value(g, instruction->operand); bool wanted_is_ptr = handle_is_ptr(g, wanted_type); bool actual_is_ptr = handle_is_ptr(g, actual_type); if (wanted_is_ptr == actual_is_ptr) { // We either bitcast the value directly or bitcast the pointer which does a pointer cast LLVMTypeRef wanted_type_ref = wanted_is_ptr ? LLVMPointerType(get_llvm_type(g, wanted_type), 0) : get_llvm_type(g, wanted_type); return LLVMBuildBitCast(g->builder, value, wanted_type_ref, ""); } else if (actual_is_ptr) { // A scalar is wanted but we got a pointer LLVMTypeRef wanted_ptr_type_ref = LLVMPointerType(get_llvm_type(g, wanted_type), 0); LLVMValueRef bitcasted_ptr = LLVMBuildBitCast(g->builder, value, wanted_ptr_type_ref, ""); uint32_t alignment = get_abi_alignment(g, actual_type); return gen_load_untyped(g, bitcasted_ptr, alignment, false, ""); } else { // A pointer is wanted but we got a scalar assert(actual_type->id == ZigTypeIdPointer); LLVMTypeRef wanted_ptr_type_ref = LLVMPointerType(get_llvm_type(g, wanted_type), 0); return LLVMBuildBitCast(g->builder, value, wanted_ptr_type_ref, ""); } } static LLVMValueRef ir_render_widen_or_shorten(CodeGen *g, IrExecutableGen *executable, IrInstGenWidenOrShorten *instruction) { ZigType *actual_type = instruction->target->value->type; // TODO instead of this logic, use the Noop instruction to change the type from // enum_tag to the underlying int type ZigType *int_type; if (actual_type->id == ZigTypeIdEnum) { int_type = actual_type->data.enumeration.tag_int_type; } else { int_type = actual_type; } LLVMValueRef target_val = ir_llvm_value(g, instruction->target); return gen_widen_or_shorten(g, ir_want_runtime_safety(g, &instruction->base), int_type, instruction->base.value->type, target_val); } static LLVMValueRef ir_render_int_to_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenIntToPtr *instruction) { ZigType *wanted_type = instruction->base.value->type; LLVMValueRef target_val = ir_llvm_value(g, instruction->target); const uint32_t align_bytes = get_ptr_align(g, wanted_type); if (ir_want_runtime_safety(g, &instruction->base)) { ZigType *usize = g->builtin_types.entry_usize; LLVMValueRef zero = LLVMConstNull(usize->llvm_type); if (!ptr_allows_addr_zero(wanted_type)) { LLVMValueRef is_zero_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, target_val, zero, ""); LLVMBasicBlockRef bad_block = LLVMAppendBasicBlock(g->cur_fn_val, "PtrToIntBad"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "PtrToIntOk"); LLVMBuildCondBr(g->builder, is_zero_bit, bad_block, ok_block); LLVMPositionBuilderAtEnd(g->builder, bad_block); gen_safety_crash(g, PanicMsgIdPtrCastNull); LLVMPositionBuilderAtEnd(g->builder, ok_block); } if (align_bytes > 1) { LLVMValueRef alignment_minus_1 = LLVMConstInt(usize->llvm_type, align_bytes - 1, false); LLVMValueRef anded_val = LLVMBuildAnd(g->builder, target_val, alignment_minus_1, ""); LLVMValueRef is_ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, anded_val, zero, ""); LLVMBasicBlockRef bad_block = LLVMAppendBasicBlock(g->cur_fn_val, "PtrToIntAlignBad"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "PtrToIntAlignOk"); LLVMBuildCondBr(g->builder, is_ok_bit, ok_block, bad_block); LLVMPositionBuilderAtEnd(g->builder, bad_block); gen_safety_crash(g, PanicMsgIdIncorrectAlignment); LLVMPositionBuilderAtEnd(g->builder, ok_block); } } return LLVMBuildIntToPtr(g->builder, target_val, get_llvm_type(g, wanted_type), ""); } static LLVMValueRef ir_render_ptr_to_int(CodeGen *g, IrExecutableGen *executable, IrInstGenPtrToInt *instruction) { ZigType *wanted_type = instruction->base.value->type; LLVMValueRef target_val = ir_llvm_value(g, instruction->target); return LLVMBuildPtrToInt(g->builder, target_val, get_llvm_type(g, wanted_type), ""); } static LLVMValueRef ir_render_int_to_enum(CodeGen *g, IrExecutableGen *executable, IrInstGenIntToEnum *instruction) { ZigType *wanted_type = instruction->base.value->type; assert(wanted_type->id == ZigTypeIdEnum); ZigType *tag_int_type = wanted_type->data.enumeration.tag_int_type; LLVMValueRef target_val = ir_llvm_value(g, instruction->target); LLVMValueRef tag_int_value = gen_widen_or_shorten(g, ir_want_runtime_safety(g, &instruction->base), instruction->target->value->type, tag_int_type, target_val); if (ir_want_runtime_safety(g, &instruction->base) && !wanted_type->data.enumeration.non_exhaustive) { LLVMBasicBlockRef bad_value_block = LLVMAppendBasicBlock(g->cur_fn_val, "BadValue"); LLVMBasicBlockRef ok_value_block = LLVMAppendBasicBlock(g->cur_fn_val, "OkValue"); size_t field_count = wanted_type->data.enumeration.src_field_count; LLVMValueRef switch_instr = LLVMBuildSwitch(g->builder, tag_int_value, bad_value_block, field_count); HashMap occupied_tag_values = {}; occupied_tag_values.init(field_count); for (size_t field_i = 0; field_i < field_count; field_i += 1) { TypeEnumField *type_enum_field = &wanted_type->data.enumeration.fields[field_i]; Buf *name = type_enum_field->name; auto entry = occupied_tag_values.put_unique(type_enum_field->value, name); if (entry != nullptr) { continue; } LLVMValueRef this_tag_int_value = bigint_to_llvm_const(get_llvm_type(g, tag_int_type), &type_enum_field->value); LLVMAddCase(switch_instr, this_tag_int_value, ok_value_block); } occupied_tag_values.deinit(); LLVMPositionBuilderAtEnd(g->builder, bad_value_block); gen_safety_crash(g, PanicMsgIdBadEnumValue); LLVMPositionBuilderAtEnd(g->builder, ok_value_block); } return tag_int_value; } static LLVMValueRef ir_render_int_to_err(CodeGen *g, IrExecutableGen *executable, IrInstGenIntToErr *instruction) { ZigType *wanted_type = instruction->base.value->type; assert(wanted_type->id == ZigTypeIdErrorSet); ZigType *actual_type = instruction->target->value->type; assert(actual_type->id == ZigTypeIdInt); assert(!actual_type->data.integral.is_signed); LLVMValueRef target_val = ir_llvm_value(g, instruction->target); if (ir_want_runtime_safety(g, &instruction->base)) { add_error_range_check(g, wanted_type, actual_type, target_val); } return gen_widen_or_shorten(g, false, actual_type, g->err_tag_type, target_val); } static LLVMValueRef ir_render_err_to_int(CodeGen *g, IrExecutableGen *executable, IrInstGenErrToInt *instruction) { ZigType *wanted_type = instruction->base.value->type; assert(wanted_type->id == ZigTypeIdInt); assert(!wanted_type->data.integral.is_signed); ZigType *actual_type = instruction->target->value->type; LLVMValueRef target_val = ir_llvm_value(g, instruction->target); if (actual_type->id == ZigTypeIdErrorSet) { return gen_widen_or_shorten(g, ir_want_runtime_safety(g, &instruction->base), g->err_tag_type, wanted_type, target_val); } else if (actual_type->id == ZigTypeIdErrorUnion) { // this should have been a compile time constant assert(type_has_bits(g, actual_type->data.error_union.err_set_type)); if (!type_has_bits(g, actual_type->data.error_union.payload_type)) { return gen_widen_or_shorten(g, ir_want_runtime_safety(g, &instruction->base), g->err_tag_type, wanted_type, target_val); } else { zig_panic("TODO err to int when error union payload type not void"); } } else { zig_unreachable(); } } static LLVMValueRef ir_render_unreachable(CodeGen *g, IrExecutableGen *executable, IrInstGenUnreachable *unreachable_instruction) { if (ir_want_runtime_safety(g, &unreachable_instruction->base)) { gen_safety_crash(g, PanicMsgIdUnreachable); } else { LLVMBuildUnreachable(g->builder); } return nullptr; } static LLVMValueRef ir_render_cond_br(CodeGen *g, IrExecutableGen *executable, IrInstGenCondBr *cond_br_instruction) { LLVMBuildCondBr(g->builder, ir_llvm_value(g, cond_br_instruction->condition), cond_br_instruction->then_block->llvm_block, cond_br_instruction->else_block->llvm_block); return nullptr; } static LLVMValueRef ir_render_br(CodeGen *g, IrExecutableGen *executable, IrInstGenBr *br_instruction) { LLVMBuildBr(g->builder, br_instruction->dest_block->llvm_block); return nullptr; } static LLVMValueRef ir_render_binary_not(CodeGen *g, IrExecutableGen *executable, IrInstGenBinaryNot *inst) { LLVMValueRef operand = ir_llvm_value(g, inst->operand); return LLVMBuildNot(g->builder, operand, ""); } static LLVMValueRef ir_gen_negation(CodeGen *g, IrInstGen *inst, IrInstGen *operand, bool wrapping) { LLVMValueRef llvm_operand = ir_llvm_value(g, operand); ZigType *operand_type = operand->value->type; ZigType *scalar_type = (operand_type->id == ZigTypeIdVector) ? operand_type->data.vector.elem_type : operand_type; if (scalar_type->id == ZigTypeIdFloat) { ZigLLVMSetFastMath(g->builder, ir_want_fast_math(g, inst)); return LLVMBuildFNeg(g->builder, llvm_operand, ""); } else if (scalar_type->id == ZigTypeIdInt) { if (wrapping) { return LLVMBuildNeg(g->builder, llvm_operand, ""); } else if (ir_want_runtime_safety(g, inst)) { LLVMValueRef zero = LLVMConstNull(LLVMTypeOf(llvm_operand)); return gen_overflow_op(g, operand_type, AddSubMulSub, zero, llvm_operand); } else if (scalar_type->data.integral.is_signed) { return LLVMBuildNSWNeg(g->builder, llvm_operand, ""); } else { zig_unreachable(); } } else { zig_unreachable(); } } static LLVMValueRef ir_render_negation(CodeGen *g, IrExecutableGen *executable, IrInstGenNegation *inst) { return ir_gen_negation(g, &inst->base, inst->operand, inst->wrapping); } static LLVMValueRef ir_render_bool_not(CodeGen *g, IrExecutableGen *executable, IrInstGenBoolNot *instruction) { LLVMValueRef value = ir_llvm_value(g, instruction->value); LLVMValueRef zero = LLVMConstNull(LLVMTypeOf(value)); return LLVMBuildICmp(g->builder, LLVMIntEQ, value, zero, ""); } static void render_decl_var(CodeGen *g, ZigVar *var) { if (!type_has_bits(g, var->var_type)) return; var->value_ref = ir_llvm_value(g, var->ptr_instruction); gen_var_debug_decl(g, var); } static LLVMValueRef ir_render_decl_var(CodeGen *g, IrExecutableGen *executable, IrInstGenDeclVar *instruction) { instruction->var->ptr_instruction = instruction->var_ptr; instruction->var->did_the_decl_codegen = true; render_decl_var(g, instruction->var); return nullptr; } static LLVMValueRef ir_render_load_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenLoadPtr *instruction) { ZigType *child_type = instruction->base.value->type; if (!type_has_bits(g, child_type)) return nullptr; LLVMValueRef ptr = ir_llvm_value(g, instruction->ptr); ZigType *ptr_type = instruction->ptr->value->type; assert(ptr_type->id == ZigTypeIdPointer); ir_assert(ptr_type->data.pointer.vector_index != VECTOR_INDEX_RUNTIME, &instruction->base); if (ptr_type->data.pointer.vector_index != VECTOR_INDEX_NONE) { LLVMValueRef index_val = LLVMConstInt(LLVMInt32Type(), ptr_type->data.pointer.vector_index, false); LLVMValueRef loaded_vector = LLVMBuildLoad(g->builder, ptr, ""); return LLVMBuildExtractElement(g->builder, loaded_vector, index_val, ""); } uint32_t host_int_bytes = ptr_type->data.pointer.host_int_bytes; if (host_int_bytes == 0) return get_handle_value(g, ptr, child_type, ptr_type); bool big_endian = g->is_big_endian; LLVMTypeRef int_ptr_ty = LLVMPointerType(LLVMIntType(host_int_bytes * 8), 0); LLVMValueRef int_ptr = LLVMBuildBitCast(g->builder, ptr, int_ptr_ty, ""); LLVMValueRef containing_int = gen_load(g, int_ptr, ptr_type, ""); uint32_t host_bit_count = LLVMGetIntTypeWidth(LLVMTypeOf(containing_int)); assert(host_bit_count == host_int_bytes * 8); uint32_t size_in_bits = type_size_bits(g, child_type); uint32_t bit_offset = ptr_type->data.pointer.bit_offset_in_host; uint32_t shift_amt = big_endian ? host_bit_count - bit_offset - size_in_bits : bit_offset; LLVMValueRef shift_amt_val = LLVMConstInt(LLVMTypeOf(containing_int), shift_amt, false); LLVMValueRef shifted_value = LLVMBuildLShr(g->builder, containing_int, shift_amt_val, ""); if (handle_is_ptr(g, child_type)) { LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); LLVMTypeRef same_size_int = LLVMIntType(size_in_bits); LLVMValueRef truncated_int = LLVMBuildTrunc(g->builder, shifted_value, same_size_int, ""); LLVMValueRef bitcasted_ptr = LLVMBuildBitCast(g->builder, result_loc, LLVMPointerType(same_size_int, 0), ""); LLVMBuildStore(g->builder, truncated_int, bitcasted_ptr); return result_loc; } if (child_type->id == ZigTypeIdFloat) { LLVMTypeRef same_size_int = LLVMIntType(size_in_bits); LLVMValueRef truncated_int = LLVMBuildTrunc(g->builder, shifted_value, same_size_int, ""); return LLVMBuildBitCast(g->builder, truncated_int, get_llvm_type(g, child_type), ""); } return LLVMBuildTrunc(g->builder, shifted_value, get_llvm_type(g, child_type), ""); } static bool value_is_all_undef_array(CodeGen *g, ZigValue *const_val, size_t len) { switch (const_val->data.x_array.special) { case ConstArraySpecialUndef: return true; case ConstArraySpecialBuf: return false; case ConstArraySpecialNone: for (size_t i = 0; i < len; i += 1) { if (!value_is_all_undef(g, &const_val->data.x_array.data.s_none.elements[i])) return false; } return true; } zig_unreachable(); } static bool value_is_all_undef(CodeGen *g, ZigValue *const_val) { Error err; if (const_val->special == ConstValSpecialLazy && (err = ir_resolve_lazy(g, nullptr, const_val))) codegen_report_errors_and_exit(g); switch (const_val->special) { case ConstValSpecialLazy: zig_unreachable(); case ConstValSpecialRuntime: return false; case ConstValSpecialUndef: return true; case ConstValSpecialStatic: if (const_val->type->id == ZigTypeIdStruct) { for (size_t i = 0; i < const_val->type->data.structure.src_field_count; i += 1) { if (!value_is_all_undef(g, const_val->data.x_struct.fields[i])) return false; } return true; } else if (const_val->type->id == ZigTypeIdArray) { return value_is_all_undef_array(g, const_val, const_val->type->data.array.len); } else if (const_val->type->id == ZigTypeIdVector) { return value_is_all_undef_array(g, const_val, const_val->type->data.vector.len); } else { return false; } } zig_unreachable(); } static LLVMValueRef gen_valgrind_client_request(CodeGen *g, LLVMValueRef default_value, LLVMValueRef request, LLVMValueRef a1, LLVMValueRef a2, LLVMValueRef a3, LLVMValueRef a4, LLVMValueRef a5) { if (!target_has_valgrind_support(g->zig_target)) { return default_value; } LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; bool asm_has_side_effects = true; bool asm_is_alignstack = false; if (g->zig_target->arch == ZigLLVM_x86_64) { if (g->zig_target->os == OsLinux || target_os_is_darwin(g->zig_target->os) || g->zig_target->os == OsSolaris || (g->zig_target->os == OsWindows && g->zig_target->abi != ZigLLVM_MSVC)) { if (g->cur_fn->valgrind_client_request_array == nullptr) { LLVMBasicBlockRef prev_block = LLVMGetInsertBlock(g->builder); LLVMBasicBlockRef entry_block = LLVMGetEntryBasicBlock(g->cur_fn->llvm_value); LLVMValueRef first_inst = LLVMGetFirstInstruction(entry_block); LLVMPositionBuilderBefore(g->builder, first_inst); LLVMTypeRef array_type_ref = LLVMArrayType(usize_type_ref, 6); g->cur_fn->valgrind_client_request_array = LLVMBuildAlloca(g->builder, array_type_ref, ""); LLVMPositionBuilderAtEnd(g->builder, prev_block); } LLVMValueRef array_ptr = g->cur_fn->valgrind_client_request_array; LLVMValueRef array_elements[] = {request, a1, a2, a3, a4, a5}; LLVMValueRef zero = LLVMConstInt(usize_type_ref, 0, false); for (unsigned i = 0; i < 6; i += 1) { LLVMValueRef indexes[] = { zero, LLVMConstInt(usize_type_ref, i, false), }; LLVMValueRef elem_ptr = LLVMBuildInBoundsGEP(g->builder, array_ptr, indexes, 2, ""); LLVMBuildStore(g->builder, array_elements[i], elem_ptr); } Buf *asm_template = buf_create_from_str( "rolq $$3, %rdi ; rolq $$13, %rdi\n" "rolq $$61, %rdi ; rolq $$51, %rdi\n" "xchgq %rbx,%rbx\n" ); Buf *asm_constraints = buf_create_from_str( "={rdx},{rax},0,~{cc},~{memory}" ); unsigned input_and_output_count = 2; LLVMValueRef array_ptr_as_usize = LLVMBuildPtrToInt(g->builder, array_ptr, usize_type_ref, ""); LLVMValueRef param_values[] = { array_ptr_as_usize, default_value }; LLVMTypeRef param_types[] = {usize_type_ref, usize_type_ref}; LLVMTypeRef function_type = LLVMFunctionType(usize_type_ref, param_types, input_and_output_count, false); LLVMValueRef asm_fn = LLVMGetInlineAsm(function_type, buf_ptr(asm_template), buf_len(asm_template), buf_ptr(asm_constraints), buf_len(asm_constraints), asm_has_side_effects, asm_is_alignstack, LLVMInlineAsmDialectATT); return LLVMBuildCall(g->builder, asm_fn, param_values, input_and_output_count, ""); } } zig_unreachable(); } static void gen_valgrind_undef(CodeGen *g, LLVMValueRef dest_ptr, LLVMValueRef byte_count) { static const uint32_t VG_USERREQ__MAKE_MEM_UNDEFINED = 1296236545; ZigType *usize = g->builtin_types.entry_usize; LLVMValueRef zero = LLVMConstInt(usize->llvm_type, 0, false); LLVMValueRef req = LLVMConstInt(usize->llvm_type, VG_USERREQ__MAKE_MEM_UNDEFINED, false); LLVMValueRef ptr_as_usize = LLVMBuildPtrToInt(g->builder, dest_ptr, usize->llvm_type, ""); gen_valgrind_client_request(g, zero, req, ptr_as_usize, byte_count, zero, zero, zero); } static void gen_undef_init(CodeGen *g, ZigType *ptr_type, ZigType *value_type, LLVMValueRef ptr) { assert(type_has_bits(g, value_type)); uint64_t ptr_align_bytes = get_ptr_align(g, ptr_type); assert(ptr_align_bytes > 0); uint64_t size_bytes = LLVMStoreSizeOfType(g->target_data_ref, get_llvm_type(g, value_type)); assert(size_bytes > 0); if (ptr_type->data.pointer.host_int_bytes == 0) { // memset uninitialized memory to 0xaa LLVMTypeRef ptr_u8 = LLVMPointerType(LLVMInt8Type(), 0); LLVMValueRef fill_char = LLVMConstInt(LLVMInt8Type(), 0xaa, false); LLVMValueRef dest_ptr = LLVMBuildBitCast(g->builder, ptr, ptr_u8, ""); ZigType *usize = g->builtin_types.entry_usize; LLVMValueRef byte_count = LLVMConstInt(usize->llvm_type, size_bytes, false); ZigLLVMBuildMemSet(g->builder, dest_ptr, fill_char, byte_count, ptr_align_bytes, false); // then tell valgrind that the memory is undefined even though we just memset it if (g->valgrind_enabled) { gen_valgrind_undef(g, dest_ptr, byte_count); } return; } // This is a pointer into a packed struct, we can't use memset here. // The jury is still out on what pattern should be written here so clear the // old value and call it a day. Generating a 0xAA...AA mask for this n-bit // value is left as an exercise for the (bored) reader. LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, value_type)); gen_assign_raw(g, ptr, ptr_type, zero); } static LLVMValueRef ir_render_store_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenStorePtr *instruction) { Error err; ZigType *ptr_type = instruction->ptr->value->type; assert(ptr_type->id == ZigTypeIdPointer); bool ptr_type_has_bits; if ((err = type_has_bits2(g, ptr_type, &ptr_type_has_bits))) codegen_report_errors_and_exit(g); if (!ptr_type_has_bits) return nullptr; if (instruction->ptr->base.ref_count == 0) { // In this case, this StorePtr instruction should be elided. Something happened like this: // var t = true; // const x = if (t) Num.Two else unreachable; // The if condition is a runtime value, so the StorePtr for `x = Num.Two` got generated // (this instruction being rendered) but because of `else unreachable` the result ended // up being a comptime const value. return nullptr; } bool have_init_expr = !value_is_all_undef(g, instruction->value->value); if (have_init_expr) { LLVMValueRef ptr = ir_llvm_value(g, instruction->ptr); LLVMValueRef value = ir_llvm_value(g, instruction->value); gen_assign_raw(g, ptr, ptr_type, value); } else if (ir_want_runtime_safety(g, &instruction->base)) { gen_undef_init(g, ptr_type, instruction->value->value->type, ir_llvm_value(g, instruction->ptr)); } return nullptr; } static LLVMValueRef ir_render_vector_store_elem(CodeGen *g, IrExecutableGen *executable, IrInstGenVectorStoreElem *instruction) { LLVMValueRef vector_ptr = ir_llvm_value(g, instruction->vector_ptr); LLVMValueRef index = ir_llvm_value(g, instruction->index); LLVMValueRef value = ir_llvm_value(g, instruction->value); LLVMValueRef loaded_vector = gen_load(g, vector_ptr, instruction->vector_ptr->value->type, ""); LLVMValueRef modified_vector = LLVMBuildInsertElement(g->builder, loaded_vector, value, index, ""); gen_store(g, modified_vector, vector_ptr, instruction->vector_ptr->value->type); return nullptr; } static LLVMValueRef ir_render_var_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenVarPtr *instruction) { if (instruction->base.value->special != ConstValSpecialRuntime) return ir_llvm_value(g, &instruction->base); ZigVar *var = instruction->var; if (type_has_bits(g, var->var_type)) { assert(var->value_ref); return var->value_ref; } else { return nullptr; } } static LLVMValueRef ir_render_return_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenReturnPtr *instruction) { if (!type_has_bits(g, instruction->base.value->type)) return nullptr; ir_assert(g->cur_ret_ptr != nullptr, &instruction->base); return g->cur_ret_ptr; } static LLVMValueRef ir_render_elem_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenElemPtr *instruction) { LLVMValueRef array_ptr_ptr = ir_llvm_value(g, instruction->array_ptr); ZigType *array_ptr_type = instruction->array_ptr->value->type; assert(array_ptr_type->id == ZigTypeIdPointer); ZigType *array_type = array_ptr_type->data.pointer.child_type; LLVMValueRef subscript_value = ir_llvm_value(g, instruction->elem_index); assert(subscript_value); if (!type_has_bits(g, array_type)) return nullptr; bool safety_check_on = ir_want_runtime_safety(g, &instruction->base) && instruction->safety_check_on; if (array_type->id == ZigTypeIdArray || (array_type->id == ZigTypeIdPointer && array_type->data.pointer.ptr_len == PtrLenSingle)) { LLVMValueRef array_ptr = get_handle_value(g, array_ptr_ptr, array_type, array_ptr_type); if (array_type->id == ZigTypeIdPointer) { assert(array_type->data.pointer.child_type->id == ZigTypeIdArray); array_type = array_type->data.pointer.child_type; } assert(array_type->data.array.len != 0 || array_type->data.array.sentinel != nullptr); if (safety_check_on) { uint64_t extra_len_from_sentinel = (array_type->data.array.sentinel != nullptr) ? 1 : 0; uint64_t full_len = array_type->data.array.len + extra_len_from_sentinel; LLVMValueRef end = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, full_len, false); add_bounds_check(g, subscript_value, LLVMIntEQ, nullptr, LLVMIntULT, end); } if (array_ptr_type->data.pointer.host_int_bytes != 0) { return array_ptr_ptr; } ZigType *child_type = array_type->data.array.child_type; if (child_type->id == ZigTypeIdStruct && child_type->data.structure.layout == ContainerLayoutPacked) { ZigType *ptr_type = instruction->base.value->type; size_t host_int_bytes = ptr_type->data.pointer.host_int_bytes; if (host_int_bytes != 0) { uint32_t size_in_bits = type_size_bits(g, ptr_type->data.pointer.child_type); LLVMTypeRef ptr_u8_type_ref = LLVMPointerType(LLVMInt8Type(), 0); LLVMValueRef u8_array_ptr = LLVMBuildBitCast(g->builder, array_ptr, ptr_u8_type_ref, ""); assert(size_in_bits % 8 == 0); LLVMValueRef elem_size_bytes = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, size_in_bits / 8, false); LLVMValueRef byte_offset = LLVMBuildNUWMul(g->builder, subscript_value, elem_size_bytes, ""); LLVMValueRef indices[] = { byte_offset }; LLVMValueRef elem_byte_ptr = LLVMBuildInBoundsGEP(g->builder, u8_array_ptr, indices, 1, ""); return LLVMBuildBitCast(g->builder, elem_byte_ptr, LLVMPointerType(get_llvm_type(g, child_type), 0), ""); } } LLVMValueRef indices[] = { LLVMConstNull(g->builtin_types.entry_usize->llvm_type), subscript_value }; return LLVMBuildInBoundsGEP(g->builder, array_ptr, indices, 2, ""); } else if (array_type->id == ZigTypeIdPointer) { LLVMValueRef array_ptr = get_handle_value(g, array_ptr_ptr, array_type, array_ptr_type); assert(LLVMGetTypeKind(LLVMTypeOf(array_ptr)) == LLVMPointerTypeKind); LLVMValueRef indices[] = { subscript_value }; return LLVMBuildInBoundsGEP(g->builder, array_ptr, indices, 1, ""); } else if (array_type->id == ZigTypeIdStruct) { LLVMValueRef array_ptr = get_handle_value(g, array_ptr_ptr, array_type, array_ptr_type); assert(array_type->data.structure.special == StructSpecialSlice); ZigType *ptr_type = array_type->data.structure.fields[slice_ptr_index]->type_entry; if (!type_has_bits(g, ptr_type)) { if (safety_check_on) { assert(LLVMGetTypeKind(LLVMTypeOf(array_ptr)) == LLVMIntegerTypeKind); add_bounds_check(g, subscript_value, LLVMIntEQ, nullptr, LLVMIntULT, array_ptr); } return nullptr; } assert(LLVMGetTypeKind(LLVMTypeOf(array_ptr)) == LLVMPointerTypeKind); assert(LLVMGetTypeKind(LLVMGetElementType(LLVMTypeOf(array_ptr))) == LLVMStructTypeKind); if (safety_check_on) { size_t len_index = array_type->data.structure.fields[slice_len_index]->gen_index; assert(len_index != SIZE_MAX); LLVMValueRef len_ptr = LLVMBuildStructGEP(g->builder, array_ptr, (unsigned)len_index, ""); LLVMValueRef len = gen_load_untyped(g, len_ptr, 0, false, ""); LLVMIntPredicate upper_op = (ptr_type->data.pointer.sentinel != nullptr) ? LLVMIntULE : LLVMIntULT; add_bounds_check(g, subscript_value, LLVMIntEQ, nullptr, upper_op, len); } size_t ptr_index = array_type->data.structure.fields[slice_ptr_index]->gen_index; assert(ptr_index != SIZE_MAX); LLVMValueRef ptr_ptr = LLVMBuildStructGEP(g->builder, array_ptr, (unsigned)ptr_index, ""); LLVMValueRef ptr = gen_load_untyped(g, ptr_ptr, 0, false, ""); return LLVMBuildInBoundsGEP(g->builder, ptr, &subscript_value, 1, ""); } else if (array_type->id == ZigTypeIdVector) { return array_ptr_ptr; } else { zig_unreachable(); } } static LLVMValueRef get_new_stack_addr(CodeGen *g, LLVMValueRef new_stack) { LLVMValueRef ptr_field_ptr = LLVMBuildStructGEP(g->builder, new_stack, (unsigned)slice_ptr_index, ""); LLVMValueRef len_field_ptr = LLVMBuildStructGEP(g->builder, new_stack, (unsigned)slice_len_index, ""); LLVMValueRef ptr_value = gen_load_untyped(g, ptr_field_ptr, 0, false, ""); LLVMValueRef len_value = gen_load_untyped(g, len_field_ptr, 0, false, ""); LLVMValueRef ptr_addr = LLVMBuildPtrToInt(g->builder, ptr_value, LLVMTypeOf(len_value), ""); LLVMValueRef end_addr = LLVMBuildNUWAdd(g->builder, ptr_addr, len_value, ""); const unsigned alignment_factor = ZigLLVMDataLayoutGetStackAlignment(g->target_data_ref); LLVMValueRef align_amt = LLVMConstInt(LLVMTypeOf(end_addr), alignment_factor, false); LLVMValueRef align_adj = LLVMBuildURem(g->builder, end_addr, align_amt, ""); return LLVMBuildNUWSub(g->builder, end_addr, align_adj, ""); } static void gen_set_stack_pointer(CodeGen *g, LLVMValueRef aligned_end_addr) { LLVMValueRef write_register_fn_val = get_write_register_fn_val(g); if (g->sp_md_node == nullptr) { Buf *sp_reg_name = buf_create_from_str(arch_stack_pointer_register_name(g->zig_target->arch)); LLVMValueRef str_node = LLVMMDString(buf_ptr(sp_reg_name), buf_len(sp_reg_name) + 1); g->sp_md_node = LLVMMDNode(&str_node, 1); } LLVMValueRef params[] = { g->sp_md_node, aligned_end_addr, }; LLVMBuildCall(g->builder, write_register_fn_val, params, 2, ""); } static void set_call_instr_sret(CodeGen *g, LLVMValueRef call_instr) { unsigned attr_kind_id = LLVMGetEnumAttributeKindForName("sret", 4); LLVMAttributeRef sret_attr = LLVMCreateEnumAttribute(LLVMGetGlobalContext(), attr_kind_id, 0); LLVMAddCallSiteAttribute(call_instr, 1, sret_attr); } static void render_async_spills(CodeGen *g) { ZigType *fn_type = g->cur_fn->type_entry; ZigType *import = get_scope_import(&g->cur_fn->fndef_scope->base); CalcLLVMFieldIndex arg_calc = {0}; frame_index_arg_calc(g, &arg_calc, fn_type->data.fn.fn_type_id.return_type); for (size_t var_i = 0; var_i < g->cur_fn->variable_list.length; var_i += 1) { ZigVar *var = g->cur_fn->variable_list.at(var_i); if (!type_has_bits(g, var->var_type)) { continue; } if (ir_get_var_is_comptime(var)) continue; switch (type_requires_comptime(g, var->var_type)) { case ReqCompTimeInvalid: zig_unreachable(); case ReqCompTimeYes: continue; case ReqCompTimeNo: break; } if (var->src_arg_index == SIZE_MAX) { continue; } calc_llvm_field_index_add(g, &arg_calc, var->var_type); var->value_ref = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, arg_calc.field_index - 1, var->name); if (var->decl_node) { var->di_loc_var = ZigLLVMCreateAutoVariable(g->dbuilder, get_di_scope(g, var->parent_scope), var->name, import->data.structure.root_struct->di_file, (unsigned)(var->decl_node->line + 1), get_llvm_di_type(g, var->var_type), !g->strip_debug_symbols, 0); gen_var_debug_decl(g, var); } } ZigType *frame_type = g->cur_fn->frame_type->data.frame.locals_struct; for (size_t alloca_i = 0; alloca_i < g->cur_fn->alloca_gen_list.length; alloca_i += 1) { IrInstGenAlloca *instruction = g->cur_fn->alloca_gen_list.at(alloca_i); if (instruction->field_index == SIZE_MAX) continue; size_t gen_index = frame_type->data.structure.fields[instruction->field_index]->gen_index; instruction->base.llvm_value = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, gen_index, instruction->name_hint); } } static void render_async_var_decls(CodeGen *g, Scope *scope) { for (;;) { switch (scope->id) { case ScopeIdCImport: zig_unreachable(); case ScopeIdFnDef: return; case ScopeIdVarDecl: { ZigVar *var = reinterpret_cast(scope)->var; if (var->did_the_decl_codegen) { render_decl_var(g, var); } } ZIG_FALLTHROUGH; case ScopeIdDecls: case ScopeIdBlock: case ScopeIdDefer: case ScopeIdDeferExpr: case ScopeIdLoop: case ScopeIdSuspend: case ScopeIdCompTime: case ScopeIdNoSuspend: case ScopeIdRuntime: case ScopeIdTypeOf: case ScopeIdExpr: scope = scope->parent; continue; } } } static LLVMValueRef gen_frame_size(CodeGen *g, LLVMValueRef fn_val) { assert(g->need_frame_size_prefix_data); LLVMTypeRef usize_llvm_type = g->builtin_types.entry_usize->llvm_type; LLVMTypeRef ptr_usize_llvm_type = LLVMPointerType(usize_llvm_type, 0); LLVMValueRef casted_fn_val = LLVMBuildBitCast(g->builder, fn_val, ptr_usize_llvm_type, ""); LLVMValueRef negative_one = LLVMConstInt(LLVMInt32Type(), -1, true); LLVMValueRef prefix_ptr = LLVMBuildInBoundsGEP(g->builder, casted_fn_val, &negative_one, 1, ""); return LLVMBuildLoad(g->builder, prefix_ptr, ""); } static void gen_init_stack_trace(CodeGen *g, LLVMValueRef trace_field_ptr, LLVMValueRef addrs_field_ptr) { LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMValueRef zero = LLVMConstNull(usize_type_ref); LLVMValueRef index_ptr = LLVMBuildStructGEP(g->builder, trace_field_ptr, 0, ""); LLVMBuildStore(g->builder, zero, index_ptr); LLVMValueRef addrs_slice_ptr = LLVMBuildStructGEP(g->builder, trace_field_ptr, 1, ""); LLVMValueRef addrs_ptr_ptr = LLVMBuildStructGEP(g->builder, addrs_slice_ptr, slice_ptr_index, ""); LLVMValueRef indices[] = { LLVMConstNull(usize_type_ref), LLVMConstNull(usize_type_ref) }; LLVMValueRef trace_field_addrs_as_ptr = LLVMBuildInBoundsGEP(g->builder, addrs_field_ptr, indices, 2, ""); LLVMBuildStore(g->builder, trace_field_addrs_as_ptr, addrs_ptr_ptr); LLVMValueRef addrs_len_ptr = LLVMBuildStructGEP(g->builder, addrs_slice_ptr, slice_len_index, ""); LLVMBuildStore(g->builder, LLVMConstInt(usize_type_ref, stack_trace_ptr_count, false), addrs_len_ptr); } static LLVMValueRef ir_render_call(CodeGen *g, IrExecutableGen *executable, IrInstGenCall *instruction) { Error err; LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMValueRef fn_val; ZigType *fn_type; bool callee_is_async; if (instruction->fn_entry) { fn_val = fn_llvm_value(g, instruction->fn_entry); fn_type = instruction->fn_entry->type_entry; callee_is_async = fn_is_async(instruction->fn_entry); } else { assert(instruction->fn_ref); fn_val = ir_llvm_value(g, instruction->fn_ref); fn_type = instruction->fn_ref->value->type; callee_is_async = fn_type->data.fn.fn_type_id.cc == CallingConventionAsync; } FnTypeId *fn_type_id = &fn_type->data.fn.fn_type_id; ZigType *src_return_type = fn_type_id->return_type; bool ret_has_bits = type_has_bits(g, src_return_type); CallingConvention cc = fn_type->data.fn.fn_type_id.cc; bool first_arg_ret = ret_has_bits && want_first_arg_sret(g, fn_type_id); bool prefix_arg_err_ret_stack = codegen_fn_has_err_ret_tracing_arg(g, fn_type_id->return_type); bool is_var_args = fn_type_id->is_var_args; ZigList gen_param_values = {}; ZigList gen_param_types = {}; LLVMValueRef result_loc = instruction->result_loc ? ir_llvm_value(g, instruction->result_loc) : nullptr; LLVMValueRef zero = LLVMConstNull(usize_type_ref); bool need_frame_ptr_ptr_spill = false; ZigType *anyframe_type = nullptr; LLVMValueRef frame_result_loc_uncasted = nullptr; LLVMValueRef frame_result_loc; LLVMValueRef awaiter_init_val; LLVMValueRef ret_ptr; if (callee_is_async) { if (instruction->new_stack == nullptr) { if (instruction->modifier == CallModifierAsync) { frame_result_loc = result_loc; } else { ir_assert(instruction->frame_result_loc != nullptr, &instruction->base); frame_result_loc_uncasted = ir_llvm_value(g, instruction->frame_result_loc); ir_assert(instruction->fn_entry != nullptr, &instruction->base); frame_result_loc = LLVMBuildBitCast(g->builder, frame_result_loc_uncasted, LLVMPointerType(get_llvm_type(g, instruction->fn_entry->frame_type), 0), ""); } } else { if (instruction->new_stack->value->type->id == ZigTypeIdPointer && instruction->new_stack->value->type->data.pointer.child_type->id == ZigTypeIdFnFrame) { frame_result_loc = ir_llvm_value(g, instruction->new_stack); } else { LLVMValueRef frame_slice_ptr = ir_llvm_value(g, instruction->new_stack); if (ir_want_runtime_safety(g, &instruction->base)) { LLVMValueRef given_len_ptr = LLVMBuildStructGEP(g->builder, frame_slice_ptr, slice_len_index, ""); LLVMValueRef given_frame_len = LLVMBuildLoad(g->builder, given_len_ptr, ""); LLVMValueRef actual_frame_len = gen_frame_size(g, fn_val); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "FrameSizeCheckFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "FrameSizeCheckOk"); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntUGE, given_frame_len, actual_frame_len, ""); LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdFrameTooSmall); LLVMPositionBuilderAtEnd(g->builder, ok_block); } need_frame_ptr_ptr_spill = true; LLVMValueRef frame_ptr_ptr = LLVMBuildStructGEP(g->builder, frame_slice_ptr, slice_ptr_index, ""); LLVMValueRef frame_ptr = LLVMBuildLoad(g->builder, frame_ptr_ptr, ""); if (instruction->fn_entry == nullptr) { anyframe_type = get_any_frame_type(g, src_return_type); frame_result_loc = LLVMBuildBitCast(g->builder, frame_ptr, get_llvm_type(g, anyframe_type), ""); } else { ZigType *frame_type = get_fn_frame_type(g, instruction->fn_entry); if ((err = type_resolve(g, frame_type, ResolveStatusLLVMFull))) codegen_report_errors_and_exit(g); ZigType *ptr_frame_type = get_pointer_to_type(g, frame_type, false); frame_result_loc = LLVMBuildBitCast(g->builder, frame_ptr, get_llvm_type(g, ptr_frame_type), ""); } } } if (instruction->modifier == CallModifierAsync) { if (instruction->new_stack == nullptr) { awaiter_init_val = zero; if (ret_has_bits) { // Use the result location which is inside the frame if this is an async call. ret_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start + 2, ""); } } else { awaiter_init_val = zero; if (ret_has_bits) { if (result_loc != nullptr) { // Use the result location provided to the @asyncCall builtin ret_ptr = result_loc; } else { // no result location provided to @asyncCall - use the one inside the frame. ret_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start + 2, ""); } } } // even if prefix_arg_err_ret_stack is true, let the async function do its own // initialization. } else { if (instruction->modifier == CallModifierNoSuspend && !fn_is_async(g->cur_fn)) { // Async function called as a normal function, and calling function is not async. // This is allowed because it was called with `nosuspend` which asserts that it will // never suspend. awaiter_init_val = zero; } else { // async function called as a normal function awaiter_init_val = LLVMBuildPtrToInt(g->builder, g->cur_frame_ptr, usize_type_ref, ""); // caller's own frame pointer } if (ret_has_bits) { if (result_loc == nullptr) { // return type is a scalar, but we still need a pointer to it. Use the async fn frame. ret_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start + 2, ""); } else { // Use the call instruction's result location. ret_ptr = result_loc; } // Store a zero in the awaiter's result ptr to indicate we do not need a copy made. LLVMValueRef awaiter_ret_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start + 1, ""); LLVMValueRef zero_ptr = LLVMConstNull(LLVMGetElementType(LLVMTypeOf(awaiter_ret_ptr))); LLVMBuildStore(g->builder, zero_ptr, awaiter_ret_ptr); } if (prefix_arg_err_ret_stack) { LLVMValueRef err_ret_trace_ptr_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_index_trace_arg(g, src_return_type) + 1, ""); bool is_llvm_alloca; LLVMValueRef my_err_ret_trace_val = get_cur_err_ret_trace_val(g, instruction->base.base.scope, &is_llvm_alloca); LLVMBuildStore(g->builder, my_err_ret_trace_val, err_ret_trace_ptr_ptr); } } assert(frame_result_loc != nullptr); LLVMValueRef fn_ptr_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_fn_ptr_index, ""); LLVMValueRef bitcasted_fn_val = LLVMBuildBitCast(g->builder, fn_val, LLVMGetElementType(LLVMTypeOf(fn_ptr_ptr)), ""); LLVMBuildStore(g->builder, bitcasted_fn_val, fn_ptr_ptr); LLVMValueRef resume_index_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_resume_index, ""); LLVMBuildStore(g->builder, zero, resume_index_ptr); LLVMValueRef awaiter_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_awaiter_index, ""); LLVMBuildStore(g->builder, awaiter_init_val, awaiter_ptr); if (ret_has_bits) { LLVMValueRef ret_ptr_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start, ""); LLVMBuildStore(g->builder, ret_ptr, ret_ptr_ptr); } } else if (instruction->modifier == CallModifierAsync) { // Async call of blocking function if (instruction->new_stack != nullptr) { zig_panic("TODO @asyncCall of non-async function"); } frame_result_loc = result_loc; awaiter_init_val = LLVMConstAllOnes(usize_type_ref); LLVMValueRef awaiter_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_awaiter_index, ""); LLVMBuildStore(g->builder, awaiter_init_val, awaiter_ptr); if (ret_has_bits) { ret_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start + 2, ""); LLVMValueRef ret_ptr_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start, ""); LLVMBuildStore(g->builder, ret_ptr, ret_ptr_ptr); if (first_arg_ret) { gen_param_values.append(ret_ptr); } if (prefix_arg_err_ret_stack) { // Set up the callee stack trace pointer pointing into the frame. // Then we have to wire up the StackTrace pointers. // Await is responsible for merging error return traces. uint32_t trace_field_index_start = frame_index_trace_arg(g, src_return_type); LLVMValueRef callee_trace_ptr_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, trace_field_index_start, ""); LLVMValueRef trace_field_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, trace_field_index_start + 2, ""); LLVMValueRef addrs_field_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, trace_field_index_start + 3, ""); LLVMBuildStore(g->builder, trace_field_ptr, callee_trace_ptr_ptr); gen_init_stack_trace(g, trace_field_ptr, addrs_field_ptr); bool is_llvm_alloca; gen_param_values.append(get_cur_err_ret_trace_val(g, instruction->base.base.scope, &is_llvm_alloca)); } } } else { if (first_arg_ret) { gen_param_values.append(result_loc); } if (prefix_arg_err_ret_stack) { bool is_llvm_alloca; gen_param_values.append(get_cur_err_ret_trace_val(g, instruction->base.base.scope, &is_llvm_alloca)); } } FnWalk fn_walk = {}; fn_walk.id = FnWalkIdCall; fn_walk.data.call.inst = instruction; fn_walk.data.call.is_var_args = is_var_args; fn_walk.data.call.gen_param_values = &gen_param_values; fn_walk.data.call.gen_param_types = &gen_param_types; walk_function_params(g, fn_type, &fn_walk); ZigLLVM_CallAttr call_attr; switch (instruction->modifier) { case CallModifierBuiltin: case CallModifierCompileTime: zig_unreachable(); case CallModifierNone: case CallModifierNoSuspend: case CallModifierAsync: call_attr = ZigLLVM_CallAttrAuto; break; case CallModifierNeverTail: call_attr = ZigLLVM_CallAttrNeverTail; break; case CallModifierNeverInline: call_attr = ZigLLVM_CallAttrNeverInline; break; case CallModifierAlwaysTail: call_attr = ZigLLVM_CallAttrAlwaysTail; break; case CallModifierAlwaysInline: ir_assert(instruction->fn_entry != nullptr, &instruction->base); call_attr = ZigLLVM_CallAttrAlwaysInline; break; } ZigLLVM_CallingConv llvm_cc = get_llvm_cc(g, cc); LLVMValueRef result; if (callee_is_async) { CalcLLVMFieldIndex arg_calc_start = {0}; frame_index_arg_calc(g, &arg_calc_start, fn_type->data.fn.fn_type_id.return_type); LLVMValueRef casted_frame; if (instruction->new_stack != nullptr && instruction->fn_entry == nullptr) { // We need the frame type to be a pointer to a struct that includes the args // Count ahead to determine how many llvm struct fields we need. CalcLLVMFieldIndex arg_calc = arg_calc_start; for (size_t i = 0; i < gen_param_types.length; i += 1) { calc_llvm_field_index_add(g, &arg_calc, gen_param_types.at(i)); } size_t field_count = arg_calc.field_index; LLVMTypeRef *field_types = heap::c_allocator.allocate_nonzero(field_count); LLVMGetStructElementTypes(LLVMGetElementType(LLVMTypeOf(frame_result_loc)), field_types); assert(LLVMCountStructElementTypes(LLVMGetElementType(LLVMTypeOf(frame_result_loc))) == arg_calc_start.field_index); arg_calc = arg_calc_start; for (size_t arg_i = 0; arg_i < gen_param_values.length; arg_i += 1) { CalcLLVMFieldIndex prev = arg_calc; // Use the declared argument type and not the value one to be // consistent with the assignment operation below. calc_llvm_field_index_add(g, &arg_calc, gen_param_types.at(arg_i)); field_types[arg_calc.field_index - 1] = get_llvm_type(g, gen_param_types.at(arg_i)); if (arg_calc.field_index - prev.field_index > 1) { // Padding field uint32_t pad_bytes = arg_calc.offset - prev.offset - gen_param_types.at(arg_i)->abi_size; LLVMTypeRef pad_llvm_type = LLVMArrayType(LLVMInt8Type(), pad_bytes); field_types[arg_calc.field_index - 2] = pad_llvm_type; } } LLVMTypeRef frame_with_args_type = LLVMStructType(field_types, field_count, false); heap::c_allocator.deallocate(field_types, field_count); LLVMTypeRef ptr_frame_with_args_type = LLVMPointerType(frame_with_args_type, 0); casted_frame = LLVMBuildBitCast(g->builder, frame_result_loc, ptr_frame_with_args_type, ""); } else { casted_frame = frame_result_loc; } CalcLLVMFieldIndex arg_calc = arg_calc_start; for (size_t arg_i = 0; arg_i < gen_param_values.length; arg_i += 1) { calc_llvm_field_index_add(g, &arg_calc, gen_param_types.at(arg_i)); LLVMValueRef arg_ptr = LLVMBuildStructGEP(g->builder, casted_frame, arg_calc.field_index - 1, ""); gen_assign_raw(g, arg_ptr, get_pointer_to_type(g, gen_param_types.at(arg_i), true), gen_param_values.at(arg_i)); } gen_param_types.deinit(); if (instruction->modifier == CallModifierAsync) { gen_resume(g, fn_val, frame_result_loc, ResumeIdCall); if (instruction->new_stack != nullptr) { return LLVMBuildBitCast(g->builder, frame_result_loc, get_llvm_type(g, instruction->base.value->type), ""); } return nullptr; } else if (instruction->modifier == CallModifierNoSuspend && !fn_is_async(g->cur_fn)) { gen_resume(g, fn_val, frame_result_loc, ResumeIdCall); if (ir_want_runtime_safety(g, &instruction->base)) { LLVMValueRef awaiter_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_awaiter_index, ""); LLVMValueRef all_ones = LLVMConstAllOnes(usize_type_ref); LLVMValueRef prev_val = gen_maybe_atomic_op(g, LLVMAtomicRMWBinOpXchg, awaiter_ptr, all_ones, LLVMAtomicOrderingRelease); LLVMValueRef ok_val = LLVMBuildICmp(g->builder, LLVMIntEQ, prev_val, all_ones, ""); LLVMBasicBlockRef bad_block = LLVMAppendBasicBlock(g->cur_fn_val, "NoSuspendPanic"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "NoSuspendOk"); LLVMBuildCondBr(g->builder, ok_val, ok_block, bad_block); // The async function suspended, but this nosuspend call asserted it wouldn't. LLVMPositionBuilderAtEnd(g->builder, bad_block); gen_safety_crash(g, PanicMsgIdBadNoSuspendCall); LLVMPositionBuilderAtEnd(g->builder, ok_block); } ZigType *result_type = instruction->base.value->type; ZigType *ptr_result_type = get_pointer_to_type(g, result_type, true); return gen_await_early_return(g, &instruction->base, frame_result_loc, result_type, ptr_result_type, result_loc, true); } else { ZigType *ptr_result_type = get_pointer_to_type(g, src_return_type, true); LLVMBasicBlockRef call_bb = gen_suspend_begin(g, "CallResume"); LLVMValueRef call_inst = gen_resume(g, fn_val, frame_result_loc, ResumeIdCall); set_tail_call_if_appropriate(g, call_inst); LLVMBuildRetVoid(g->builder); LLVMPositionBuilderAtEnd(g->builder, call_bb); gen_assert_resume_id(g, &instruction->base, ResumeIdReturn, PanicMsgIdResumedAnAwaitingFn, nullptr); render_async_var_decls(g, instruction->base.base.scope); if (!type_has_bits(g, src_return_type)) return nullptr; if (result_loc != nullptr) { if (instruction->result_loc->id == IrInstGenIdReturnPtr) { instruction->base.spill = nullptr; return g->cur_ret_ptr; } else { return get_handle_value(g, result_loc, src_return_type, ptr_result_type); } } if (need_frame_ptr_ptr_spill) { LLVMValueRef frame_slice_ptr = ir_llvm_value(g, instruction->new_stack); LLVMValueRef frame_ptr_ptr = LLVMBuildStructGEP(g->builder, frame_slice_ptr, slice_ptr_index, ""); frame_result_loc_uncasted = LLVMBuildLoad(g->builder, frame_ptr_ptr, ""); } if (frame_result_loc_uncasted != nullptr) { if (instruction->fn_entry != nullptr) { frame_result_loc = LLVMBuildBitCast(g->builder, frame_result_loc_uncasted, LLVMPointerType(get_llvm_type(g, instruction->fn_entry->frame_type), 0), ""); } else { frame_result_loc = LLVMBuildBitCast(g->builder, frame_result_loc_uncasted, get_llvm_type(g, anyframe_type), ""); } } LLVMValueRef result_ptr = LLVMBuildStructGEP(g->builder, frame_result_loc, frame_ret_start + 2, ""); return LLVMBuildLoad(g->builder, result_ptr, ""); } } else { gen_param_types.deinit(); } if (instruction->new_stack == nullptr || instruction->is_async_call_builtin) { result = ZigLLVMBuildCall(g->builder, fn_val, gen_param_values.items, (unsigned)gen_param_values.length, llvm_cc, call_attr, ""); } else if (instruction->modifier == CallModifierAsync) { zig_panic("TODO @asyncCall of non-async function"); } else { LLVMValueRef new_stack_addr = get_new_stack_addr(g, ir_llvm_value(g, instruction->new_stack)); LLVMValueRef old_stack_ref; if (src_return_type->id != ZigTypeIdUnreachable) { LLVMValueRef stacksave_fn_val = get_stacksave_fn_val(g); old_stack_ref = LLVMBuildCall(g->builder, stacksave_fn_val, nullptr, 0, ""); } gen_set_stack_pointer(g, new_stack_addr); result = ZigLLVMBuildCall(g->builder, fn_val, gen_param_values.items, (unsigned)gen_param_values.length, llvm_cc, call_attr, ""); if (src_return_type->id != ZigTypeIdUnreachable) { LLVMValueRef stackrestore_fn_val = get_stackrestore_fn_val(g); LLVMBuildCall(g->builder, stackrestore_fn_val, &old_stack_ref, 1, ""); } } if (src_return_type->id == ZigTypeIdUnreachable) { return LLVMBuildUnreachable(g->builder); } else if (!ret_has_bits) { return nullptr; } else if (first_arg_ret) { set_call_instr_sret(g, result); return result_loc; } else if (handle_is_ptr(g, src_return_type)) { LLVMValueRef store_instr = LLVMBuildStore(g->builder, result, result_loc); LLVMSetAlignment(store_instr, get_ptr_align(g, instruction->result_loc->value->type)); return result_loc; } else if (!callee_is_async && instruction->modifier == CallModifierAsync) { LLVMBuildStore(g->builder, result, ret_ptr); return result_loc; } else { return result; } } static LLVMValueRef ir_render_struct_field_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenStructFieldPtr *instruction) { Error err; if (instruction->base.value->special != ConstValSpecialRuntime) return nullptr; LLVMValueRef struct_ptr = ir_llvm_value(g, instruction->struct_ptr); // not necessarily a pointer. could be ZigTypeIdStruct ZigType *struct_ptr_type = instruction->struct_ptr->value->type; TypeStructField *field = instruction->field; if (!type_has_bits(g, field->type_entry)) return nullptr; if (struct_ptr_type->id == ZigTypeIdPointer && struct_ptr_type->data.pointer.host_int_bytes != 0) { return struct_ptr; } ZigType *struct_type; if (struct_ptr_type->id == ZigTypeIdPointer) { if (struct_ptr_type->data.pointer.inferred_struct_field != nullptr) { struct_type = struct_ptr_type->data.pointer.inferred_struct_field->inferred_struct_type; } else { struct_type = struct_ptr_type->data.pointer.child_type; } } else { struct_type = struct_ptr_type; } if ((err = type_resolve(g, struct_type, ResolveStatusLLVMFull))) codegen_report_errors_and_exit(g); ir_assert(field->gen_index != SIZE_MAX, &instruction->base); LLVMValueRef field_ptr_val = LLVMBuildStructGEP(g->builder, struct_ptr, (unsigned)field->gen_index, ""); ZigType *res_type = instruction->base.value->type; ir_assert(res_type->id == ZigTypeIdPointer, &instruction->base); if (res_type->data.pointer.host_int_bytes != 0) { // We generate packed structs with get_llvm_type_of_n_bytes, which is // u8 for 1 byte or [n]u8 for multiple bytes. But the pointer to the type // is supposed to be a pointer to the integer. So we bitcast it here. LLVMTypeRef int_elem_type = LLVMIntType(8*res_type->data.pointer.host_int_bytes); LLVMTypeRef integer_ptr_type = LLVMPointerType(int_elem_type, 0); return LLVMBuildBitCast(g->builder, field_ptr_val, integer_ptr_type, ""); } return field_ptr_val; } static LLVMValueRef ir_render_union_field_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenUnionFieldPtr *instruction) { if (instruction->base.value->special != ConstValSpecialRuntime) return nullptr; ZigType *union_ptr_type = instruction->union_ptr->value->type; assert(union_ptr_type->id == ZigTypeIdPointer); ZigType *union_type = union_ptr_type->data.pointer.child_type; assert(union_type->id == ZigTypeIdUnion); TypeUnionField *field = instruction->field; if (!type_has_bits(g, field->type_entry)) { ZigType *tag_type = union_type->data.unionation.tag_type; if (!instruction->initializing || tag_type == nullptr || !type_has_bits(g, tag_type)) return nullptr; // The field has no bits but we still have to change the discriminant // value here LLVMValueRef union_ptr = ir_llvm_value(g, instruction->union_ptr); LLVMTypeRef tag_type_ref = get_llvm_type(g, tag_type); LLVMValueRef tag_field_ptr = nullptr; if (union_type->data.unionation.gen_field_count == 0) { assert(union_type->data.unionation.gen_tag_index == SIZE_MAX); // The whole union is collapsed into the discriminant tag_field_ptr = LLVMBuildBitCast(g->builder, union_ptr, LLVMPointerType(tag_type_ref, 0), ""); } else { assert(union_type->data.unionation.gen_tag_index != SIZE_MAX); tag_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, union_type->data.unionation.gen_tag_index, ""); } LLVMValueRef tag_value = bigint_to_llvm_const(tag_type_ref, &field->enum_field->value); assert(tag_field_ptr != nullptr); gen_store_untyped(g, tag_value, tag_field_ptr, 0, false); return nullptr; } LLVMValueRef union_ptr = ir_llvm_value(g, instruction->union_ptr); LLVMTypeRef field_type_ref = LLVMPointerType(get_llvm_type(g, field->type_entry), 0); if (union_type->data.unionation.gen_tag_index == SIZE_MAX) { LLVMValueRef union_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, 0, ""); LLVMValueRef bitcasted_union_field_ptr = LLVMBuildBitCast(g->builder, union_field_ptr, field_type_ref, ""); return bitcasted_union_field_ptr; } if (instruction->initializing) { LLVMValueRef tag_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, union_type->data.unionation.gen_tag_index, ""); LLVMValueRef tag_value = bigint_to_llvm_const(get_llvm_type(g, union_type->data.unionation.tag_type), &field->enum_field->value); gen_store_untyped(g, tag_value, tag_field_ptr, 0, false); } else if (instruction->safety_check_on && ir_want_runtime_safety(g, &instruction->base)) { LLVMValueRef tag_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, union_type->data.unionation.gen_tag_index, ""); LLVMValueRef tag_value = gen_load_untyped(g, tag_field_ptr, 0, false, ""); LLVMValueRef expected_tag_value = bigint_to_llvm_const(get_llvm_type(g, union_type->data.unionation.tag_type), &field->enum_field->value); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnionCheckOk"); LLVMBasicBlockRef bad_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnionCheckFail"); LLVMValueRef ok_val = LLVMBuildICmp(g->builder, LLVMIntEQ, tag_value, expected_tag_value, ""); LLVMBuildCondBr(g->builder, ok_val, ok_block, bad_block); LLVMPositionBuilderAtEnd(g->builder, bad_block); gen_safety_crash(g, PanicMsgIdBadUnionField); LLVMPositionBuilderAtEnd(g->builder, ok_block); } LLVMValueRef union_field_ptr = LLVMBuildStructGEP(g->builder, union_ptr, union_type->data.unionation.gen_union_index, ""); LLVMValueRef bitcasted_union_field_ptr = LLVMBuildBitCast(g->builder, union_field_ptr, field_type_ref, ""); return bitcasted_union_field_ptr; } static size_t find_asm_index(CodeGen *g, AstNode *node, AsmToken *tok, Buf *src_template) { const char *ptr = buf_ptr(src_template) + tok->start + 2; size_t len = tok->end - tok->start - 2; size_t result = 0; for (size_t i = 0; i < node->data.asm_expr.output_list.length; i += 1, result += 1) { AsmOutput *asm_output = node->data.asm_expr.output_list.at(i); if (buf_eql_mem(asm_output->asm_symbolic_name, ptr, len)) { return result; } } for (size_t i = 0; i < node->data.asm_expr.input_list.length; i += 1, result += 1) { AsmInput *asm_input = node->data.asm_expr.input_list.at(i); if (buf_eql_mem(asm_input->asm_symbolic_name, ptr, len)) { return result; } } return SIZE_MAX; } static LLVMValueRef ir_render_asm_gen(CodeGen *g, IrExecutableGen *executable, IrInstGenAsm *instruction) { AstNode *asm_node = instruction->base.base.source_node; assert(asm_node->type == NodeTypeAsmExpr); AstNodeAsmExpr *asm_expr = &asm_node->data.asm_expr; Buf *src_template = instruction->asm_template; Buf llvm_template = BUF_INIT; buf_resize(&llvm_template, 0); for (size_t token_i = 0; token_i < instruction->token_list_len; token_i += 1) { AsmToken *asm_token = &instruction->token_list[token_i]; switch (asm_token->id) { case AsmTokenIdTemplate: for (size_t offset = asm_token->start; offset < asm_token->end; offset += 1) { uint8_t c = *((uint8_t*)(buf_ptr(src_template) + offset)); if (c == '$') { buf_append_str(&llvm_template, "$$"); } else { buf_append_char(&llvm_template, c); } } break; case AsmTokenIdPercent: buf_append_char(&llvm_template, '%'); break; case AsmTokenIdVar: { size_t index = find_asm_index(g, asm_node, asm_token, src_template); assert(index < SIZE_MAX); buf_appendf(&llvm_template, "$%" ZIG_PRI_usize "", index); break; } case AsmTokenIdUniqueId: buf_append_str(&llvm_template, "${:uid}"); break; } } Buf constraint_buf = BUF_INIT; buf_resize(&constraint_buf, 0); assert(instruction->return_count == 0 || instruction->return_count == 1); size_t total_constraint_count = asm_expr->output_list.length + asm_expr->input_list.length + asm_expr->clobber_list.length; size_t input_and_output_count = asm_expr->output_list.length + asm_expr->input_list.length - instruction->return_count; size_t total_index = 0; size_t param_index = 0; LLVMTypeRef *param_types = heap::c_allocator.allocate(input_and_output_count); LLVMValueRef *param_values = heap::c_allocator.allocate(input_and_output_count); for (size_t i = 0; i < asm_expr->output_list.length; i += 1, total_index += 1) { AsmOutput *asm_output = asm_expr->output_list.at(i); bool is_return = (asm_output->return_type != nullptr); assert(*buf_ptr(asm_output->constraint) == '='); // LLVM uses commas internally to separate different constraints, // alternative constraints are achieved with pipes. // We still allow the user to use commas in a way that is similar // to GCC's inline assembly. // http://llvm.org/docs/LangRef.html#constraint-codes buf_replace(asm_output->constraint, ',', '|'); if (is_return) { buf_appendf(&constraint_buf, "=%s", buf_ptr(asm_output->constraint) + 1); } else { buf_appendf(&constraint_buf, "=*%s", buf_ptr(asm_output->constraint) + 1); } if (total_index + 1 < total_constraint_count) { buf_append_char(&constraint_buf, ','); } if (!is_return) { ZigVar *variable = instruction->output_vars[i]; assert(variable); param_types[param_index] = LLVMTypeOf(variable->value_ref); param_values[param_index] = variable->value_ref; param_index += 1; } } for (size_t i = 0; i < asm_expr->input_list.length; i += 1, total_index += 1, param_index += 1) { AsmInput *asm_input = asm_expr->input_list.at(i); buf_replace(asm_input->constraint, ',', '|'); IrInstGen *ir_input = instruction->input_list[i]; buf_append_buf(&constraint_buf, asm_input->constraint); if (total_index + 1 < total_constraint_count) { buf_append_char(&constraint_buf, ','); } ZigType *const type = ir_input->value->type; LLVMTypeRef type_ref = get_llvm_type(g, type); LLVMValueRef value_ref = ir_llvm_value(g, ir_input); // Handle integers of non pot bitsize by widening them. if (type->id == ZigTypeIdInt) { const size_t bitsize = type->data.integral.bit_count; if (bitsize < 8 || !is_power_of_2(bitsize)) { const bool is_signed = type->data.integral.is_signed; const size_t wider_bitsize = bitsize < 8 ? 8 : round_to_next_power_of_2(bitsize); ZigType *const wider_type = get_int_type(g, is_signed, wider_bitsize); type_ref = get_llvm_type(g, wider_type); value_ref = gen_widen_or_shorten(g, false, type, wider_type, value_ref); } } param_types[param_index] = type_ref; param_values[param_index] = value_ref; } for (size_t i = 0; i < asm_expr->clobber_list.length; i += 1, total_index += 1) { Buf *clobber_buf = asm_expr->clobber_list.at(i); buf_appendf(&constraint_buf, "~{%s}", buf_ptr(clobber_buf)); if (total_index + 1 < total_constraint_count) { buf_append_char(&constraint_buf, ','); } } // Add some architecture-specific clobbers. const char *arch_clobbers = nullptr; switch (g->zig_target->arch) { case ZigLLVM_x86: case ZigLLVM_x86_64: arch_clobbers = "~{dirflag},~{fpsr},~{flags}"; break; case ZigLLVM_mips: case ZigLLVM_mipsel: case ZigLLVM_mips64: case ZigLLVM_mips64el: arch_clobbers = "~{$1}"; break; default: break; } if (arch_clobbers != nullptr) { if (buf_len(&constraint_buf)) buf_append_char(&constraint_buf, ','); buf_append_str(&constraint_buf, arch_clobbers); } LLVMTypeRef ret_type; if (instruction->return_count == 0) { ret_type = LLVMVoidType(); } else { ret_type = get_llvm_type(g, instruction->base.value->type); } LLVMTypeRef function_type = LLVMFunctionType(ret_type, param_types, (unsigned)input_and_output_count, false); heap::c_allocator.deallocate(param_types, input_and_output_count); bool is_volatile = instruction->has_side_effects || (asm_expr->output_list.length == 0); LLVMValueRef asm_fn = LLVMGetInlineAsm(function_type, buf_ptr(&llvm_template), buf_len(&llvm_template), buf_ptr(&constraint_buf), buf_len(&constraint_buf), is_volatile, false, LLVMInlineAsmDialectATT); LLVMValueRef built_call = LLVMBuildCall(g->builder, asm_fn, param_values, (unsigned)input_and_output_count, ""); heap::c_allocator.deallocate(param_values, input_and_output_count); return built_call; } static LLVMValueRef gen_non_null_bit(CodeGen *g, ZigType *maybe_type, LLVMValueRef maybe_handle) { assert(maybe_type->id == ZigTypeIdOptional || (maybe_type->id == ZigTypeIdPointer && maybe_type->data.pointer.allow_zero)); ZigType *child_type = maybe_type->data.maybe.child_type; if (!type_has_bits(g, child_type)) return maybe_handle; bool is_scalar = !handle_is_ptr(g, maybe_type); if (is_scalar) return LLVMBuildICmp(g->builder, LLVMIntNE, maybe_handle, LLVMConstNull(get_llvm_type(g, maybe_type)), ""); LLVMValueRef maybe_field_ptr = LLVMBuildStructGEP(g->builder, maybe_handle, maybe_null_index, ""); return gen_load_untyped(g, maybe_field_ptr, 0, false, ""); } static LLVMValueRef ir_render_test_non_null(CodeGen *g, IrExecutableGen *executable, IrInstGenTestNonNull *instruction) { return gen_non_null_bit(g, instruction->value->value->type, ir_llvm_value(g, instruction->value)); } static LLVMValueRef ir_render_optional_unwrap_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenOptionalUnwrapPtr *instruction) { if (instruction->base.value->special != ConstValSpecialRuntime) return nullptr; ZigType *ptr_type = instruction->base_ptr->value->type; assert(ptr_type->id == ZigTypeIdPointer); ZigType *maybe_type = ptr_type->data.pointer.child_type; assert(maybe_type->id == ZigTypeIdOptional); ZigType *child_type = maybe_type->data.maybe.child_type; LLVMValueRef base_ptr = ir_llvm_value(g, instruction->base_ptr); if (instruction->safety_check_on && ir_want_runtime_safety(g, &instruction->base)) { LLVMValueRef maybe_handle = get_handle_value(g, base_ptr, maybe_type, ptr_type); LLVMValueRef non_null_bit = gen_non_null_bit(g, maybe_type, maybe_handle); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnwrapOptionalFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnwrapOptionalOk"); LLVMBuildCondBr(g->builder, non_null_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdUnwrapOptionalFail); LLVMPositionBuilderAtEnd(g->builder, ok_block); } if (!type_has_bits(g, child_type)) { if (instruction->initializing) { LLVMValueRef non_null_bit = LLVMConstInt(LLVMInt1Type(), 1, false); gen_store_untyped(g, non_null_bit, base_ptr, 0, false); } return nullptr; } else { bool is_scalar = !handle_is_ptr(g, maybe_type); if (is_scalar) { return base_ptr; } else { LLVMValueRef optional_struct_ref = get_handle_value(g, base_ptr, maybe_type, ptr_type); if (instruction->initializing) { LLVMValueRef non_null_bit_ptr = LLVMBuildStructGEP(g->builder, optional_struct_ref, maybe_null_index, ""); LLVMValueRef non_null_bit = LLVMConstInt(LLVMInt1Type(), 1, false); gen_store_untyped(g, non_null_bit, non_null_bit_ptr, 0, false); } return LLVMBuildStructGEP(g->builder, optional_struct_ref, maybe_child_index, ""); } } } static LLVMValueRef get_int_builtin_fn(CodeGen *g, ZigType *expr_type, BuiltinFnId fn_id) { bool is_vector = expr_type->id == ZigTypeIdVector; ZigType *int_type = is_vector ? expr_type->data.vector.elem_type : expr_type; assert(int_type->id == ZigTypeIdInt); uint32_t vector_len = is_vector ? expr_type->data.vector.len : 0; ZigLLVMFnKey key = {}; const char *fn_name; uint32_t n_args; if (fn_id == BuiltinFnIdCtz) { fn_name = "cttz"; n_args = 2; key.id = ZigLLVMFnIdCtz; key.data.ctz.bit_count = (uint32_t)int_type->data.integral.bit_count; } else if (fn_id == BuiltinFnIdClz) { fn_name = "ctlz"; n_args = 2; key.id = ZigLLVMFnIdClz; key.data.clz.bit_count = (uint32_t)int_type->data.integral.bit_count; } else if (fn_id == BuiltinFnIdPopCount) { fn_name = "ctpop"; n_args = 1; key.id = ZigLLVMFnIdPopCount; key.data.pop_count.bit_count = (uint32_t)int_type->data.integral.bit_count; } else if (fn_id == BuiltinFnIdBswap) { fn_name = "bswap"; n_args = 1; key.id = ZigLLVMFnIdBswap; key.data.bswap.bit_count = (uint32_t)int_type->data.integral.bit_count; key.data.bswap.vector_len = vector_len; } else if (fn_id == BuiltinFnIdBitReverse) { fn_name = "bitreverse"; n_args = 1; key.id = ZigLLVMFnIdBitReverse; key.data.bit_reverse.bit_count = (uint32_t)int_type->data.integral.bit_count; } else { zig_unreachable(); } auto existing_entry = g->llvm_fn_table.maybe_get(key); if (existing_entry) return existing_entry->value; char llvm_name[64]; if (is_vector) sprintf(llvm_name, "llvm.%s.v%" PRIu32 "i%" PRIu32, fn_name, vector_len, int_type->data.integral.bit_count); else sprintf(llvm_name, "llvm.%s.i%" PRIu32, fn_name, int_type->data.integral.bit_count); LLVMTypeRef param_types[] = { get_llvm_type(g, expr_type), LLVMInt1Type(), }; LLVMTypeRef fn_type = LLVMFunctionType(get_llvm_type(g, expr_type), param_types, n_args, false); LLVMValueRef fn_val = LLVMAddFunction(g->module, llvm_name, fn_type); assert(LLVMGetIntrinsicID(fn_val)); g->llvm_fn_table.put(key, fn_val); return fn_val; } static LLVMValueRef ir_render_clz(CodeGen *g, IrExecutableGen *executable, IrInstGenClz *instruction) { ZigType *int_type = instruction->op->value->type; LLVMValueRef fn_val = get_int_builtin_fn(g, int_type, BuiltinFnIdClz); LLVMValueRef operand = ir_llvm_value(g, instruction->op); LLVMValueRef params[] { operand, LLVMConstNull(LLVMInt1Type()), }; LLVMValueRef wrong_size_int = LLVMBuildCall(g->builder, fn_val, params, 2, ""); return gen_widen_or_shorten(g, false, int_type, instruction->base.value->type, wrong_size_int); } static LLVMValueRef ir_render_ctz(CodeGen *g, IrExecutableGen *executable, IrInstGenCtz *instruction) { ZigType *int_type = instruction->op->value->type; LLVMValueRef fn_val = get_int_builtin_fn(g, int_type, BuiltinFnIdCtz); LLVMValueRef operand = ir_llvm_value(g, instruction->op); LLVMValueRef params[] { operand, LLVMConstNull(LLVMInt1Type()), }; LLVMValueRef wrong_size_int = LLVMBuildCall(g->builder, fn_val, params, 2, ""); return gen_widen_or_shorten(g, false, int_type, instruction->base.value->type, wrong_size_int); } static LLVMValueRef ir_render_shuffle_vector(CodeGen *g, IrExecutableGen *executable, IrInstGenShuffleVector *instruction) { uint64_t len_a = instruction->a->value->type->data.vector.len; uint64_t len_mask = instruction->mask->value->type->data.vector.len; // LLVM uses integers larger than the length of the first array to // index into the second array. This was deemed unnecessarily fragile // when changing code, so Zig uses negative numbers to index the // second vector. These start at -1 and go down, and are easiest to use // with the ~ operator. Here we convert between the two formats. IrInstGen *mask = instruction->mask; LLVMValueRef *values = heap::c_allocator.allocate(len_mask); for (uint64_t i = 0; i < len_mask; i++) { if (mask->value->data.x_array.data.s_none.elements[i].special == ConstValSpecialUndef) { values[i] = LLVMGetUndef(LLVMInt32Type()); } else { int32_t v = bigint_as_signed(&mask->value->data.x_array.data.s_none.elements[i].data.x_bigint); uint32_t index_val = (v >= 0) ? (uint32_t)v : (uint32_t)~v + (uint32_t)len_a; values[i] = LLVMConstInt(LLVMInt32Type(), index_val, false); } } LLVMValueRef llvm_mask_value = LLVMConstVector(values, len_mask); heap::c_allocator.deallocate(values, len_mask); return LLVMBuildShuffleVector(g->builder, ir_llvm_value(g, instruction->a), ir_llvm_value(g, instruction->b), llvm_mask_value, ""); } static LLVMValueRef ir_render_splat(CodeGen *g, IrExecutableGen *executable, IrInstGenSplat *instruction) { ZigType *result_type = instruction->base.value->type; ir_assert(result_type->id == ZigTypeIdVector, &instruction->base); uint32_t len = result_type->data.vector.len; LLVMTypeRef op_llvm_type = LLVMVectorType(get_llvm_type(g, instruction->scalar->value->type), 1); LLVMTypeRef mask_llvm_type = LLVMVectorType(LLVMInt32Type(), len); LLVMValueRef undef_vector = LLVMGetUndef(op_llvm_type); LLVMValueRef op_vector = LLVMBuildInsertElement(g->builder, undef_vector, ir_llvm_value(g, instruction->scalar), LLVMConstInt(LLVMInt32Type(), 0, false), ""); return LLVMBuildShuffleVector(g->builder, op_vector, undef_vector, LLVMConstNull(mask_llvm_type), ""); } static LLVMValueRef ir_render_pop_count(CodeGen *g, IrExecutableGen *executable, IrInstGenPopCount *instruction) { ZigType *int_type = instruction->op->value->type; LLVMValueRef fn_val = get_int_builtin_fn(g, int_type, BuiltinFnIdPopCount); LLVMValueRef operand = ir_llvm_value(g, instruction->op); LLVMValueRef wrong_size_int = LLVMBuildCall(g->builder, fn_val, &operand, 1, ""); return gen_widen_or_shorten(g, false, int_type, instruction->base.value->type, wrong_size_int); } static LLVMValueRef ir_render_switch_br(CodeGen *g, IrExecutableGen *executable, IrInstGenSwitchBr *instruction) { ZigType *target_type = instruction->target_value->value->type; LLVMBasicBlockRef else_block = instruction->else_block->llvm_block; LLVMValueRef target_value = ir_llvm_value(g, instruction->target_value); if (target_type->id == ZigTypeIdPointer) { const ZigType *usize = g->builtin_types.entry_usize; target_value = LLVMBuildPtrToInt(g->builder, target_value, usize->llvm_type, ""); } LLVMValueRef switch_instr = LLVMBuildSwitch(g->builder, target_value, else_block, (unsigned)instruction->case_count); for (size_t i = 0; i < instruction->case_count; i += 1) { IrInstGenSwitchBrCase *this_case = &instruction->cases[i]; LLVMValueRef case_value = ir_llvm_value(g, this_case->value); if (target_type->id == ZigTypeIdPointer) { const ZigType *usize = g->builtin_types.entry_usize; case_value = LLVMBuildPtrToInt(g->builder, case_value, usize->llvm_type, ""); } LLVMAddCase(switch_instr, case_value, this_case->block->llvm_block); } return nullptr; } static LLVMValueRef ir_render_phi(CodeGen *g, IrExecutableGen *executable, IrInstGenPhi *instruction) { if (!type_has_bits(g, instruction->base.value->type)) return nullptr; LLVMTypeRef phi_type; if (handle_is_ptr(g, instruction->base.value->type)) { phi_type = LLVMPointerType(get_llvm_type(g,instruction->base.value->type), 0); } else { phi_type = get_llvm_type(g, instruction->base.value->type); } LLVMValueRef phi = LLVMBuildPhi(g->builder, phi_type, ""); LLVMValueRef *incoming_values = heap::c_allocator.allocate(instruction->incoming_count); LLVMBasicBlockRef *incoming_blocks = heap::c_allocator.allocate(instruction->incoming_count); for (size_t i = 0; i < instruction->incoming_count; i += 1) { incoming_values[i] = ir_llvm_value(g, instruction->incoming_values[i]); incoming_blocks[i] = instruction->incoming_blocks[i]->llvm_exit_block; } LLVMAddIncoming(phi, incoming_values, incoming_blocks, (unsigned)instruction->incoming_count); heap::c_allocator.deallocate(incoming_values, instruction->incoming_count); heap::c_allocator.deallocate(incoming_blocks, instruction->incoming_count); return phi; } static LLVMValueRef ir_render_ref(CodeGen *g, IrExecutableGen *executable, IrInstGenRef *instruction) { if (!type_has_bits(g, instruction->base.value->type)) { return nullptr; } if (instruction->operand->id == IrInstGenIdCall) { IrInstGenCall *call = reinterpret_cast(instruction->operand); if (call->result_loc != nullptr) { return ir_llvm_value(g, call->result_loc); } } LLVMValueRef value = ir_llvm_value(g, instruction->operand); if (handle_is_ptr(g, instruction->operand->value->type)) { return value; } else { LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); gen_store_untyped(g, value, result_loc, 0, false); return result_loc; } } static LLVMValueRef ir_render_err_name(CodeGen *g, IrExecutableGen *executable, IrInstGenErrName *instruction) { assert(g->generate_error_name_table); if (g->errors_by_index.length == 1) { LLVMBuildUnreachable(g->builder); return nullptr; } LLVMValueRef err_val = ir_llvm_value(g, instruction->value); if (ir_want_runtime_safety(g, &instruction->base)) { LLVMValueRef zero = LLVMConstNull(LLVMTypeOf(err_val)); LLVMValueRef end_val = LLVMConstInt(LLVMTypeOf(err_val), g->errors_by_index.length, false); add_bounds_check(g, err_val, LLVMIntNE, zero, LLVMIntULT, end_val); } LLVMValueRef indices[] = { LLVMConstNull(g->builtin_types.entry_usize->llvm_type), err_val, }; return LLVMBuildInBoundsGEP(g->builder, g->err_name_table, indices, 2, ""); } static LLVMValueRef get_enum_tag_name_function(CodeGen *g, ZigType *enum_type) { assert(enum_type->id == ZigTypeIdEnum); if (enum_type->data.enumeration.name_function) return enum_type->data.enumeration.name_function; ZigType *u8_ptr_type = get_pointer_to_type_extra(g, g->builtin_types.entry_u8, false, false, PtrLenUnknown, get_abi_alignment(g, g->builtin_types.entry_u8), 0, 0, false); ZigType *u8_slice_type = get_slice_type(g, u8_ptr_type); ZigType *tag_int_type = enum_type->data.enumeration.tag_int_type; LLVMTypeRef tag_int_llvm_type = get_llvm_type(g, tag_int_type); LLVMTypeRef fn_type_ref = LLVMFunctionType(LLVMPointerType(get_llvm_type(g, u8_slice_type), 0), &tag_int_llvm_type, 1, false); const char *fn_name = get_mangled_name(g, buf_ptr(buf_sprintf("__zig_tag_name_%s", buf_ptr(&enum_type->name)))); LLVMValueRef fn_val = LLVMAddFunction(g->module, fn_name, fn_type_ref); LLVMSetLinkage(fn_val, LLVMInternalLinkage); ZigLLVMFunctionSetCallingConv(fn_val, get_llvm_cc(g, CallingConventionUnspecified)); addLLVMFnAttr(fn_val, "nounwind"); add_uwtable_attr(g, fn_val); if (codegen_have_frame_pointer(g)) { ZigLLVMAddFunctionAttr(fn_val, "frame-pointer", "all"); } LLVMBasicBlockRef prev_block = LLVMGetInsertBlock(g->builder); LLVMValueRef prev_debug_location = LLVMGetCurrentDebugLocation(g->builder); ZigFn *prev_cur_fn = g->cur_fn; LLVMValueRef prev_cur_fn_val = g->cur_fn_val; LLVMBasicBlockRef entry_block = LLVMAppendBasicBlock(fn_val, "Entry"); LLVMPositionBuilderAtEnd(g->builder, entry_block); ZigLLVMClearCurrentDebugLocation(g->builder); g->cur_fn = nullptr; g->cur_fn_val = fn_val; size_t field_count = enum_type->data.enumeration.src_field_count; LLVMBasicBlockRef bad_value_block = LLVMAppendBasicBlock(g->cur_fn_val, "BadValue"); LLVMValueRef tag_int_value = LLVMGetParam(fn_val, 0); LLVMValueRef switch_instr = LLVMBuildSwitch(g->builder, tag_int_value, bad_value_block, field_count); ZigType *usize = g->builtin_types.entry_usize; LLVMValueRef array_ptr_indices[] = { LLVMConstNull(usize->llvm_type), LLVMConstNull(usize->llvm_type), }; HashMap occupied_tag_values = {}; occupied_tag_values.init(field_count); for (size_t field_i = 0; field_i < field_count; field_i += 1) { TypeEnumField *type_enum_field = &enum_type->data.enumeration.fields[field_i]; Buf *name = type_enum_field->name; auto entry = occupied_tag_values.put_unique(type_enum_field->value, name); if (entry != nullptr) { continue; } LLVMValueRef str_init = LLVMConstString(buf_ptr(name), (unsigned)buf_len(name), true); LLVMValueRef str_global = LLVMAddGlobal(g->module, LLVMTypeOf(str_init), ""); LLVMSetInitializer(str_global, str_init); LLVMSetLinkage(str_global, LLVMPrivateLinkage); LLVMSetGlobalConstant(str_global, true); LLVMSetUnnamedAddr(str_global, true); LLVMSetAlignment(str_global, LLVMABIAlignmentOfType(g->target_data_ref, LLVMTypeOf(str_init))); LLVMValueRef fields[] = { LLVMConstGEP(str_global, array_ptr_indices, 2), LLVMConstInt(g->builtin_types.entry_usize->llvm_type, buf_len(name), false), }; LLVMValueRef slice_init_value = LLVMConstNamedStruct(get_llvm_type(g, u8_slice_type), fields, 2); LLVMValueRef slice_global = LLVMAddGlobal(g->module, LLVMTypeOf(slice_init_value), ""); LLVMSetInitializer(slice_global, slice_init_value); LLVMSetLinkage(slice_global, LLVMPrivateLinkage); LLVMSetGlobalConstant(slice_global, true); LLVMSetUnnamedAddr(slice_global, true); LLVMSetAlignment(slice_global, LLVMABIAlignmentOfType(g->target_data_ref, LLVMTypeOf(slice_init_value))); LLVMBasicBlockRef return_block = LLVMAppendBasicBlock(g->cur_fn_val, "Name"); LLVMValueRef this_tag_int_value = bigint_to_llvm_const(get_llvm_type(g, tag_int_type), &enum_type->data.enumeration.fields[field_i].value); LLVMAddCase(switch_instr, this_tag_int_value, return_block); LLVMPositionBuilderAtEnd(g->builder, return_block); LLVMBuildRet(g->builder, slice_global); } occupied_tag_values.deinit(); LLVMPositionBuilderAtEnd(g->builder, bad_value_block); if (g->build_mode == BuildModeDebug || g->build_mode == BuildModeSafeRelease) { gen_safety_crash(g, PanicMsgIdBadEnumValue); } else { LLVMBuildUnreachable(g->builder); } g->cur_fn = prev_cur_fn; g->cur_fn_val = prev_cur_fn_val; LLVMPositionBuilderAtEnd(g->builder, prev_block); if (!g->strip_debug_symbols) { LLVMSetCurrentDebugLocation(g->builder, prev_debug_location); } enum_type->data.enumeration.name_function = fn_val; return fn_val; } static LLVMValueRef ir_render_enum_tag_name(CodeGen *g, IrExecutableGen *executable, IrInstGenTagName *instruction) { ZigType *enum_type = instruction->target->value->type; assert(enum_type->id == ZigTypeIdEnum); LLVMValueRef enum_name_function = get_enum_tag_name_function(g, enum_type); LLVMValueRef enum_tag_value = ir_llvm_value(g, instruction->target); return ZigLLVMBuildCall(g->builder, enum_name_function, &enum_tag_value, 1, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAuto, ""); } static LLVMValueRef ir_render_field_parent_ptr(CodeGen *g, IrExecutableGen *executable, IrInstGenFieldParentPtr *instruction) { ZigType *container_ptr_type = instruction->base.value->type; assert(container_ptr_type->id == ZigTypeIdPointer); ZigType *container_type = container_ptr_type->data.pointer.child_type; size_t byte_offset = LLVMOffsetOfElement(g->target_data_ref, get_llvm_type(g, container_type), instruction->field->gen_index); LLVMValueRef field_ptr_val = ir_llvm_value(g, instruction->field_ptr); if (byte_offset == 0) { return LLVMBuildBitCast(g->builder, field_ptr_val, get_llvm_type(g, container_ptr_type), ""); } else { ZigType *usize = g->builtin_types.entry_usize; LLVMValueRef field_ptr_int = LLVMBuildPtrToInt(g->builder, field_ptr_val, usize->llvm_type, ""); LLVMValueRef base_ptr_int = LLVMBuildNUWSub(g->builder, field_ptr_int, LLVMConstInt(usize->llvm_type, byte_offset, false), ""); return LLVMBuildIntToPtr(g->builder, base_ptr_int, get_llvm_type(g, container_ptr_type), ""); } } static LLVMValueRef ir_render_align_cast(CodeGen *g, IrExecutableGen *executable, IrInstGenAlignCast *instruction) { LLVMValueRef target_val = ir_llvm_value(g, instruction->target); assert(target_val); bool want_runtime_safety = ir_want_runtime_safety(g, &instruction->base); if (!want_runtime_safety) { return target_val; } ZigType *target_type = instruction->base.value->type; uint32_t align_bytes; LLVMValueRef ptr_val; if (target_type->id == ZigTypeIdPointer) { align_bytes = get_ptr_align(g, target_type); ptr_val = target_val; } else if (target_type->id == ZigTypeIdFn) { align_bytes = target_type->data.fn.fn_type_id.alignment; ptr_val = target_val; } else if (target_type->id == ZigTypeIdOptional && target_type->data.maybe.child_type->id == ZigTypeIdPointer) { align_bytes = get_ptr_align(g, target_type->data.maybe.child_type); ptr_val = target_val; } else if (target_type->id == ZigTypeIdOptional && target_type->data.maybe.child_type->id == ZigTypeIdFn) { align_bytes = target_type->data.maybe.child_type->data.fn.fn_type_id.alignment; ptr_val = target_val; } else if (target_type->id == ZigTypeIdStruct && target_type->data.structure.special == StructSpecialSlice) { ZigType *slice_ptr_type = target_type->data.structure.fields[slice_ptr_index]->type_entry; align_bytes = get_ptr_align(g, slice_ptr_type); size_t ptr_index = target_type->data.structure.fields[slice_ptr_index]->gen_index; LLVMValueRef ptr_val_ptr = LLVMBuildStructGEP(g->builder, target_val, (unsigned)ptr_index, ""); ptr_val = gen_load_untyped(g, ptr_val_ptr, 0, false, ""); } else { zig_unreachable(); } assert(align_bytes != 1); ZigType *usize = g->builtin_types.entry_usize; LLVMValueRef ptr_as_int_val = LLVMBuildPtrToInt(g->builder, ptr_val, usize->llvm_type, ""); LLVMValueRef alignment_minus_1 = LLVMConstInt(usize->llvm_type, align_bytes - 1, false); LLVMValueRef anded_val = LLVMBuildAnd(g->builder, ptr_as_int_val, alignment_minus_1, ""); LLVMValueRef ok_bit = LLVMBuildICmp(g->builder, LLVMIntEQ, anded_val, LLVMConstNull(usize->llvm_type), ""); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "AlignCastOk"); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "AlignCastFail"); LLVMBuildCondBr(g->builder, ok_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_safety_crash(g, PanicMsgIdIncorrectAlignment); LLVMPositionBuilderAtEnd(g->builder, ok_block); return target_val; } static LLVMValueRef ir_render_error_return_trace(CodeGen *g, IrExecutableGen *executable, IrInstGenErrorReturnTrace *instruction) { bool is_llvm_alloca; LLVMValueRef cur_err_ret_trace_val = get_cur_err_ret_trace_val(g, instruction->base.base.scope, &is_llvm_alloca); if (cur_err_ret_trace_val == nullptr) { return LLVMConstNull(get_llvm_type(g, ptr_to_stack_trace_type(g))); } return cur_err_ret_trace_val; } static LLVMAtomicOrdering to_LLVMAtomicOrdering(AtomicOrder atomic_order) { switch (atomic_order) { case AtomicOrderUnordered: return LLVMAtomicOrderingUnordered; case AtomicOrderMonotonic: return LLVMAtomicOrderingMonotonic; case AtomicOrderAcquire: return LLVMAtomicOrderingAcquire; case AtomicOrderRelease: return LLVMAtomicOrderingRelease; case AtomicOrderAcqRel: return LLVMAtomicOrderingAcquireRelease; case AtomicOrderSeqCst: return LLVMAtomicOrderingSequentiallyConsistent; } zig_unreachable(); } static enum ZigLLVM_AtomicRMWBinOp to_ZigLLVMAtomicRMWBinOp(AtomicRmwOp op, bool is_signed, bool is_float) { switch (op) { case AtomicRmwOp_xchg: return ZigLLVMAtomicRMWBinOpXchg; case AtomicRmwOp_add: return is_float ? ZigLLVMAtomicRMWBinOpFAdd : ZigLLVMAtomicRMWBinOpAdd; case AtomicRmwOp_sub: return is_float ? ZigLLVMAtomicRMWBinOpFSub : ZigLLVMAtomicRMWBinOpSub; case AtomicRmwOp_and: return ZigLLVMAtomicRMWBinOpAnd; case AtomicRmwOp_nand: return ZigLLVMAtomicRMWBinOpNand; case AtomicRmwOp_or: return ZigLLVMAtomicRMWBinOpOr; case AtomicRmwOp_xor: return ZigLLVMAtomicRMWBinOpXor; case AtomicRmwOp_max: return is_signed ? ZigLLVMAtomicRMWBinOpMax : ZigLLVMAtomicRMWBinOpUMax; case AtomicRmwOp_min: return is_signed ? ZigLLVMAtomicRMWBinOpMin : ZigLLVMAtomicRMWBinOpUMin; } zig_unreachable(); } static LLVMTypeRef get_atomic_abi_type(CodeGen *g, IrInstGen *instruction) { // If the operand type of an atomic operation is not a power of two sized // we need to widen it before using it and then truncate the result. ir_assert(instruction->value->type->id == ZigTypeIdPointer, instruction); ZigType *operand_type = instruction->value->type->data.pointer.child_type; if (operand_type->id == ZigTypeIdInt || operand_type->id == ZigTypeIdEnum) { if (operand_type->id == ZigTypeIdEnum) { operand_type = operand_type->data.enumeration.tag_int_type; } auto bit_count = operand_type->data.integral.bit_count; bool is_signed = operand_type->data.integral.is_signed; ir_assert(bit_count != 0, instruction); if (bit_count == 1 || !is_power_of_2(bit_count)) { return get_llvm_type(g, get_int_type(g, is_signed, operand_type->abi_size * 8)); } else { return nullptr; } } else if (operand_type->id == ZigTypeIdFloat) { return nullptr; } else if (operand_type->id == ZigTypeIdBool) { return g->builtin_types.entry_u8->llvm_type; } else { ir_assert(get_codegen_ptr_type_bail(g, operand_type) != nullptr, instruction); return nullptr; } } static LLVMValueRef ir_render_cmpxchg(CodeGen *g, IrExecutableGen *executable, IrInstGenCmpxchg *instruction) { LLVMValueRef ptr_val = ir_llvm_value(g, instruction->ptr); LLVMValueRef cmp_val = ir_llvm_value(g, instruction->cmp_value); LLVMValueRef new_val = ir_llvm_value(g, instruction->new_value); ZigType *operand_type = instruction->new_value->value->type; LLVMTypeRef actual_abi_type = get_atomic_abi_type(g, instruction->ptr); if (actual_abi_type != nullptr) { // operand needs widening and truncating ptr_val = LLVMBuildBitCast(g->builder, ptr_val, LLVMPointerType(actual_abi_type, 0), ""); if (operand_type->data.integral.is_signed) { cmp_val = LLVMBuildSExt(g->builder, cmp_val, actual_abi_type, ""); new_val = LLVMBuildSExt(g->builder, new_val, actual_abi_type, ""); } else { cmp_val = LLVMBuildZExt(g->builder, cmp_val, actual_abi_type, ""); new_val = LLVMBuildZExt(g->builder, new_val, actual_abi_type, ""); } } LLVMAtomicOrdering success_order = to_LLVMAtomicOrdering(instruction->success_order); LLVMAtomicOrdering failure_order = to_LLVMAtomicOrdering(instruction->failure_order); LLVMValueRef result_val = ZigLLVMBuildCmpXchg(g->builder, ptr_val, cmp_val, new_val, success_order, failure_order, instruction->is_weak); ZigType *optional_type = instruction->base.value->type; assert(optional_type->id == ZigTypeIdOptional); ZigType *child_type = optional_type->data.maybe.child_type; if (!handle_is_ptr(g, optional_type)) { LLVMValueRef payload_val = LLVMBuildExtractValue(g->builder, result_val, 0, ""); if (actual_abi_type != nullptr) { payload_val = LLVMBuildTrunc(g->builder, payload_val, get_llvm_type(g, operand_type), ""); } LLVMValueRef success_bit = LLVMBuildExtractValue(g->builder, result_val, 1, ""); return LLVMBuildSelect(g->builder, success_bit, LLVMConstNull(get_llvm_type(g, child_type)), payload_val, ""); } // When the cmpxchg is discarded, the result location will have no bits. if (!type_has_bits(g, instruction->result_loc->value->type)) { return nullptr; } LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); ir_assert(result_loc != nullptr, &instruction->base); ir_assert(type_has_bits(g, child_type), &instruction->base); LLVMValueRef payload_val = LLVMBuildExtractValue(g->builder, result_val, 0, ""); if (actual_abi_type != nullptr) { payload_val = LLVMBuildTrunc(g->builder, payload_val, get_llvm_type(g, operand_type), ""); } LLVMValueRef val_ptr = LLVMBuildStructGEP(g->builder, result_loc, maybe_child_index, ""); gen_assign_raw(g, val_ptr, get_pointer_to_type(g, child_type, false), payload_val); LLVMValueRef success_bit = LLVMBuildExtractValue(g->builder, result_val, 1, ""); LLVMValueRef nonnull_bit = LLVMBuildNot(g->builder, success_bit, ""); LLVMValueRef maybe_ptr = LLVMBuildStructGEP(g->builder, result_loc, maybe_null_index, ""); gen_store_untyped(g, nonnull_bit, maybe_ptr, 0, false); return result_loc; } static LLVMValueRef ir_render_reduce(CodeGen *g, IrExecutableGen *executable, IrInstGenReduce *instruction) { LLVMValueRef value = ir_llvm_value(g, instruction->value); ZigType *value_type = instruction->value->value->type; assert(value_type->id == ZigTypeIdVector); ZigType *scalar_type = value_type->data.vector.elem_type; ZigLLVMSetFastMath(g->builder, ir_want_fast_math(g, &instruction->base)); LLVMValueRef result_val; switch (instruction->op) { case ReduceOp_and: assert(scalar_type->id == ZigTypeIdInt || scalar_type->id == ZigTypeIdBool); result_val = ZigLLVMBuildAndReduce(g->builder, value); break; case ReduceOp_or: assert(scalar_type->id == ZigTypeIdInt || scalar_type->id == ZigTypeIdBool); result_val = ZigLLVMBuildOrReduce(g->builder, value); break; case ReduceOp_xor: assert(scalar_type->id == ZigTypeIdInt || scalar_type->id == ZigTypeIdBool); result_val = ZigLLVMBuildXorReduce(g->builder, value); break; case ReduceOp_min: { if (scalar_type->id == ZigTypeIdInt) { const bool is_signed = scalar_type->data.integral.is_signed; result_val = ZigLLVMBuildIntMinReduce(g->builder, value, is_signed); } else if (scalar_type->id == ZigTypeIdFloat) { result_val = ZigLLVMBuildFPMinReduce(g->builder, value); } else zig_unreachable(); } break; case ReduceOp_max: { if (scalar_type->id == ZigTypeIdInt) { const bool is_signed = scalar_type->data.integral.is_signed; result_val = ZigLLVMBuildIntMaxReduce(g->builder, value, is_signed); } else if (scalar_type->id == ZigTypeIdFloat) { result_val = ZigLLVMBuildFPMaxReduce(g->builder, value); } else zig_unreachable(); } break; case ReduceOp_add: { if (scalar_type->id == ZigTypeIdInt) { result_val = ZigLLVMBuildAddReduce(g->builder, value); } else if (scalar_type->id == ZigTypeIdFloat) { LLVMValueRef neutral_value = LLVMConstReal( get_llvm_type(g, scalar_type), -0.0); result_val = ZigLLVMBuildFPAddReduce(g->builder, neutral_value, value); } else zig_unreachable(); } break; case ReduceOp_mul: { if (scalar_type->id == ZigTypeIdInt) { result_val = ZigLLVMBuildMulReduce(g->builder, value); } else if (scalar_type->id == ZigTypeIdFloat) { LLVMValueRef neutral_value = LLVMConstReal( get_llvm_type(g, scalar_type), 1.0); result_val = ZigLLVMBuildFPMulReduce(g->builder, neutral_value, value); } else zig_unreachable(); } break; default: zig_unreachable(); } return result_val; } static LLVMValueRef ir_render_fence(CodeGen *g, IrExecutableGen *executable, IrInstGenFence *instruction) { LLVMAtomicOrdering atomic_order = to_LLVMAtomicOrdering(instruction->order); LLVMBuildFence(g->builder, atomic_order, false, ""); return nullptr; } static LLVMValueRef ir_render_truncate(CodeGen *g, IrExecutableGen *executable, IrInstGenTruncate *instruction) { LLVMValueRef target_val = ir_llvm_value(g, instruction->target); ZigType *dest_type = instruction->base.value->type; ZigType *src_type = instruction->target->value->type; if (dest_type == src_type) { // no-op return target_val; } if (src_type->data.integral.bit_count == dest_type->data.integral.bit_count) { return LLVMBuildBitCast(g->builder, target_val, get_llvm_type(g, dest_type), ""); } else { LLVMValueRef target_val = ir_llvm_value(g, instruction->target); return LLVMBuildTrunc(g->builder, target_val, get_llvm_type(g, dest_type), ""); } } static LLVMValueRef ir_render_memset(CodeGen *g, IrExecutableGen *executable, IrInstGenMemset *instruction) { LLVMValueRef dest_ptr = ir_llvm_value(g, instruction->dest_ptr); LLVMValueRef len_val = ir_llvm_value(g, instruction->count); LLVMTypeRef ptr_u8 = LLVMPointerType(LLVMInt8Type(), 0); LLVMValueRef dest_ptr_casted = LLVMBuildBitCast(g->builder, dest_ptr, ptr_u8, ""); ZigType *ptr_type = instruction->dest_ptr->value->type; assert(ptr_type->id == ZigTypeIdPointer); bool val_is_undef = value_is_all_undef(g, instruction->byte->value); LLVMValueRef fill_char; if (val_is_undef) { if (ir_want_runtime_safety_scope(g, instruction->base.base.scope)) { fill_char = LLVMConstInt(LLVMInt8Type(), 0xaa, false); } else { return nullptr; } } else { fill_char = ir_llvm_value(g, instruction->byte); } ZigLLVMBuildMemSet(g->builder, dest_ptr_casted, fill_char, len_val, get_ptr_align(g, ptr_type), ptr_type->data.pointer.is_volatile); if (val_is_undef && g->valgrind_enabled) { gen_valgrind_undef(g, dest_ptr_casted, len_val); } return nullptr; } static LLVMValueRef ir_render_memcpy(CodeGen *g, IrExecutableGen *executable, IrInstGenMemcpy *instruction) { LLVMValueRef dest_ptr = ir_llvm_value(g, instruction->dest_ptr); LLVMValueRef src_ptr = ir_llvm_value(g, instruction->src_ptr); LLVMValueRef len_val = ir_llvm_value(g, instruction->count); LLVMTypeRef ptr_u8 = LLVMPointerType(LLVMInt8Type(), 0); LLVMValueRef dest_ptr_casted = LLVMBuildBitCast(g->builder, dest_ptr, ptr_u8, ""); LLVMValueRef src_ptr_casted = LLVMBuildBitCast(g->builder, src_ptr, ptr_u8, ""); ZigType *dest_ptr_type = instruction->dest_ptr->value->type; ZigType *src_ptr_type = instruction->src_ptr->value->type; assert(dest_ptr_type->id == ZigTypeIdPointer); assert(src_ptr_type->id == ZigTypeIdPointer); bool is_volatile = (dest_ptr_type->data.pointer.is_volatile || src_ptr_type->data.pointer.is_volatile); ZigLLVMBuildMemCpy(g->builder, dest_ptr_casted, get_ptr_align(g, dest_ptr_type), src_ptr_casted, get_ptr_align(g, src_ptr_type), len_val, is_volatile); return nullptr; } static LLVMValueRef ir_render_wasm_memory_size(CodeGen *g, IrExecutableGen *executable, IrInstGenWasmMemorySize *instruction) { // TODO adjust for wasm64 LLVMValueRef param = ir_llvm_value(g, instruction->index); LLVMValueRef val = LLVMBuildCall(g->builder, gen_wasm_memory_size(g), ¶m, 1, ""); return val; } static LLVMValueRef ir_render_wasm_memory_grow(CodeGen *g, IrExecutableGen *executable, IrInstGenWasmMemoryGrow *instruction) { // TODO adjust for wasm64 LLVMValueRef params[] = { ir_llvm_value(g, instruction->index), ir_llvm_value(g, instruction->delta), }; LLVMValueRef val = LLVMBuildCall(g->builder, gen_wasm_memory_grow(g), params, 2, ""); return val; } static LLVMValueRef ir_render_slice(CodeGen *g, IrExecutableGen *executable, IrInstGenSlice *instruction) { Error err; LLVMValueRef array_ptr_ptr = ir_llvm_value(g, instruction->ptr); ZigType *array_ptr_type = instruction->ptr->value->type; assert(array_ptr_type->id == ZigTypeIdPointer); ZigType *array_type = array_ptr_type->data.pointer.child_type; LLVMValueRef array_ptr = get_handle_value(g, array_ptr_ptr, array_type, array_ptr_type); bool want_runtime_safety = instruction->safety_check_on && ir_want_runtime_safety(g, &instruction->base); // The result is either a slice or a pointer to an array ZigType *result_type = instruction->base.value->type; // This is not whether the result type has a sentinel, but whether there should be a sentinel check, // e.g. if they used [a..b :s] syntax. ZigValue *sentinel = instruction->sentinel; LLVMValueRef slice_start_ptr = nullptr; LLVMValueRef len_value = nullptr; if (array_type->id == ZigTypeIdArray || (array_type->id == ZigTypeIdPointer && array_type->data.pointer.ptr_len == PtrLenSingle)) { if (array_type->id == ZigTypeIdPointer) { array_type = array_type->data.pointer.child_type; } LLVMValueRef start_val = ir_llvm_value(g, instruction->start); LLVMValueRef end_val; if (instruction->end) { end_val = ir_llvm_value(g, instruction->end); } else { end_val = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, array_type->data.array.len, false); } if (want_runtime_safety) { // Safety check: start <= end if (instruction->start->value->special == ConstValSpecialRuntime || instruction->end) { add_bounds_check(g, start_val, LLVMIntEQ, nullptr, LLVMIntULE, end_val); } // Safety check: the last element of the slice (the sentinel if // requested) must be inside the array // XXX: Overflow is not checked here... const size_t full_len = array_type->data.array.len + (array_type->data.array.sentinel != nullptr); LLVMValueRef array_end = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, full_len, false); LLVMValueRef check_end_val = end_val; if (sentinel != nullptr) { LLVMValueRef usize_one = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, 1, false); check_end_val = LLVMBuildNUWAdd(g->builder, end_val, usize_one, ""); } add_bounds_check(g, check_end_val, LLVMIntEQ, nullptr, LLVMIntULE, array_end); } bool value_has_bits; if ((err = type_has_bits2(g, array_type, &value_has_bits))) codegen_report_errors_and_exit(g); if (value_has_bits) { if (want_runtime_safety && sentinel != nullptr) { LLVMValueRef indices[] = { LLVMConstNull(g->builtin_types.entry_usize->llvm_type), end_val, }; LLVMValueRef sentinel_elem_ptr = LLVMBuildInBoundsGEP(g->builder, array_ptr, indices, 2, ""); add_sentinel_check(g, sentinel_elem_ptr, sentinel); } LLVMValueRef indices[] = { LLVMConstNull(g->builtin_types.entry_usize->llvm_type), start_val, }; slice_start_ptr = LLVMBuildInBoundsGEP(g->builder, array_ptr, indices, 2, ""); } len_value = LLVMBuildNUWSub(g->builder, end_val, start_val, ""); } else if (array_type->id == ZigTypeIdPointer) { assert(array_type->data.pointer.ptr_len != PtrLenSingle); LLVMValueRef start_val = ir_llvm_value(g, instruction->start); LLVMValueRef end_val = ir_llvm_value(g, instruction->end); if (want_runtime_safety) { // Safety check: start <= end add_bounds_check(g, start_val, LLVMIntEQ, nullptr, LLVMIntULE, end_val); } bool value_has_bits; if ((err = type_has_bits2(g, array_type, &value_has_bits))) codegen_report_errors_and_exit(g); if (value_has_bits) { if (want_runtime_safety && sentinel != nullptr) { LLVMValueRef sentinel_elem_ptr = LLVMBuildInBoundsGEP(g->builder, array_ptr, &end_val, 1, ""); add_sentinel_check(g, sentinel_elem_ptr, sentinel); } slice_start_ptr = LLVMBuildInBoundsGEP(g->builder, array_ptr, &start_val, 1, ""); } len_value = LLVMBuildNUWSub(g->builder, end_val, start_val, ""); } else if (array_type->id == ZigTypeIdStruct) { assert(array_type->data.structure.special == StructSpecialSlice); assert(LLVMGetTypeKind(LLVMTypeOf(array_ptr)) == LLVMPointerTypeKind); assert(LLVMGetTypeKind(LLVMGetElementType(LLVMTypeOf(array_ptr))) == LLVMStructTypeKind); const size_t gen_len_index = array_type->data.structure.fields[slice_len_index]->gen_index; assert(gen_len_index != SIZE_MAX); LLVMValueRef prev_end = nullptr; if (!instruction->end || want_runtime_safety) { LLVMValueRef src_len_ptr = LLVMBuildStructGEP(g->builder, array_ptr, gen_len_index, ""); prev_end = gen_load_untyped(g, src_len_ptr, 0, false, ""); } LLVMValueRef start_val = ir_llvm_value(g, instruction->start); LLVMValueRef end_val; if (instruction->end) { end_val = ir_llvm_value(g, instruction->end); } else { end_val = prev_end; } ZigType *ptr_field_type = array_type->data.structure.fields[slice_ptr_index]->type_entry; if (want_runtime_safety) { assert(prev_end); // Safety check: start <= end add_bounds_check(g, start_val, LLVMIntEQ, nullptr, LLVMIntULE, end_val); // Safety check: the sentinel counts as one more element // XXX: Overflow is not checked here... LLVMValueRef check_prev_end = prev_end; if (ptr_field_type->data.pointer.sentinel != nullptr) { LLVMValueRef usize_one = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, 1, false); check_prev_end = LLVMBuildNUWAdd(g->builder, prev_end, usize_one, ""); } LLVMValueRef check_end_val = end_val; if (sentinel != nullptr) { LLVMValueRef usize_one = LLVMConstInt(g->builtin_types.entry_usize->llvm_type, 1, false); check_end_val = LLVMBuildNUWAdd(g->builder, end_val, usize_one, ""); } add_bounds_check(g, check_end_val, LLVMIntEQ, nullptr, LLVMIntULE, check_prev_end); } bool ptr_has_bits; if ((err = type_has_bits2(g, ptr_field_type, &ptr_has_bits))) codegen_report_errors_and_exit(g); if (ptr_has_bits) { const size_t gen_ptr_index = array_type->data.structure.fields[slice_ptr_index]->gen_index; assert(gen_ptr_index != SIZE_MAX); LLVMValueRef src_ptr_ptr = LLVMBuildStructGEP(g->builder, array_ptr, gen_ptr_index, ""); LLVMValueRef src_ptr = gen_load_untyped(g, src_ptr_ptr, 0, false, ""); if (sentinel != nullptr) { LLVMValueRef sentinel_elem_ptr = LLVMBuildInBoundsGEP(g->builder, src_ptr, &end_val, 1, ""); add_sentinel_check(g, sentinel_elem_ptr, sentinel); } slice_start_ptr = LLVMBuildInBoundsGEP(g->builder, src_ptr, &start_val, 1, ""); } len_value = LLVMBuildNUWSub(g->builder, end_val, start_val, ""); } else { zig_unreachable(); } bool result_has_bits; if ((err = type_has_bits2(g, result_type, &result_has_bits))) codegen_report_errors_and_exit(g); // Nothing to do, we're only interested in the bound checks emitted above if (!result_has_bits) return nullptr; // The starting pointer for the slice may be null in case of zero-sized // arrays, the length value is always defined. assert(len_value != nullptr); // The slice decays into a pointer to an array, the size is tracked in the // type itself if (result_type->id == ZigTypeIdPointer) { ir_assert(instruction->result_loc == nullptr, &instruction->base); LLVMTypeRef result_ptr_type = get_llvm_type(g, result_type); if (slice_start_ptr != nullptr) { return LLVMBuildBitCast(g->builder, slice_start_ptr, result_ptr_type, ""); } return LLVMGetUndef(result_ptr_type); } ir_assert(instruction->result_loc != nullptr, &instruction->base); // Create a new slice LLVMValueRef tmp_struct_ptr = ir_llvm_value(g, instruction->result_loc); ZigType *slice_ptr_type = result_type->data.structure.fields[slice_ptr_index]->type_entry; // The slice may not have a pointer at all if it points to a zero-sized type const size_t gen_ptr_index = result_type->data.structure.fields[slice_ptr_index]->gen_index; if (gen_ptr_index != SIZE_MAX) { LLVMValueRef ptr_field_ptr = LLVMBuildStructGEP(g->builder, tmp_struct_ptr, gen_ptr_index, ""); if (slice_start_ptr != nullptr) { gen_store_untyped(g, slice_start_ptr, ptr_field_ptr, 0, false); } else if (want_runtime_safety) { gen_undef_init(g, slice_ptr_type, slice_ptr_type, ptr_field_ptr); } else { gen_store_untyped(g, LLVMGetUndef(get_llvm_type(g, slice_ptr_type)), ptr_field_ptr, 0, false); } } const size_t gen_len_index = result_type->data.structure.fields[slice_len_index]->gen_index; assert(gen_len_index != SIZE_MAX); LLVMValueRef len_field_ptr = LLVMBuildStructGEP(g->builder, tmp_struct_ptr, gen_len_index, ""); gen_store_untyped(g, len_value, len_field_ptr, 0, false); return tmp_struct_ptr; } static LLVMValueRef get_trap_fn_val(CodeGen *g) { if (g->trap_fn_val) return g->trap_fn_val; LLVMTypeRef fn_type = LLVMFunctionType(LLVMVoidType(), nullptr, 0, false); g->trap_fn_val = LLVMAddFunction(g->module, "llvm.debugtrap", fn_type); assert(LLVMGetIntrinsicID(g->trap_fn_val)); return g->trap_fn_val; } static LLVMValueRef ir_render_breakpoint(CodeGen *g, IrExecutableGen *executable, IrInstGenBreakpoint *instruction) { LLVMBuildCall(g->builder, get_trap_fn_val(g), nullptr, 0, ""); return nullptr; } static LLVMValueRef ir_render_return_address(CodeGen *g, IrExecutableGen *executable, IrInstGenReturnAddress *instruction) { if (target_is_wasm(g->zig_target) && g->zig_target->os != OsEmscripten) { // I got this error from LLVM 10: // "Non-Emscripten WebAssembly hasn't implemented __builtin_return_address" return LLVMConstNull(get_llvm_type(g, instruction->base.value->type)); } LLVMValueRef zero = LLVMConstNull(g->builtin_types.entry_i32->llvm_type); LLVMValueRef ptr_val = LLVMBuildCall(g->builder, get_return_address_fn_val(g), &zero, 1, ""); return LLVMBuildPtrToInt(g->builder, ptr_val, g->builtin_types.entry_usize->llvm_type, ""); } static LLVMValueRef get_frame_address_fn_val(CodeGen *g) { if (g->frame_address_fn_val) return g->frame_address_fn_val; ZigType *return_type = get_pointer_to_type(g, g->builtin_types.entry_u8, true); LLVMTypeRef fn_type = LLVMFunctionType(get_llvm_type(g, return_type), &g->builtin_types.entry_i32->llvm_type, 1, false); g->frame_address_fn_val = LLVMAddFunction(g->module, "llvm.frameaddress.p0i8", fn_type); assert(LLVMGetIntrinsicID(g->frame_address_fn_val)); return g->frame_address_fn_val; } static LLVMValueRef ir_render_frame_address(CodeGen *g, IrExecutableGen *executable, IrInstGenFrameAddress *instruction) { LLVMValueRef zero = LLVMConstNull(g->builtin_types.entry_i32->llvm_type); LLVMValueRef ptr_val = LLVMBuildCall(g->builder, get_frame_address_fn_val(g), &zero, 1, ""); return LLVMBuildPtrToInt(g->builder, ptr_val, g->builtin_types.entry_usize->llvm_type, ""); } static LLVMValueRef ir_render_handle(CodeGen *g, IrExecutableGen *executable, IrInstGenFrameHandle *instruction) { return g->cur_frame_ptr; } static LLVMValueRef render_shl_with_overflow(CodeGen *g, IrInstGenOverflowOp *instruction) { ZigType *int_type = instruction->result_ptr_type; assert(int_type->id == ZigTypeIdInt); LLVMValueRef op1 = ir_llvm_value(g, instruction->op1); LLVMValueRef op2 = ir_llvm_value(g, instruction->op2); LLVMValueRef ptr_result = ir_llvm_value(g, instruction->result_ptr); LLVMValueRef op2_casted = gen_widen_or_shorten(g, false, instruction->op2->value->type, instruction->op1->value->type, op2); LLVMValueRef result = LLVMBuildShl(g->builder, op1, op2_casted, ""); LLVMValueRef orig_val; if (int_type->data.integral.is_signed) { orig_val = LLVMBuildAShr(g->builder, result, op2_casted, ""); } else { orig_val = LLVMBuildLShr(g->builder, result, op2_casted, ""); } LLVMValueRef overflow_bit = LLVMBuildICmp(g->builder, LLVMIntNE, op1, orig_val, ""); gen_store(g, result, ptr_result, instruction->result_ptr->value->type); return overflow_bit; } static LLVMValueRef ir_render_overflow_op(CodeGen *g, IrExecutableGen *executable, IrInstGenOverflowOp *instruction) { AddSubMul add_sub_mul; switch (instruction->op) { case IrOverflowOpAdd: add_sub_mul = AddSubMulAdd; break; case IrOverflowOpSub: add_sub_mul = AddSubMulSub; break; case IrOverflowOpMul: add_sub_mul = AddSubMulMul; break; case IrOverflowOpShl: return render_shl_with_overflow(g, instruction); } ZigType *int_type = instruction->result_ptr_type; assert(int_type->id == ZigTypeIdInt); LLVMValueRef fn_val = get_int_overflow_fn(g, int_type, add_sub_mul); LLVMValueRef op1 = ir_llvm_value(g, instruction->op1); LLVMValueRef op2 = ir_llvm_value(g, instruction->op2); LLVMValueRef ptr_result = ir_llvm_value(g, instruction->result_ptr); LLVMValueRef params[] = { op1, op2, }; LLVMValueRef result_struct = LLVMBuildCall(g->builder, fn_val, params, 2, ""); LLVMValueRef result = LLVMBuildExtractValue(g->builder, result_struct, 0, ""); LLVMValueRef overflow_bit = LLVMBuildExtractValue(g->builder, result_struct, 1, ""); gen_store(g, result, ptr_result, instruction->result_ptr->value->type); return overflow_bit; } static LLVMValueRef ir_render_test_err(CodeGen *g, IrExecutableGen *executable, IrInstGenTestErr *instruction) { ZigType *err_union_type = instruction->err_union->value->type; ZigType *payload_type = err_union_type->data.error_union.payload_type; LLVMValueRef err_union_handle = ir_llvm_value(g, instruction->err_union); LLVMValueRef err_val; if (type_has_bits(g, payload_type)) { LLVMValueRef err_val_ptr = LLVMBuildStructGEP(g->builder, err_union_handle, err_union_err_index, ""); err_val = gen_load_untyped(g, err_val_ptr, 0, false, ""); } else { err_val = err_union_handle; } LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, g->err_tag_type)); return LLVMBuildICmp(g->builder, LLVMIntNE, err_val, zero, ""); } static LLVMValueRef ir_render_unwrap_err_code(CodeGen *g, IrExecutableGen *executable, IrInstGenUnwrapErrCode *instruction) { if (instruction->base.value->special != ConstValSpecialRuntime) return nullptr; ZigType *ptr_type = instruction->err_union_ptr->value->type; assert(ptr_type->id == ZigTypeIdPointer); ZigType *err_union_type = ptr_type->data.pointer.child_type; ZigType *payload_type = err_union_type->data.error_union.payload_type; LLVMValueRef err_union_ptr = ir_llvm_value(g, instruction->err_union_ptr); if (!type_has_bits(g, payload_type)) { return err_union_ptr; } else { // TODO assign undef to the payload LLVMValueRef err_union_handle = get_handle_value(g, err_union_ptr, err_union_type, ptr_type); return LLVMBuildStructGEP(g->builder, err_union_handle, err_union_err_index, ""); } } static LLVMValueRef ir_render_unwrap_err_payload(CodeGen *g, IrExecutableGen *executable, IrInstGenUnwrapErrPayload *instruction) { Error err; if (instruction->base.value->special != ConstValSpecialRuntime) return nullptr; bool want_safety = instruction->safety_check_on && ir_want_runtime_safety(g, &instruction->base) && g->errors_by_index.length > 1; ZigType *ptr_type = instruction->value->value->type; assert(ptr_type->id == ZigTypeIdPointer); ZigType *err_union_type = ptr_type->data.pointer.child_type; ZigType *payload_type = err_union_type->data.error_union.payload_type; LLVMValueRef err_union_ptr = ir_llvm_value(g, instruction->value); LLVMValueRef zero = LLVMConstNull(get_llvm_type(g, g->err_tag_type)); bool value_has_bits; if ((err = type_has_bits2(g, instruction->base.value->type, &value_has_bits))) codegen_report_errors_and_exit(g); if (!want_safety && !value_has_bits) { if (instruction->initializing) { gen_store_untyped(g, zero, err_union_ptr, 0, false); } return nullptr; } LLVMValueRef err_union_handle = get_handle_value(g, err_union_ptr, err_union_type, ptr_type); if (!type_has_bits(g, err_union_type->data.error_union.err_set_type)) { return err_union_handle; } if (want_safety) { LLVMValueRef err_val; if (type_has_bits(g, payload_type)) { LLVMValueRef err_val_ptr = LLVMBuildStructGEP(g->builder, err_union_handle, err_union_err_index, ""); err_val = gen_load_untyped(g, err_val_ptr, 0, false, ""); } else { err_val = err_union_handle; } LLVMValueRef cond_val = LLVMBuildICmp(g->builder, LLVMIntEQ, err_val, zero, ""); LLVMBasicBlockRef err_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnwrapErrError"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "UnwrapErrOk"); LLVMBuildCondBr(g->builder, cond_val, ok_block, err_block); LLVMPositionBuilderAtEnd(g->builder, err_block); gen_safety_crash_for_err(g, err_val, instruction->base.base.scope); LLVMPositionBuilderAtEnd(g->builder, ok_block); } if (type_has_bits(g, payload_type)) { if (instruction->initializing) { LLVMValueRef err_tag_ptr = LLVMBuildStructGEP(g->builder, err_union_handle, err_union_err_index, ""); LLVMValueRef ok_err_val = LLVMConstNull(get_llvm_type(g, g->err_tag_type)); gen_store_untyped(g, ok_err_val, err_tag_ptr, 0, false); } return LLVMBuildStructGEP(g->builder, err_union_handle, err_union_payload_index, ""); } else { if (instruction->initializing) { gen_store_untyped(g, zero, err_union_ptr, 0, false); } return nullptr; } } static LLVMValueRef ir_render_optional_wrap(CodeGen *g, IrExecutableGen *executable, IrInstGenOptionalWrap *instruction) { ZigType *wanted_type = instruction->base.value->type; assert(wanted_type->id == ZigTypeIdOptional); ZigType *child_type = wanted_type->data.maybe.child_type; if (!type_has_bits(g, child_type)) { LLVMValueRef result = LLVMConstAllOnes(LLVMInt1Type()); if (instruction->result_loc != nullptr) { LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); gen_store_untyped(g, result, result_loc, 0, false); } return result; } LLVMValueRef payload_val = ir_llvm_value(g, instruction->operand); if (!handle_is_ptr(g, wanted_type)) { if (instruction->result_loc != nullptr) { LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); gen_store_untyped(g, payload_val, result_loc, 0, false); } return payload_val; } LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); LLVMValueRef val_ptr = LLVMBuildStructGEP(g->builder, result_loc, maybe_child_index, ""); // child_type and instruction->value->value->type may differ by constness gen_assign_raw(g, val_ptr, get_pointer_to_type(g, child_type, false), payload_val); LLVMValueRef maybe_ptr = LLVMBuildStructGEP(g->builder, result_loc, maybe_null_index, ""); gen_store_untyped(g, LLVMConstAllOnes(LLVMInt1Type()), maybe_ptr, 0, false); return result_loc; } static LLVMValueRef ir_render_err_wrap_code(CodeGen *g, IrExecutableGen *executable, IrInstGenErrWrapCode *instruction) { ZigType *wanted_type = instruction->base.value->type; assert(wanted_type->id == ZigTypeIdErrorUnion); LLVMValueRef err_val = ir_llvm_value(g, instruction->operand); if (!handle_is_ptr(g, wanted_type)) return err_val; LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); LLVMValueRef err_tag_ptr = LLVMBuildStructGEP(g->builder, result_loc, err_union_err_index, ""); gen_store_untyped(g, err_val, err_tag_ptr, 0, false); // TODO store undef to the payload return result_loc; } static LLVMValueRef ir_render_err_wrap_payload(CodeGen *g, IrExecutableGen *executable, IrInstGenErrWrapPayload *instruction) { ZigType *wanted_type = instruction->base.value->type; assert(wanted_type->id == ZigTypeIdErrorUnion); ZigType *payload_type = wanted_type->data.error_union.payload_type; ZigType *err_set_type = wanted_type->data.error_union.err_set_type; if (!type_has_bits(g, err_set_type)) { return ir_llvm_value(g, instruction->operand); } LLVMValueRef ok_err_val = LLVMConstNull(get_llvm_type(g, g->err_tag_type)); if (!type_has_bits(g, payload_type)) return ok_err_val; LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); LLVMValueRef payload_val = ir_llvm_value(g, instruction->operand); LLVMValueRef err_tag_ptr = LLVMBuildStructGEP(g->builder, result_loc, err_union_err_index, ""); gen_store_untyped(g, ok_err_val, err_tag_ptr, 0, false); LLVMValueRef payload_ptr = LLVMBuildStructGEP(g->builder, result_loc, err_union_payload_index, ""); gen_assign_raw(g, payload_ptr, get_pointer_to_type(g, payload_type, false), payload_val); return result_loc; } static LLVMValueRef ir_render_union_tag(CodeGen *g, IrExecutableGen *executable, IrInstGenUnionTag *instruction) { ZigType *union_type = instruction->value->value->type; ZigType *tag_type = union_type->data.unionation.tag_type; if (!type_has_bits(g, tag_type)) return nullptr; LLVMValueRef union_val = ir_llvm_value(g, instruction->value); if (union_type->data.unionation.gen_field_count == 0) return union_val; assert(union_type->data.unionation.gen_tag_index != SIZE_MAX); LLVMValueRef tag_field_ptr = LLVMBuildStructGEP(g->builder, union_val, union_type->data.unionation.gen_tag_index, ""); ZigType *ptr_type = get_pointer_to_type(g, tag_type, false); return get_handle_value(g, tag_field_ptr, tag_type, ptr_type); } static LLVMValueRef ir_render_panic(CodeGen *g, IrExecutableGen *executable, IrInstGenPanic *instruction) { bool is_llvm_alloca; LLVMValueRef err_ret_trace_val = get_cur_err_ret_trace_val(g, instruction->base.base.scope, &is_llvm_alloca); gen_panic(g, ir_llvm_value(g, instruction->msg), err_ret_trace_val, is_llvm_alloca); return nullptr; } static LLVMValueRef ir_render_atomic_rmw(CodeGen *g, IrExecutableGen *executable, IrInstGenAtomicRmw *instruction) { bool is_signed; ZigType *operand_type = instruction->operand->value->type; bool is_float = operand_type->id == ZigTypeIdFloat; if (operand_type->id == ZigTypeIdInt) { is_signed = operand_type->data.integral.is_signed; } else { is_signed = false; } enum ZigLLVM_AtomicRMWBinOp op = to_ZigLLVMAtomicRMWBinOp(instruction->op, is_signed, is_float); LLVMAtomicOrdering ordering = to_LLVMAtomicOrdering(instruction->ordering); LLVMValueRef ptr = ir_llvm_value(g, instruction->ptr); LLVMValueRef operand = ir_llvm_value(g, instruction->operand); LLVMTypeRef actual_abi_type = get_atomic_abi_type(g, instruction->ptr); if (actual_abi_type != nullptr) { // operand needs widening and truncating LLVMValueRef casted_ptr = LLVMBuildBitCast(g->builder, ptr, LLVMPointerType(actual_abi_type, 0), ""); LLVMValueRef casted_operand; if (operand_type->data.integral.is_signed) { casted_operand = LLVMBuildSExt(g->builder, operand, actual_abi_type, ""); } else { casted_operand = LLVMBuildZExt(g->builder, operand, actual_abi_type, ""); } LLVMValueRef uncasted_result = ZigLLVMBuildAtomicRMW(g->builder, op, casted_ptr, casted_operand, ordering, g->is_single_threaded); return LLVMBuildTrunc(g->builder, uncasted_result, get_llvm_type(g, operand_type), ""); } if (get_codegen_ptr_type_bail(g, operand_type) == nullptr) { return ZigLLVMBuildAtomicRMW(g->builder, op, ptr, operand, ordering, g->is_single_threaded); } // it's a pointer but we need to treat it as an int LLVMValueRef casted_ptr = LLVMBuildBitCast(g->builder, ptr, LLVMPointerType(g->builtin_types.entry_usize->llvm_type, 0), ""); LLVMValueRef casted_operand = LLVMBuildPtrToInt(g->builder, operand, g->builtin_types.entry_usize->llvm_type, ""); LLVMValueRef uncasted_result = ZigLLVMBuildAtomicRMW(g->builder, op, casted_ptr, casted_operand, ordering, g->is_single_threaded); return LLVMBuildIntToPtr(g->builder, uncasted_result, get_llvm_type(g, operand_type), ""); } static LLVMValueRef ir_render_atomic_load(CodeGen *g, IrExecutableGen *executable, IrInstGenAtomicLoad *instruction) { LLVMAtomicOrdering ordering = to_LLVMAtomicOrdering(instruction->ordering); LLVMValueRef ptr = ir_llvm_value(g, instruction->ptr); ZigType *operand_type = instruction->ptr->value->type->data.pointer.child_type; LLVMTypeRef actual_abi_type = get_atomic_abi_type(g, instruction->ptr); if (actual_abi_type != nullptr) { // operand needs widening and truncating ptr = LLVMBuildBitCast(g->builder, ptr, LLVMPointerType(actual_abi_type, 0), ""); LLVMValueRef load_inst = gen_load(g, ptr, instruction->ptr->value->type, ""); LLVMSetOrdering(load_inst, ordering); return LLVMBuildTrunc(g->builder, load_inst, get_llvm_type(g, operand_type), ""); } LLVMValueRef load_inst = gen_load(g, ptr, instruction->ptr->value->type, ""); LLVMSetOrdering(load_inst, ordering); return load_inst; } static LLVMValueRef ir_render_atomic_store(CodeGen *g, IrExecutableGen *executable, IrInstGenAtomicStore *instruction) { LLVMAtomicOrdering ordering = to_LLVMAtomicOrdering(instruction->ordering); LLVMValueRef ptr = ir_llvm_value(g, instruction->ptr); LLVMValueRef value = ir_llvm_value(g, instruction->value); LLVMTypeRef actual_abi_type = get_atomic_abi_type(g, instruction->ptr); if (actual_abi_type != nullptr) { // operand needs widening ptr = LLVMBuildBitCast(g->builder, ptr, LLVMPointerType(actual_abi_type, 0), ""); if (instruction->value->value->type->data.integral.is_signed) { value = LLVMBuildSExt(g->builder, value, actual_abi_type, ""); } else { value = LLVMBuildZExt(g->builder, value, actual_abi_type, ""); } } LLVMValueRef store_inst = gen_store(g, value, ptr, instruction->ptr->value->type); LLVMSetOrdering(store_inst, ordering); return nullptr; } static LLVMValueRef ir_render_float_op(CodeGen *g, IrExecutableGen *executable, IrInstGenFloatOp *instruction) { LLVMValueRef operand = ir_llvm_value(g, instruction->operand); LLVMValueRef fn_val = get_float_fn(g, instruction->base.value->type, ZigLLVMFnIdFloatOp, instruction->fn_id); return LLVMBuildCall(g->builder, fn_val, &operand, 1, ""); } static LLVMValueRef ir_render_mul_add(CodeGen *g, IrExecutableGen *executable, IrInstGenMulAdd *instruction) { LLVMValueRef op1 = ir_llvm_value(g, instruction->op1); LLVMValueRef op2 = ir_llvm_value(g, instruction->op2); LLVMValueRef op3 = ir_llvm_value(g, instruction->op3); assert(instruction->base.value->type->id == ZigTypeIdFloat || instruction->base.value->type->id == ZigTypeIdVector); LLVMValueRef fn_val = get_float_fn(g, instruction->base.value->type, ZigLLVMFnIdFMA, BuiltinFnIdMulAdd); LLVMValueRef args[3] = { op1, op2, op3, }; return LLVMBuildCall(g->builder, fn_val, args, 3, ""); } static LLVMValueRef ir_render_bswap(CodeGen *g, IrExecutableGen *executable, IrInstGenBswap *instruction) { LLVMValueRef op = ir_llvm_value(g, instruction->op); ZigType *expr_type = instruction->base.value->type; bool is_vector = expr_type->id == ZigTypeIdVector; ZigType *int_type = is_vector ? expr_type->data.vector.elem_type : expr_type; assert(int_type->id == ZigTypeIdInt); if (int_type->data.integral.bit_count % 16 == 0) { LLVMValueRef fn_val = get_int_builtin_fn(g, expr_type, BuiltinFnIdBswap); return LLVMBuildCall(g->builder, fn_val, &op, 1, ""); } // Not an even number of bytes, so we zext 1 byte, then bswap, shift right 1 byte, truncate ZigType *extended_type = get_int_type(g, int_type->data.integral.is_signed, int_type->data.integral.bit_count + 8); LLVMValueRef shift_amt = LLVMConstInt(get_llvm_type(g, extended_type), 8, false); if (is_vector) { extended_type = get_vector_type(g, expr_type->data.vector.len, extended_type); LLVMValueRef *values = heap::c_allocator.allocate_nonzero(expr_type->data.vector.len); for (uint32_t i = 0; i < expr_type->data.vector.len; i += 1) { values[i] = shift_amt; } shift_amt = LLVMConstVector(values, expr_type->data.vector.len); heap::c_allocator.deallocate(values, expr_type->data.vector.len); } // aabbcc LLVMValueRef extended = LLVMBuildZExt(g->builder, op, get_llvm_type(g, extended_type), ""); // 00aabbcc LLVMValueRef fn_val = get_int_builtin_fn(g, extended_type, BuiltinFnIdBswap); LLVMValueRef swapped = LLVMBuildCall(g->builder, fn_val, &extended, 1, ""); // ccbbaa00 LLVMValueRef shifted = ZigLLVMBuildLShrExact(g->builder, swapped, shift_amt, ""); // 00ccbbaa return LLVMBuildTrunc(g->builder, shifted, get_llvm_type(g, expr_type), ""); } static LLVMValueRef ir_render_extern(CodeGen *g, IrExecutableGen *executable, IrInstGenExtern *instruction) { ZigType *expr_type = instruction->base.value->type; assert(get_src_ptr_type(expr_type)); const char *symbol_name = buf_ptr(instruction->name); const LLVMLinkage linkage = to_llvm_linkage(instruction->linkage, true); LLVMValueRef global_value = LLVMGetNamedGlobal(g->module, symbol_name); if (global_value == nullptr) { global_value = LLVMAddGlobal(g->module, get_llvm_type(g, expr_type), symbol_name); LLVMSetLinkage(global_value, linkage); LLVMSetGlobalConstant(global_value, true); if (instruction->is_thread_local) LLVMSetThreadLocalMode(global_value, LLVMGeneralDynamicTLSModel); } else if (LLVMGetLinkage(global_value) != linkage) { // XXX: Handle this case better! zig_panic("duplicate extern symbol"); } return LLVMBuildBitCast(g->builder, global_value, get_llvm_type(g, expr_type), ""); } static LLVMValueRef ir_render_bit_reverse(CodeGen *g, IrExecutableGen *executable, IrInstGenBitReverse *instruction) { LLVMValueRef op = ir_llvm_value(g, instruction->op); ZigType *int_type = instruction->base.value->type; assert(int_type->id == ZigTypeIdInt); LLVMValueRef fn_val = get_int_builtin_fn(g, instruction->base.value->type, BuiltinFnIdBitReverse); return LLVMBuildCall(g->builder, fn_val, &op, 1, ""); } static LLVMValueRef ir_render_vector_to_array(CodeGen *g, IrExecutableGen *executable, IrInstGenVectorToArray *instruction) { ZigType *array_type = instruction->base.value->type; assert(array_type->id == ZigTypeIdArray); assert(handle_is_ptr(g, array_type)); LLVMValueRef result_loc = ir_llvm_value(g, instruction->result_loc); LLVMValueRef vector = ir_llvm_value(g, instruction->vector); ZigType *elem_type = array_type->data.array.child_type; bool bitcast_ok = elem_type->size_in_bits == elem_type->abi_size * 8; if (bitcast_ok) { LLVMValueRef casted_ptr = LLVMBuildBitCast(g->builder, result_loc, LLVMPointerType(get_llvm_type(g, instruction->vector->value->type), 0), ""); uint32_t alignment = get_ptr_align(g, instruction->result_loc->value->type); gen_store_untyped(g, vector, casted_ptr, alignment, false); } else { // If the ABI size of the element type is not evenly divisible by size_in_bits, a simple bitcast // will not work, and we fall back to extractelement. LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMTypeRef u32_type_ref = LLVMInt32Type(); LLVMValueRef zero = LLVMConstInt(usize_type_ref, 0, false); for (uintptr_t i = 0; i < instruction->vector->value->type->data.vector.len; i++) { LLVMValueRef index_usize = LLVMConstInt(usize_type_ref, i, false); LLVMValueRef index_u32 = LLVMConstInt(u32_type_ref, i, false); LLVMValueRef indexes[] = { zero, index_usize }; LLVMValueRef elem_ptr = LLVMBuildInBoundsGEP(g->builder, result_loc, indexes, 2, ""); LLVMValueRef elem = LLVMBuildExtractElement(g->builder, vector, index_u32, ""); LLVMBuildStore(g->builder, elem, elem_ptr); } } return result_loc; } static LLVMValueRef ir_render_array_to_vector(CodeGen *g, IrExecutableGen *executable, IrInstGenArrayToVector *instruction) { ZigType *vector_type = instruction->base.value->type; assert(vector_type->id == ZigTypeIdVector); assert(!handle_is_ptr(g, vector_type)); LLVMValueRef array_ptr = ir_llvm_value(g, instruction->array); LLVMTypeRef vector_type_ref = get_llvm_type(g, vector_type); ZigType *elem_type = vector_type->data.vector.elem_type; bool bitcast_ok = elem_type->size_in_bits == elem_type->abi_size * 8; if (bitcast_ok) { LLVMValueRef casted_ptr = LLVMBuildBitCast(g->builder, array_ptr, LLVMPointerType(vector_type_ref, 0), ""); ZigType *array_type = instruction->array->value->type; assert(array_type->id == ZigTypeIdArray); uint32_t alignment = get_abi_alignment(g, array_type->data.array.child_type); return gen_load_untyped(g, casted_ptr, alignment, false, ""); } else { // If the ABI size of the element type is not evenly divisible by size_in_bits, a simple bitcast // will not work, and we fall back to insertelement. LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMTypeRef u32_type_ref = LLVMInt32Type(); LLVMValueRef zero = LLVMConstInt(usize_type_ref, 0, false); LLVMValueRef vector = LLVMGetUndef(vector_type_ref); for (uintptr_t i = 0; i < instruction->base.value->type->data.vector.len; i++) { LLVMValueRef index_usize = LLVMConstInt(usize_type_ref, i, false); LLVMValueRef index_u32 = LLVMConstInt(u32_type_ref, i, false); LLVMValueRef indexes[] = { zero, index_usize }; LLVMValueRef elem_ptr = LLVMBuildInBoundsGEP(g->builder, array_ptr, indexes, 2, ""); LLVMValueRef elem = LLVMBuildLoad(g->builder, elem_ptr, ""); vector = LLVMBuildInsertElement(g->builder, vector, elem, index_u32, ""); } return vector; } } static LLVMValueRef ir_render_assert_zero(CodeGen *g, IrExecutableGen *executable, IrInstGenAssertZero *instruction) { LLVMValueRef target = ir_llvm_value(g, instruction->target); ZigType *int_type = instruction->target->value->type; if (ir_want_runtime_safety(g, &instruction->base)) { return gen_assert_zero(g, target, int_type); } return nullptr; } static LLVMValueRef ir_render_assert_non_null(CodeGen *g, IrExecutableGen *executable, IrInstGenAssertNonNull *instruction) { LLVMValueRef target = ir_llvm_value(g, instruction->target); ZigType *target_type = instruction->target->value->type; if (target_type->id == ZigTypeIdPointer) { assert(target_type->data.pointer.ptr_len == PtrLenC); LLVMValueRef non_null_bit = LLVMBuildICmp(g->builder, LLVMIntNE, target, LLVMConstNull(get_llvm_type(g, target_type)), ""); LLVMBasicBlockRef fail_block = LLVMAppendBasicBlock(g->cur_fn_val, "AssertNonNullFail"); LLVMBasicBlockRef ok_block = LLVMAppendBasicBlock(g->cur_fn_val, "AssertNonNullOk"); LLVMBuildCondBr(g->builder, non_null_bit, ok_block, fail_block); LLVMPositionBuilderAtEnd(g->builder, fail_block); gen_assertion(g, PanicMsgIdUnwrapOptionalFail, &instruction->base); LLVMPositionBuilderAtEnd(g->builder, ok_block); } else { zig_unreachable(); } return nullptr; } static LLVMValueRef ir_render_suspend_begin(CodeGen *g, IrExecutableGen *executable, IrInstGenSuspendBegin *instruction) { if (fn_is_async(g->cur_fn)) { instruction->resume_bb = gen_suspend_begin(g, "SuspendResume"); } return nullptr; } static LLVMValueRef ir_render_suspend_finish(CodeGen *g, IrExecutableGen *executable, IrInstGenSuspendFinish *instruction) { LLVMBuildRetVoid(g->builder); LLVMPositionBuilderAtEnd(g->builder, instruction->begin->resume_bb); if (ir_want_runtime_safety(g, &instruction->base)) { LLVMBuildStore(g->builder, g->cur_bad_not_suspended_index, g->cur_async_resume_index_ptr); } render_async_var_decls(g, instruction->base.base.scope); return nullptr; } static LLVMValueRef gen_await_early_return(CodeGen *g, IrInstGen *source_instr, LLVMValueRef target_frame_ptr, ZigType *result_type, ZigType *ptr_result_type, LLVMValueRef result_loc, bool non_async) { LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMValueRef their_result_ptr = nullptr; if (type_has_bits(g, result_type) && (non_async || result_loc != nullptr)) { LLVMValueRef their_result_ptr_ptr = LLVMBuildStructGEP(g->builder, target_frame_ptr, frame_ret_start, ""); their_result_ptr = LLVMBuildLoad(g->builder, their_result_ptr_ptr, ""); if (result_loc != nullptr) { LLVMTypeRef ptr_u8 = LLVMPointerType(LLVMInt8Type(), 0); LLVMValueRef dest_ptr_casted = LLVMBuildBitCast(g->builder, result_loc, ptr_u8, ""); LLVMValueRef src_ptr_casted = LLVMBuildBitCast(g->builder, their_result_ptr, ptr_u8, ""); bool is_volatile = false; uint32_t abi_align = get_abi_alignment(g, result_type); LLVMValueRef byte_count_val = LLVMConstInt(usize_type_ref, type_size(g, result_type), false); ZigLLVMBuildMemCpy(g->builder, dest_ptr_casted, abi_align, src_ptr_casted, abi_align, byte_count_val, is_volatile); } } if (codegen_fn_has_err_ret_tracing_arg(g, result_type)) { LLVMValueRef their_trace_ptr_ptr = LLVMBuildStructGEP(g->builder, target_frame_ptr, frame_index_trace_arg(g, result_type), ""); LLVMValueRef src_trace_ptr = LLVMBuildLoad(g->builder, their_trace_ptr_ptr, ""); bool is_llvm_alloca; LLVMValueRef dest_trace_ptr = get_cur_err_ret_trace_val(g, source_instr->base.scope, &is_llvm_alloca); LLVMValueRef args[] = { dest_trace_ptr, src_trace_ptr }; ZigLLVMBuildCall(g->builder, get_merge_err_ret_traces_fn_val(g), args, 2, get_llvm_cc(g, CallingConventionUnspecified), ZigLLVM_CallAttrAuto, ""); } if (non_async && type_has_bits(g, result_type)) { LLVMValueRef result_ptr = (result_loc == nullptr) ? their_result_ptr : result_loc; return get_handle_value(g, result_ptr, result_type, ptr_result_type); } else { return nullptr; } } static LLVMValueRef ir_render_await(CodeGen *g, IrExecutableGen *executable, IrInstGenAwait *instruction) { LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMValueRef zero = LLVMConstNull(usize_type_ref); LLVMValueRef target_frame_ptr = ir_llvm_value(g, instruction->frame); ZigType *result_type = instruction->base.value->type; ZigType *ptr_result_type = get_pointer_to_type(g, result_type, true); LLVMValueRef result_loc = (instruction->result_loc == nullptr) ? nullptr : ir_llvm_value(g, instruction->result_loc); if (instruction->is_nosuspend || (instruction->target_fn != nullptr && !fn_is_async(instruction->target_fn))) { return gen_await_early_return(g, &instruction->base, target_frame_ptr, result_type, ptr_result_type, result_loc, true); } // Prepare to be suspended LLVMBasicBlockRef resume_bb = gen_suspend_begin(g, "AwaitResume"); LLVMBasicBlockRef end_bb = LLVMAppendBasicBlock(g->cur_fn_val, "AwaitEnd"); // At this point resuming the function will continue from resume_bb. // This code is as if it is running inside the suspend block. // supply the awaiter return pointer if (type_has_bits(g, result_type)) { LLVMValueRef awaiter_ret_ptr_ptr = LLVMBuildStructGEP(g->builder, target_frame_ptr, frame_ret_start + 1, ""); if (result_loc == nullptr) { // no copy needed LLVMBuildStore(g->builder, LLVMConstNull(LLVMGetElementType(LLVMTypeOf(awaiter_ret_ptr_ptr))), awaiter_ret_ptr_ptr); } else { LLVMBuildStore(g->builder, result_loc, awaiter_ret_ptr_ptr); } } // supply the error return trace pointer if (codegen_fn_has_err_ret_tracing_arg(g, result_type)) { bool is_llvm_alloca; LLVMValueRef my_err_ret_trace_val = get_cur_err_ret_trace_val(g, instruction->base.base.scope, &is_llvm_alloca); assert(my_err_ret_trace_val != nullptr); LLVMValueRef err_ret_trace_ptr_ptr = LLVMBuildStructGEP(g->builder, target_frame_ptr, frame_index_trace_arg(g, result_type) + 1, ""); LLVMBuildStore(g->builder, my_err_ret_trace_val, err_ret_trace_ptr_ptr); } // caller's own frame pointer LLVMValueRef awaiter_init_val = LLVMBuildPtrToInt(g->builder, g->cur_frame_ptr, usize_type_ref, ""); LLVMValueRef awaiter_ptr = LLVMBuildStructGEP(g->builder, target_frame_ptr, frame_awaiter_index, ""); LLVMValueRef prev_val = gen_maybe_atomic_op(g, LLVMAtomicRMWBinOpXchg, awaiter_ptr, awaiter_init_val, LLVMAtomicOrderingRelease); LLVMBasicBlockRef bad_await_block = LLVMAppendBasicBlock(g->cur_fn_val, "BadAwait"); LLVMBasicBlockRef complete_suspend_block = LLVMAppendBasicBlock(g->cur_fn_val, "CompleteSuspend"); LLVMBasicBlockRef early_return_block = LLVMAppendBasicBlock(g->cur_fn_val, "EarlyReturn"); LLVMValueRef all_ones = LLVMConstAllOnes(usize_type_ref); LLVMValueRef switch_instr = LLVMBuildSwitch(g->builder, prev_val, bad_await_block, 2); LLVMAddCase(switch_instr, zero, complete_suspend_block); LLVMAddCase(switch_instr, all_ones, early_return_block); // We discovered that another awaiter was already here. LLVMPositionBuilderAtEnd(g->builder, bad_await_block); gen_assertion(g, PanicMsgIdBadAwait, &instruction->base); // Rely on the target to resume us from suspension. LLVMPositionBuilderAtEnd(g->builder, complete_suspend_block); LLVMBuildRetVoid(g->builder); // Early return: The async function has already completed. We must copy the result and // the error return trace if applicable. LLVMPositionBuilderAtEnd(g->builder, early_return_block); gen_await_early_return(g, &instruction->base, target_frame_ptr, result_type, ptr_result_type, result_loc, false); LLVMBuildBr(g->builder, end_bb); LLVMPositionBuilderAtEnd(g->builder, resume_bb); gen_assert_resume_id(g, &instruction->base, ResumeIdReturn, PanicMsgIdResumedAnAwaitingFn, nullptr); LLVMBuildBr(g->builder, end_bb); LLVMPositionBuilderAtEnd(g->builder, end_bb); // Rely on the spill for the llvm_value to be populated. // See the implementation of ir_llvm_value. return nullptr; } static LLVMValueRef ir_render_resume(CodeGen *g, IrExecutableGen *executable, IrInstGenResume *instruction) { LLVMValueRef frame = ir_llvm_value(g, instruction->frame); ZigType *frame_type = instruction->frame->value->type; assert(frame_type->id == ZigTypeIdAnyFrame); gen_resume(g, nullptr, frame, ResumeIdManual); return nullptr; } static LLVMValueRef ir_render_frame_size(CodeGen *g, IrExecutableGen *executable, IrInstGenFrameSize *instruction) { LLVMValueRef fn_val = ir_llvm_value(g, instruction->fn); return gen_frame_size(g, fn_val); } static LLVMValueRef ir_render_spill_begin(CodeGen *g, IrExecutableGen *executable, IrInstGenSpillBegin *instruction) { if (!fn_is_async(g->cur_fn)) return nullptr; switch (instruction->spill_id) { case SpillIdInvalid: zig_unreachable(); case SpillIdRetErrCode: { LLVMValueRef operand = ir_llvm_value(g, instruction->operand); LLVMValueRef ptr = ir_llvm_value(g, g->cur_fn->err_code_spill); LLVMBuildStore(g->builder, operand, ptr); return nullptr; } } zig_unreachable(); } static LLVMValueRef ir_render_spill_end(CodeGen *g, IrExecutableGen *executable, IrInstGenSpillEnd *instruction) { if (!fn_is_async(g->cur_fn)) return ir_llvm_value(g, instruction->begin->operand); switch (instruction->begin->spill_id) { case SpillIdInvalid: zig_unreachable(); case SpillIdRetErrCode: { LLVMValueRef ptr = ir_llvm_value(g, g->cur_fn->err_code_spill); return LLVMBuildLoad(g->builder, ptr, ""); } } zig_unreachable(); } static LLVMValueRef ir_render_vector_extract_elem(CodeGen *g, IrExecutableGen *executable, IrInstGenVectorExtractElem *instruction) { LLVMValueRef vector = ir_llvm_value(g, instruction->vector); LLVMValueRef index = ir_llvm_value(g, instruction->index); return LLVMBuildExtractElement(g->builder, vector, index, ""); } static void set_debug_location(CodeGen *g, IrInstGen *instruction) { AstNode *source_node = instruction->base.source_node; Scope *scope = instruction->base.scope; assert(source_node); assert(scope); ZigLLVMSetCurrentDebugLocation(g->builder, (int)source_node->line + 1, (int)source_node->column + 1, get_di_scope(g, scope)); } static LLVMValueRef ir_render_instruction(CodeGen *g, IrExecutableGen *executable, IrInstGen *instruction) { switch (instruction->id) { case IrInstGenIdInvalid: case IrInstGenIdConst: case IrInstGenIdAlloca: zig_unreachable(); case IrInstGenIdDeclVar: return ir_render_decl_var(g, executable, (IrInstGenDeclVar *)instruction); case IrInstGenIdReturn: return ir_render_return(g, executable, (IrInstGenReturn *)instruction); case IrInstGenIdBinOp: return ir_render_bin_op(g, executable, (IrInstGenBinOp *)instruction); case IrInstGenIdCast: return ir_render_cast(g, executable, (IrInstGenCast *)instruction); case IrInstGenIdUnreachable: return ir_render_unreachable(g, executable, (IrInstGenUnreachable *)instruction); case IrInstGenIdCondBr: return ir_render_cond_br(g, executable, (IrInstGenCondBr *)instruction); case IrInstGenIdBr: return ir_render_br(g, executable, (IrInstGenBr *)instruction); case IrInstGenIdBinaryNot: return ir_render_binary_not(g, executable, (IrInstGenBinaryNot *)instruction); case IrInstGenIdNegation: return ir_render_negation(g, executable, (IrInstGenNegation *)instruction); case IrInstGenIdLoadPtr: return ir_render_load_ptr(g, executable, (IrInstGenLoadPtr *)instruction); case IrInstGenIdStorePtr: return ir_render_store_ptr(g, executable, (IrInstGenStorePtr *)instruction); case IrInstGenIdVectorStoreElem: return ir_render_vector_store_elem(g, executable, (IrInstGenVectorStoreElem *)instruction); case IrInstGenIdVarPtr: return ir_render_var_ptr(g, executable, (IrInstGenVarPtr *)instruction); case IrInstGenIdReturnPtr: return ir_render_return_ptr(g, executable, (IrInstGenReturnPtr *)instruction); case IrInstGenIdElemPtr: return ir_render_elem_ptr(g, executable, (IrInstGenElemPtr *)instruction); case IrInstGenIdCall: return ir_render_call(g, executable, (IrInstGenCall *)instruction); case IrInstGenIdStructFieldPtr: return ir_render_struct_field_ptr(g, executable, (IrInstGenStructFieldPtr *)instruction); case IrInstGenIdUnionFieldPtr: return ir_render_union_field_ptr(g, executable, (IrInstGenUnionFieldPtr *)instruction); case IrInstGenIdAsm: return ir_render_asm_gen(g, executable, (IrInstGenAsm *)instruction); case IrInstGenIdTestNonNull: return ir_render_test_non_null(g, executable, (IrInstGenTestNonNull *)instruction); case IrInstGenIdOptionalUnwrapPtr: return ir_render_optional_unwrap_ptr(g, executable, (IrInstGenOptionalUnwrapPtr *)instruction); case IrInstGenIdClz: return ir_render_clz(g, executable, (IrInstGenClz *)instruction); case IrInstGenIdCtz: return ir_render_ctz(g, executable, (IrInstGenCtz *)instruction); case IrInstGenIdPopCount: return ir_render_pop_count(g, executable, (IrInstGenPopCount *)instruction); case IrInstGenIdSwitchBr: return ir_render_switch_br(g, executable, (IrInstGenSwitchBr *)instruction); case IrInstGenIdBswap: return ir_render_bswap(g, executable, (IrInstGenBswap *)instruction); case IrInstGenIdBitReverse: return ir_render_bit_reverse(g, executable, (IrInstGenBitReverse *)instruction); case IrInstGenIdPhi: return ir_render_phi(g, executable, (IrInstGenPhi *)instruction); case IrInstGenIdRef: return ir_render_ref(g, executable, (IrInstGenRef *)instruction); case IrInstGenIdErrName: return ir_render_err_name(g, executable, (IrInstGenErrName *)instruction); case IrInstGenIdCmpxchg: return ir_render_cmpxchg(g, executable, (IrInstGenCmpxchg *)instruction); case IrInstGenIdFence: return ir_render_fence(g, executable, (IrInstGenFence *)instruction); case IrInstGenIdReduce: return ir_render_reduce(g, executable, (IrInstGenReduce *)instruction); case IrInstGenIdTruncate: return ir_render_truncate(g, executable, (IrInstGenTruncate *)instruction); case IrInstGenIdBoolNot: return ir_render_bool_not(g, executable, (IrInstGenBoolNot *)instruction); case IrInstGenIdMemset: return ir_render_memset(g, executable, (IrInstGenMemset *)instruction); case IrInstGenIdMemcpy: return ir_render_memcpy(g, executable, (IrInstGenMemcpy *)instruction); case IrInstGenIdSlice: return ir_render_slice(g, executable, (IrInstGenSlice *)instruction); case IrInstGenIdBreakpoint: return ir_render_breakpoint(g, executable, (IrInstGenBreakpoint *)instruction); case IrInstGenIdReturnAddress: return ir_render_return_address(g, executable, (IrInstGenReturnAddress *)instruction); case IrInstGenIdFrameAddress: return ir_render_frame_address(g, executable, (IrInstGenFrameAddress *)instruction); case IrInstGenIdFrameHandle: return ir_render_handle(g, executable, (IrInstGenFrameHandle *)instruction); case IrInstGenIdOverflowOp: return ir_render_overflow_op(g, executable, (IrInstGenOverflowOp *)instruction); case IrInstGenIdTestErr: return ir_render_test_err(g, executable, (IrInstGenTestErr *)instruction); case IrInstGenIdUnwrapErrCode: return ir_render_unwrap_err_code(g, executable, (IrInstGenUnwrapErrCode *)instruction); case IrInstGenIdUnwrapErrPayload: return ir_render_unwrap_err_payload(g, executable, (IrInstGenUnwrapErrPayload *)instruction); case IrInstGenIdOptionalWrap: return ir_render_optional_wrap(g, executable, (IrInstGenOptionalWrap *)instruction); case IrInstGenIdErrWrapCode: return ir_render_err_wrap_code(g, executable, (IrInstGenErrWrapCode *)instruction); case IrInstGenIdErrWrapPayload: return ir_render_err_wrap_payload(g, executable, (IrInstGenErrWrapPayload *)instruction); case IrInstGenIdUnionTag: return ir_render_union_tag(g, executable, (IrInstGenUnionTag *)instruction); case IrInstGenIdPtrCast: return ir_render_ptr_cast(g, executable, (IrInstGenPtrCast *)instruction); case IrInstGenIdBitCast: return ir_render_bit_cast(g, executable, (IrInstGenBitCast *)instruction); case IrInstGenIdWidenOrShorten: return ir_render_widen_or_shorten(g, executable, (IrInstGenWidenOrShorten *)instruction); case IrInstGenIdPtrToInt: return ir_render_ptr_to_int(g, executable, (IrInstGenPtrToInt *)instruction); case IrInstGenIdIntToPtr: return ir_render_int_to_ptr(g, executable, (IrInstGenIntToPtr *)instruction); case IrInstGenIdIntToEnum: return ir_render_int_to_enum(g, executable, (IrInstGenIntToEnum *)instruction); case IrInstGenIdIntToErr: return ir_render_int_to_err(g, executable, (IrInstGenIntToErr *)instruction); case IrInstGenIdErrToInt: return ir_render_err_to_int(g, executable, (IrInstGenErrToInt *)instruction); case IrInstGenIdPanic: return ir_render_panic(g, executable, (IrInstGenPanic *)instruction); case IrInstGenIdTagName: return ir_render_enum_tag_name(g, executable, (IrInstGenTagName *)instruction); case IrInstGenIdFieldParentPtr: return ir_render_field_parent_ptr(g, executable, (IrInstGenFieldParentPtr *)instruction); case IrInstGenIdAlignCast: return ir_render_align_cast(g, executable, (IrInstGenAlignCast *)instruction); case IrInstGenIdErrorReturnTrace: return ir_render_error_return_trace(g, executable, (IrInstGenErrorReturnTrace *)instruction); case IrInstGenIdAtomicRmw: return ir_render_atomic_rmw(g, executable, (IrInstGenAtomicRmw *)instruction); case IrInstGenIdAtomicLoad: return ir_render_atomic_load(g, executable, (IrInstGenAtomicLoad *)instruction); case IrInstGenIdAtomicStore: return ir_render_atomic_store(g, executable, (IrInstGenAtomicStore *)instruction); case IrInstGenIdSaveErrRetAddr: return ir_render_save_err_ret_addr(g, executable, (IrInstGenSaveErrRetAddr *)instruction); case IrInstGenIdFloatOp: return ir_render_float_op(g, executable, (IrInstGenFloatOp *)instruction); case IrInstGenIdMulAdd: return ir_render_mul_add(g, executable, (IrInstGenMulAdd *)instruction); case IrInstGenIdArrayToVector: return ir_render_array_to_vector(g, executable, (IrInstGenArrayToVector *)instruction); case IrInstGenIdVectorToArray: return ir_render_vector_to_array(g, executable, (IrInstGenVectorToArray *)instruction); case IrInstGenIdAssertZero: return ir_render_assert_zero(g, executable, (IrInstGenAssertZero *)instruction); case IrInstGenIdAssertNonNull: return ir_render_assert_non_null(g, executable, (IrInstGenAssertNonNull *)instruction); case IrInstGenIdPtrOfArrayToSlice: return ir_render_ptr_of_array_to_slice(g, executable, (IrInstGenPtrOfArrayToSlice *)instruction); case IrInstGenIdSuspendBegin: return ir_render_suspend_begin(g, executable, (IrInstGenSuspendBegin *)instruction); case IrInstGenIdSuspendFinish: return ir_render_suspend_finish(g, executable, (IrInstGenSuspendFinish *)instruction); case IrInstGenIdResume: return ir_render_resume(g, executable, (IrInstGenResume *)instruction); case IrInstGenIdFrameSize: return ir_render_frame_size(g, executable, (IrInstGenFrameSize *)instruction); case IrInstGenIdAwait: return ir_render_await(g, executable, (IrInstGenAwait *)instruction); case IrInstGenIdSpillBegin: return ir_render_spill_begin(g, executable, (IrInstGenSpillBegin *)instruction); case IrInstGenIdSpillEnd: return ir_render_spill_end(g, executable, (IrInstGenSpillEnd *)instruction); case IrInstGenIdShuffleVector: return ir_render_shuffle_vector(g, executable, (IrInstGenShuffleVector *) instruction); case IrInstGenIdSplat: return ir_render_splat(g, executable, (IrInstGenSplat *) instruction); case IrInstGenIdVectorExtractElem: return ir_render_vector_extract_elem(g, executable, (IrInstGenVectorExtractElem *) instruction); case IrInstGenIdWasmMemorySize: return ir_render_wasm_memory_size(g, executable, (IrInstGenWasmMemorySize *) instruction); case IrInstGenIdWasmMemoryGrow: return ir_render_wasm_memory_grow(g, executable, (IrInstGenWasmMemoryGrow *) instruction); case IrInstGenIdExtern: return ir_render_extern(g, executable, (IrInstGenExtern *) instruction); } zig_unreachable(); } static void ir_render(CodeGen *g, ZigFn *fn_entry) { assert(fn_entry); IrExecutableGen *executable = &fn_entry->analyzed_executable; assert(executable->basic_block_list.length > 0); for (size_t block_i = 0; block_i < executable->basic_block_list.length; block_i += 1) { IrBasicBlockGen *current_block = executable->basic_block_list.at(block_i); if (get_scope_typeof(current_block->scope) != nullptr) { LLVMBuildBr(g->builder, current_block->llvm_block); } assert(current_block->llvm_block); LLVMPositionBuilderAtEnd(g->builder, current_block->llvm_block); for (size_t instr_i = 0; instr_i < current_block->instruction_list.length; instr_i += 1) { IrInstGen *instruction = current_block->instruction_list.at(instr_i); if (instruction->base.ref_count == 0 && !ir_inst_gen_has_side_effects(instruction)) continue; if (get_scope_typeof(instruction->base.scope) != nullptr) continue; if (!g->strip_debug_symbols) { set_debug_location(g, instruction); } instruction->llvm_value = ir_render_instruction(g, executable, instruction); if (instruction->spill != nullptr && instruction->llvm_value != nullptr) { LLVMValueRef spill_ptr = ir_llvm_value(g, instruction->spill); gen_assign_raw(g, spill_ptr, instruction->spill->value->type, instruction->llvm_value); instruction->llvm_value = nullptr; } } current_block->llvm_exit_block = LLVMGetInsertBlock(g->builder); } } static LLVMValueRef gen_const_ptr_struct_recursive(CodeGen *g, ZigValue *struct_const_val, size_t field_index); static LLVMValueRef gen_const_ptr_array_recursive(CodeGen *g, ZigValue *array_const_val, size_t index); static LLVMValueRef gen_const_ptr_union_recursive(CodeGen *g, ZigValue *union_const_val); static LLVMValueRef gen_const_ptr_err_union_code_recursive(CodeGen *g, ZigValue *err_union_const_val); static LLVMValueRef gen_const_ptr_err_union_payload_recursive(CodeGen *g, ZigValue *err_union_const_val); static LLVMValueRef gen_const_ptr_optional_payload_recursive(CodeGen *g, ZigValue *optional_const_val); static LLVMValueRef gen_parent_ptr(CodeGen *g, ZigValue *val, ConstParent *parent) { switch (parent->id) { case ConstParentIdNone: render_const_val(g, val, ""); render_const_val_global(g, val, ""); return val->llvm_global; case ConstParentIdStruct: { ZigValue *struct_val = parent->data.p_struct.struct_val; size_t src_field_index = parent->data.p_struct.field_index; size_t gen_field_index = struct_val->type->data.structure.fields[src_field_index]->gen_index; return gen_const_ptr_struct_recursive(g, struct_val, gen_field_index); } case ConstParentIdErrUnionCode: return gen_const_ptr_err_union_code_recursive(g, parent->data.p_err_union_code.err_union_val); case ConstParentIdErrUnionPayload: return gen_const_ptr_err_union_payload_recursive(g, parent->data.p_err_union_payload.err_union_val); case ConstParentIdOptionalPayload: return gen_const_ptr_optional_payload_recursive(g, parent->data.p_optional_payload.optional_val); case ConstParentIdArray: return gen_const_ptr_array_recursive(g, parent->data.p_array.array_val, parent->data.p_array.elem_index); case ConstParentIdUnion: return gen_const_ptr_union_recursive(g, parent->data.p_union.union_val); case ConstParentIdScalar: render_const_val(g, parent->data.p_scalar.scalar_val, ""); render_const_val_global(g, parent->data.p_scalar.scalar_val, ""); return parent->data.p_scalar.scalar_val->llvm_global; } zig_unreachable(); } static LLVMValueRef gen_const_ptr_array_recursive(CodeGen *g, ZigValue *array_const_val, size_t index) { expand_undef_array(g, array_const_val); ConstParent *parent = &array_const_val->parent; LLVMValueRef base_ptr = gen_parent_ptr(g, array_const_val, parent); LLVMTypeKind el_type = LLVMGetTypeKind(LLVMGetElementType(LLVMTypeOf(base_ptr))); if (el_type == LLVMArrayTypeKind) { ZigType *usize = g->builtin_types.entry_usize; LLVMValueRef indices[] = { LLVMConstNull(usize->llvm_type), LLVMConstInt(usize->llvm_type, index, false), }; return LLVMConstInBoundsGEP(base_ptr, indices, 2); } else if (el_type == LLVMStructTypeKind) { ZigType *u32 = g->builtin_types.entry_u32; LLVMValueRef indices[] = { LLVMConstNull(get_llvm_type(g, u32)), LLVMConstInt(get_llvm_type(g, u32), index, false), }; return LLVMConstInBoundsGEP(base_ptr, indices, 2); } else { return base_ptr; } } static LLVMValueRef gen_const_ptr_struct_recursive(CodeGen *g, ZigValue *struct_const_val, size_t field_index) { ConstParent *parent = &struct_const_val->parent; LLVMValueRef base_ptr = gen_parent_ptr(g, struct_const_val, parent); ZigType *u32 = g->builtin_types.entry_u32; LLVMValueRef indices[] = { LLVMConstNull(get_llvm_type(g, u32)), LLVMConstInt(get_llvm_type(g, u32), field_index, false), }; return LLVMConstInBoundsGEP(base_ptr, indices, 2); } static LLVMValueRef gen_const_ptr_err_union_code_recursive(CodeGen *g, ZigValue *err_union_const_val) { ConstParent *parent = &err_union_const_val->parent; LLVMValueRef base_ptr = gen_parent_ptr(g, err_union_const_val, parent); ZigType *u32 = g->builtin_types.entry_u32; LLVMValueRef indices[] = { LLVMConstNull(get_llvm_type(g, u32)), LLVMConstInt(get_llvm_type(g, u32), err_union_err_index, false), }; return LLVMConstInBoundsGEP(base_ptr, indices, 2); } static LLVMValueRef gen_const_ptr_err_union_payload_recursive(CodeGen *g, ZigValue *err_union_const_val) { ConstParent *parent = &err_union_const_val->parent; LLVMValueRef base_ptr = gen_parent_ptr(g, err_union_const_val, parent); ZigType *u32 = g->builtin_types.entry_u32; LLVMValueRef indices[] = { LLVMConstNull(get_llvm_type(g, u32)), LLVMConstInt(get_llvm_type(g, u32), err_union_payload_index, false), }; return LLVMConstInBoundsGEP(base_ptr, indices, 2); } static LLVMValueRef gen_const_ptr_optional_payload_recursive(CodeGen *g, ZigValue *optional_const_val) { ConstParent *parent = &optional_const_val->parent; LLVMValueRef base_ptr = gen_parent_ptr(g, optional_const_val, parent); ZigType *u32 = g->builtin_types.entry_u32; LLVMValueRef indices[] = { LLVMConstNull(get_llvm_type(g, u32)), LLVMConstInt(get_llvm_type(g, u32), maybe_child_index, false), }; return LLVMConstInBoundsGEP(base_ptr, indices, 2); } static LLVMValueRef gen_const_ptr_union_recursive(CodeGen *g, ZigValue *union_const_val) { ConstParent *parent = &union_const_val->parent; LLVMValueRef base_ptr = gen_parent_ptr(g, union_const_val, parent); // Slot in the structure where the payload is stored, if equal to SIZE_MAX // the union has no tag and a single field and is collapsed into the field // itself size_t union_payload_index = union_const_val->type->data.unionation.gen_union_index; ZigType *u32 = g->builtin_types.entry_u32; LLVMValueRef indices[] = { LLVMConstNull(get_llvm_type(g, u32)), LLVMConstInt(get_llvm_type(g, u32), union_payload_index, false), }; return LLVMConstInBoundsGEP(base_ptr, indices, (union_payload_index != SIZE_MAX) ? 2 : 1); } static LLVMValueRef pack_const_int(CodeGen *g, LLVMTypeRef big_int_type_ref, ZigValue *const_val) { switch (const_val->special) { case ConstValSpecialLazy: case ConstValSpecialRuntime: zig_unreachable(); case ConstValSpecialUndef: return LLVMConstInt(big_int_type_ref, 0, false); case ConstValSpecialStatic: break; } ZigType *type_entry = const_val->type; assert(type_has_bits(g, type_entry)); switch (type_entry->id) { case ZigTypeIdInvalid: case ZigTypeIdMetaType: case ZigTypeIdUnreachable: case ZigTypeIdComptimeFloat: case ZigTypeIdComptimeInt: case ZigTypeIdEnumLiteral: case ZigTypeIdUndefined: case ZigTypeIdNull: case ZigTypeIdErrorUnion: case ZigTypeIdErrorSet: case ZigTypeIdBoundFn: case ZigTypeIdVoid: case ZigTypeIdOpaque: zig_unreachable(); case ZigTypeIdBool: return LLVMConstInt(big_int_type_ref, const_val->data.x_bool ? 1 : 0, false); case ZigTypeIdEnum: { assert(type_entry->data.enumeration.decl_node->data.container_decl.init_arg_expr != nullptr); LLVMValueRef int_val = gen_const_val(g, const_val, ""); return LLVMConstZExt(int_val, big_int_type_ref); } case ZigTypeIdInt: { LLVMValueRef int_val = gen_const_val(g, const_val, ""); return LLVMConstZExt(int_val, big_int_type_ref); } case ZigTypeIdFloat: { LLVMValueRef float_val = gen_const_val(g, const_val, ""); LLVMValueRef int_val = LLVMConstFPToUI(float_val, LLVMIntType((unsigned)type_entry->data.floating.bit_count)); return LLVMConstZExt(int_val, big_int_type_ref); } case ZigTypeIdPointer: case ZigTypeIdFn: case ZigTypeIdOptional: { LLVMValueRef ptr_val = gen_const_val(g, const_val, ""); LLVMValueRef ptr_size_int_val = LLVMConstPtrToInt(ptr_val, g->builtin_types.entry_usize->llvm_type); return LLVMConstZExt(ptr_size_int_val, big_int_type_ref); } case ZigTypeIdArray: { LLVMValueRef val = LLVMConstInt(big_int_type_ref, 0, false); if (const_val->data.x_array.special == ConstArraySpecialUndef) { return val; } expand_undef_array(g, const_val); bool is_big_endian = g->is_big_endian; // TODO get endianness from struct type uint32_t packed_bits_size = type_size_bits(g, type_entry->data.array.child_type); size_t used_bits = 0; for (size_t i = 0; i < type_entry->data.array.len; i += 1) { ZigValue *elem_val = &const_val->data.x_array.data.s_none.elements[i]; LLVMValueRef child_val = pack_const_int(g, big_int_type_ref, elem_val); if (is_big_endian) { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, packed_bits_size, false); val = LLVMConstShl(val, shift_amt); val = LLVMConstOr(val, child_val); } else { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, used_bits, false); LLVMValueRef child_val_shifted = LLVMConstShl(child_val, shift_amt); val = LLVMConstOr(val, child_val_shifted); used_bits += packed_bits_size; } } if (type_entry->data.array.sentinel != nullptr) { ZigValue *elem_val = type_entry->data.array.sentinel; LLVMValueRef child_val = pack_const_int(g, big_int_type_ref, elem_val); if (is_big_endian) { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, packed_bits_size, false); val = LLVMConstShl(val, shift_amt); val = LLVMConstOr(val, child_val); } else { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, used_bits, false); LLVMValueRef child_val_shifted = LLVMConstShl(child_val, shift_amt); val = LLVMConstOr(val, child_val_shifted); used_bits += packed_bits_size; } } return val; } case ZigTypeIdVector: zig_panic("TODO bit pack a vector"); case ZigTypeIdUnion: zig_panic("TODO bit pack a union"); case ZigTypeIdStruct: { assert(type_entry->data.structure.layout == ContainerLayoutPacked); bool is_big_endian = g->is_big_endian; // TODO get endianness from struct type LLVMValueRef val = LLVMConstInt(big_int_type_ref, 0, false); size_t used_bits = 0; for (size_t i = 0; i < type_entry->data.structure.src_field_count; i += 1) { TypeStructField *field = type_entry->data.structure.fields[i]; if (field->gen_index == SIZE_MAX) { continue; } LLVMValueRef child_val = pack_const_int(g, big_int_type_ref, const_val->data.x_struct.fields[i]); uint32_t packed_bits_size = type_size_bits(g, field->type_entry); if (is_big_endian) { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, packed_bits_size, false); val = LLVMConstShl(val, shift_amt); val = LLVMConstOr(val, child_val); } else { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, used_bits, false); LLVMValueRef child_val_shifted = LLVMConstShl(child_val, shift_amt); val = LLVMConstOr(val, child_val_shifted); used_bits += packed_bits_size; } } return val; } case ZigTypeIdFnFrame: zig_panic("TODO bit pack an async function frame"); case ZigTypeIdAnyFrame: zig_panic("TODO bit pack an anyframe"); } zig_unreachable(); } // We have this because union constants can't be represented by the official union type, // and this property bubbles up in whatever aggregate type contains a union constant static bool is_llvm_value_unnamed_type(CodeGen *g, ZigType *type_entry, LLVMValueRef val) { return LLVMTypeOf(val) != get_llvm_type(g, type_entry); } static LLVMValueRef gen_const_val_ptr(CodeGen *g, ZigValue *const_val, const char *name) { switch (const_val->data.x_ptr.special) { case ConstPtrSpecialInvalid: case ConstPtrSpecialDiscard: zig_unreachable(); case ConstPtrSpecialRef: { ZigValue *pointee = const_val->data.x_ptr.data.ref.pointee; render_const_val(g, pointee, ""); render_const_val_global(g, pointee, ""); const_val->llvm_value = LLVMConstBitCast(pointee->llvm_global, get_llvm_type(g, const_val->type)); return const_val->llvm_value; } case ConstPtrSpecialBaseArray: case ConstPtrSpecialSubArray: { ZigValue *array_const_val = const_val->data.x_ptr.data.base_array.array_val; assert(array_const_val->type->id == ZigTypeIdArray); if (!type_has_bits(g, array_const_val->type)) { // make this a null pointer ZigType *usize = g->builtin_types.entry_usize; const_val->llvm_value = LLVMConstIntToPtr(LLVMConstNull(usize->llvm_type), get_llvm_type(g, const_val->type)); return const_val->llvm_value; } size_t elem_index = const_val->data.x_ptr.data.base_array.elem_index; LLVMValueRef uncasted_ptr_val = gen_const_ptr_array_recursive(g, array_const_val, elem_index); LLVMValueRef ptr_val = LLVMConstBitCast(uncasted_ptr_val, get_llvm_type(g, const_val->type)); const_val->llvm_value = ptr_val; return ptr_val; } case ConstPtrSpecialBaseStruct: { ZigValue *struct_const_val = const_val->data.x_ptr.data.base_struct.struct_val; assert(struct_const_val->type->id == ZigTypeIdStruct); if (!type_has_bits(g, struct_const_val->type)) { // make this a null pointer ZigType *usize = g->builtin_types.entry_usize; const_val->llvm_value = LLVMConstIntToPtr(LLVMConstNull(usize->llvm_type), get_llvm_type(g, const_val->type)); return const_val->llvm_value; } size_t src_field_index = const_val->data.x_ptr.data.base_struct.field_index; size_t gen_field_index = struct_const_val->type->data.structure.fields[src_field_index]->gen_index; LLVMValueRef uncasted_ptr_val = gen_const_ptr_struct_recursive(g, struct_const_val, gen_field_index); LLVMValueRef ptr_val = LLVMConstBitCast(uncasted_ptr_val, get_llvm_type(g, const_val->type)); const_val->llvm_value = ptr_val; return ptr_val; } case ConstPtrSpecialBaseErrorUnionCode: { ZigValue *err_union_const_val = const_val->data.x_ptr.data.base_err_union_code.err_union_val; assert(err_union_const_val->type->id == ZigTypeIdErrorUnion); if (!type_has_bits(g, err_union_const_val->type)) { // make this a null pointer ZigType *usize = g->builtin_types.entry_usize; const_val->llvm_value = LLVMConstIntToPtr(LLVMConstNull(usize->llvm_type), get_llvm_type(g, const_val->type)); return const_val->llvm_value; } LLVMValueRef uncasted_ptr_val = gen_const_ptr_err_union_code_recursive(g, err_union_const_val); LLVMValueRef ptr_val = LLVMConstBitCast(uncasted_ptr_val, get_llvm_type(g, const_val->type)); const_val->llvm_value = ptr_val; return ptr_val; } case ConstPtrSpecialBaseErrorUnionPayload: { ZigValue *err_union_const_val = const_val->data.x_ptr.data.base_err_union_payload.err_union_val; assert(err_union_const_val->type->id == ZigTypeIdErrorUnion); if (!type_has_bits(g, err_union_const_val->type)) { // make this a null pointer ZigType *usize = g->builtin_types.entry_usize; const_val->llvm_value = LLVMConstIntToPtr(LLVMConstNull(usize->llvm_type), get_llvm_type(g, const_val->type)); return const_val->llvm_value; } LLVMValueRef uncasted_ptr_val = gen_const_ptr_err_union_payload_recursive(g, err_union_const_val); LLVMValueRef ptr_val = LLVMConstBitCast(uncasted_ptr_val, get_llvm_type(g, const_val->type)); const_val->llvm_value = ptr_val; return ptr_val; } case ConstPtrSpecialBaseOptionalPayload: { ZigValue *optional_const_val = const_val->data.x_ptr.data.base_optional_payload.optional_val; assert(optional_const_val->type->id == ZigTypeIdOptional); if (!type_has_bits(g, optional_const_val->type)) { // make this a null pointer ZigType *usize = g->builtin_types.entry_usize; const_val->llvm_value = LLVMConstIntToPtr(LLVMConstNull(usize->llvm_type), get_llvm_type(g, const_val->type)); return const_val->llvm_value; } LLVMValueRef uncasted_ptr_val = gen_const_ptr_optional_payload_recursive(g, optional_const_val); LLVMValueRef ptr_val = LLVMConstBitCast(uncasted_ptr_val, get_llvm_type(g, const_val->type)); const_val->llvm_value = ptr_val; return ptr_val; } case ConstPtrSpecialHardCodedAddr: { uint64_t addr_value = const_val->data.x_ptr.data.hard_coded_addr.addr; ZigType *usize = g->builtin_types.entry_usize; const_val->llvm_value = LLVMConstIntToPtr( LLVMConstInt(usize->llvm_type, addr_value, false), get_llvm_type(g, const_val->type)); return const_val->llvm_value; } case ConstPtrSpecialFunction: return LLVMConstBitCast(fn_llvm_value(g, const_val->data.x_ptr.data.fn.fn_entry), get_llvm_type(g, const_val->type)); case ConstPtrSpecialNull: return LLVMConstNull(get_llvm_type(g, const_val->type)); } zig_unreachable(); } static LLVMValueRef gen_const_val_err_set(CodeGen *g, ZigValue *const_val, const char *name) { uint64_t value = (const_val->data.x_err_set == nullptr) ? 0 : const_val->data.x_err_set->value; return LLVMConstInt(get_llvm_type(g, g->builtin_types.entry_global_error_set), value, false); } static LLVMValueRef gen_const_val(CodeGen *g, ZigValue *const_val, const char *name) { Error err; ZigType *type_entry = const_val->type; assert(type_has_bits(g, type_entry)); if (const_val->special == ConstValSpecialLazy && (err = ir_resolve_lazy(g, nullptr, const_val))) codegen_report_errors_and_exit(g); switch (const_val->special) { case ConstValSpecialLazy: case ConstValSpecialRuntime: zig_unreachable(); case ConstValSpecialUndef: return LLVMGetUndef(get_llvm_type(g, type_entry)); case ConstValSpecialStatic: break; } if ((err = type_resolve(g, type_entry, ResolveStatusLLVMFull))) zig_unreachable(); switch (type_entry->id) { case ZigTypeIdInt: return bigint_to_llvm_const(get_llvm_type(g, type_entry), &const_val->data.x_bigint); case ZigTypeIdErrorSet: return gen_const_val_err_set(g, const_val, name); case ZigTypeIdFloat: switch (type_entry->data.floating.bit_count) { case 16: return LLVMConstReal(get_llvm_type(g, type_entry), zig_f16_to_double(const_val->data.x_f16)); case 32: return LLVMConstReal(get_llvm_type(g, type_entry), const_val->data.x_f32); case 64: return LLVMConstReal(get_llvm_type(g, type_entry), const_val->data.x_f64); case 128: { // TODO make sure this is correct on big endian targets too uint8_t buf[16]; memcpy(buf, &const_val->data.x_f128, 16); LLVMValueRef as_int = LLVMConstIntOfArbitraryPrecision(LLVMInt128Type(), 2, (uint64_t*)buf); return LLVMConstBitCast(as_int, get_llvm_type(g, type_entry)); } default: zig_unreachable(); } case ZigTypeIdBool: if (const_val->data.x_bool) { return LLVMConstAllOnes(LLVMInt1Type()); } else { return LLVMConstNull(LLVMInt1Type()); } case ZigTypeIdOptional: { ZigType *child_type = type_entry->data.maybe.child_type; if (get_src_ptr_type(type_entry) != nullptr) { bool has_bits; if ((err = type_has_bits2(g, child_type, &has_bits))) codegen_report_errors_and_exit(g); if (has_bits) return gen_const_val_ptr(g, const_val, name); // No bits, treat this value as a boolean const unsigned bool_val = optional_value_is_null(const_val) ? 0 : 1; return LLVMConstInt(LLVMInt1Type(), bool_val, false); } else if (child_type->id == ZigTypeIdErrorSet) { return gen_const_val_err_set(g, const_val, name); } else if (!type_has_bits(g, child_type)) { return LLVMConstInt(LLVMInt1Type(), const_val->data.x_optional ? 1 : 0, false); } else { LLVMValueRef child_val; LLVMValueRef maybe_val; bool make_unnamed_struct; if (const_val->data.x_optional) { child_val = gen_const_val(g, const_val->data.x_optional, ""); maybe_val = LLVMConstAllOnes(LLVMInt1Type()); make_unnamed_struct = is_llvm_value_unnamed_type(g, const_val->type, child_val); } else { child_val = LLVMGetUndef(get_llvm_type(g, child_type)); maybe_val = LLVMConstNull(LLVMInt1Type()); make_unnamed_struct = false; } LLVMValueRef fields[] = { child_val, maybe_val, nullptr, }; if (make_unnamed_struct) { LLVMValueRef result = LLVMConstStruct(fields, 2, false); uint64_t last_field_offset = LLVMOffsetOfElement(g->target_data_ref, LLVMTypeOf(result), 1); uint64_t end_offset = last_field_offset + LLVMStoreSizeOfType(g->target_data_ref, LLVMTypeOf(fields[1])); uint64_t expected_sz = LLVMABISizeOfType(g->target_data_ref, get_llvm_type(g, type_entry)); unsigned pad_sz = expected_sz - end_offset; if (pad_sz != 0) { fields[2] = LLVMGetUndef(LLVMArrayType(LLVMInt8Type(), pad_sz)); result = LLVMConstStruct(fields, 3, false); } uint64_t actual_sz = LLVMStoreSizeOfType(g->target_data_ref, LLVMTypeOf(result)); assert(actual_sz == expected_sz); return result; } else { return LLVMConstNamedStruct(get_llvm_type(g, type_entry), fields, 2); } } } case ZigTypeIdStruct: { LLVMValueRef *fields = heap::c_allocator.allocate(type_entry->data.structure.gen_field_count); size_t src_field_count = type_entry->data.structure.src_field_count; bool make_unnamed_struct = false; assert(type_entry->data.structure.resolve_status == ResolveStatusLLVMFull); if (type_entry->data.structure.layout == ContainerLayoutPacked) { size_t src_field_index = 0; while (src_field_index < src_field_count) { TypeStructField *type_struct_field = type_entry->data.structure.fields[src_field_index]; if (type_struct_field->gen_index == SIZE_MAX) { src_field_index += 1; continue; } size_t src_field_index_end = src_field_index + 1; for (; src_field_index_end < src_field_count; src_field_index_end += 1) { TypeStructField *it_field = type_entry->data.structure.fields[src_field_index_end]; if (it_field->gen_index != type_struct_field->gen_index) break; } if (src_field_index + 1 == src_field_index_end) { ZigValue *field_val = const_val->data.x_struct.fields[src_field_index]; LLVMValueRef val = gen_const_val(g, field_val, ""); fields[type_struct_field->gen_index] = val; make_unnamed_struct = make_unnamed_struct || is_llvm_value_unnamed_type(g, field_val->type, val); } else { bool is_big_endian = g->is_big_endian; // TODO get endianness from struct type LLVMTypeRef field_ty = LLVMStructGetTypeAtIndex(get_llvm_type(g, type_entry), (unsigned)type_struct_field->gen_index); const size_t size_in_bytes = LLVMStoreSizeOfType(g->target_data_ref, field_ty); const size_t size_in_bits = size_in_bytes * 8; LLVMTypeRef big_int_type_ref = LLVMIntType(size_in_bits); LLVMValueRef val = LLVMConstInt(big_int_type_ref, 0, false); size_t used_bits = 0; for (size_t i = src_field_index; i < src_field_index_end; i += 1) { TypeStructField *it_field = type_entry->data.structure.fields[i]; if (it_field->gen_index == SIZE_MAX) { continue; } LLVMValueRef child_val = pack_const_int(g, big_int_type_ref, const_val->data.x_struct.fields[i]); uint32_t packed_bits_size = type_size_bits(g, it_field->type_entry); if (is_big_endian) { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, size_in_bits - used_bits - packed_bits_size, false); LLVMValueRef child_val_shifted = LLVMConstShl(child_val, shift_amt); val = LLVMConstOr(val, child_val_shifted); } else { LLVMValueRef shift_amt = LLVMConstInt(big_int_type_ref, used_bits, false); LLVMValueRef child_val_shifted = LLVMConstShl(child_val, shift_amt); val = LLVMConstOr(val, child_val_shifted); } used_bits += packed_bits_size; } assert(size_in_bits >= used_bits); if (LLVMGetTypeKind(field_ty) != LLVMArrayTypeKind) { assert(LLVMGetTypeKind(field_ty) == LLVMIntegerTypeKind); fields[type_struct_field->gen_index] = val; } else { const LLVMValueRef AMT = LLVMConstInt(LLVMTypeOf(val), 8, false); LLVMValueRef *values = heap::c_allocator.allocate(size_in_bytes); for (size_t i = 0; i < size_in_bytes; i++) { const size_t idx = is_big_endian ? size_in_bytes - 1 - i : i; values[idx] = LLVMConstTruncOrBitCast(val, LLVMInt8Type()); val = LLVMConstLShr(val, AMT); } fields[type_struct_field->gen_index] = LLVMConstArray(LLVMInt8Type(), values, size_in_bytes); } } src_field_index = src_field_index_end; } } else { for (uint32_t i = 0; i < src_field_count; i += 1) { TypeStructField *type_struct_field = type_entry->data.structure.fields[i]; if (type_struct_field->gen_index == SIZE_MAX) { continue; } ZigValue *field_val = const_val->data.x_struct.fields[i]; if (field_val == nullptr) { add_node_error(g, type_struct_field->decl_node, buf_sprintf("compiler bug: generating const value for struct field '%s'", buf_ptr(type_struct_field->name))); codegen_report_errors_and_exit(g); } ZigType *field_type = field_val->type; assert(field_type != nullptr); if ((err = ensure_const_val_repr(nullptr, g, nullptr, field_val, field_type))) { zig_unreachable(); } LLVMValueRef val = gen_const_val(g, field_val, ""); make_unnamed_struct = make_unnamed_struct || is_llvm_value_unnamed_type(g, field_type, val); // Find the next runtime field size_t next_rt_gen_index = type_entry->data.structure.gen_field_count; size_t next_offset = type_entry->abi_size; for (size_t j = i + 1; j < src_field_count; j++) { const size_t index = type_entry->data.structure.fields[j]->gen_index; const size_t offset = type_entry->data.structure.fields[j]->offset; if (index != SIZE_MAX) { next_rt_gen_index = index; next_offset = offset; break; } } // How much padding is needed to reach the next field const size_t pad_bytes = next_offset - (type_struct_field->offset + LLVMABISizeOfType(g->target_data_ref, LLVMTypeOf(val))); // Catch underflow assert((ssize_t)pad_bytes >= 0); if (type_struct_field->gen_index + 1 != next_rt_gen_index) { // If there's a hole between this field and the next // we have an alignment gap to fill fields[type_struct_field->gen_index] = val; fields[type_struct_field->gen_index + 1] = LLVMGetUndef(LLVMArrayType(LLVMInt8Type(), pad_bytes)); } else if (pad_bytes != 0) { LLVMValueRef padded_val[] = { val, LLVMGetUndef(LLVMArrayType(LLVMInt8Type(), pad_bytes)), }; fields[type_struct_field->gen_index] = LLVMConstStruct(padded_val, 2, true); make_unnamed_struct = true; } else { fields[type_struct_field->gen_index] = val; } } } if (make_unnamed_struct) { LLVMValueRef unnamed_struct = LLVMConstStruct(fields, type_entry->data.structure.gen_field_count, type_entry->data.structure.layout == ContainerLayoutPacked); heap::c_allocator.deallocate(fields, type_entry->data.structure.gen_field_count); return unnamed_struct; } else { LLVMValueRef named_struct = LLVMConstNamedStruct(get_llvm_type(g, type_entry), fields, type_entry->data.structure.gen_field_count); heap::c_allocator.deallocate(fields, type_entry->data.structure.gen_field_count); return named_struct; } } case ZigTypeIdArray: { uint64_t len = type_entry->data.array.len; switch (const_val->data.x_array.special) { case ConstArraySpecialUndef: return LLVMGetUndef(get_llvm_type(g, type_entry)); case ConstArraySpecialNone: { uint64_t extra_len_from_sentinel = (type_entry->data.array.sentinel != nullptr) ? 1 : 0; uint64_t full_len = len + extra_len_from_sentinel; LLVMValueRef *values = heap::c_allocator.allocate(full_len); LLVMTypeRef element_type_ref = get_llvm_type(g, type_entry->data.array.child_type); bool make_unnamed_struct = false; for (uint64_t i = 0; i < len; i += 1) { ZigValue *elem_value = &const_val->data.x_array.data.s_none.elements[i]; LLVMValueRef val = gen_const_val(g, elem_value, ""); values[i] = val; make_unnamed_struct = make_unnamed_struct || is_llvm_value_unnamed_type(g, elem_value->type, val); } if (type_entry->data.array.sentinel != nullptr) { values[len] = gen_const_val(g, type_entry->data.array.sentinel, ""); } if (make_unnamed_struct) { LLVMValueRef unnamed_struct = LLVMConstStruct(values, full_len, true); heap::c_allocator.deallocate(values, full_len); return unnamed_struct; } else { LLVMValueRef array = LLVMConstArray(element_type_ref, values, (unsigned)full_len); heap::c_allocator.deallocate(values, full_len); return array; } } case ConstArraySpecialBuf: { Buf *buf = const_val->data.x_array.data.s_buf; return LLVMConstString(buf_ptr(buf), (unsigned)buf_len(buf), type_entry->data.array.sentinel == nullptr); } } zig_unreachable(); } case ZigTypeIdVector: { uint32_t len = type_entry->data.vector.len; switch (const_val->data.x_array.special) { case ConstArraySpecialUndef: return LLVMGetUndef(get_llvm_type(g, type_entry)); case ConstArraySpecialNone: { LLVMValueRef *values = heap::c_allocator.allocate(len); for (uint64_t i = 0; i < len; i += 1) { ZigValue *elem_value = &const_val->data.x_array.data.s_none.elements[i]; values[i] = gen_const_val(g, elem_value, ""); } LLVMValueRef vector = LLVMConstVector(values, len); heap::c_allocator.deallocate(values, len); return vector; } case ConstArraySpecialBuf: { Buf *buf = const_val->data.x_array.data.s_buf; assert(buf_len(buf) == len); LLVMValueRef *values = heap::c_allocator.allocate(len); for (uint64_t i = 0; i < len; i += 1) { values[i] = LLVMConstInt(g->builtin_types.entry_u8->llvm_type, buf_ptr(buf)[i], false); } LLVMValueRef vector = LLVMConstVector(values, len); heap::c_allocator.deallocate(values, len); return vector; } } zig_unreachable(); } case ZigTypeIdUnion: { // Force type_entry->data.unionation.union_llvm_type to get resolved (void)get_llvm_type(g, type_entry); if (type_entry->data.unionation.gen_field_count == 0) { if (type_entry->data.unionation.tag_type == nullptr) { return nullptr; } else { return bigint_to_llvm_const(get_llvm_type(g, type_entry->data.unionation.tag_type), &const_val->data.x_union.tag); } } LLVMTypeRef union_type_ref = type_entry->data.unionation.union_llvm_type; assert(union_type_ref != nullptr); LLVMValueRef union_value_ref; bool make_unnamed_struct; ZigValue *payload_value = const_val->data.x_union.payload; if (payload_value == nullptr || !type_has_bits(g, payload_value->type)) { if (type_entry->data.unionation.gen_tag_index == SIZE_MAX) return LLVMGetUndef(get_llvm_type(g, type_entry)); union_value_ref = LLVMGetUndef(union_type_ref); make_unnamed_struct = false; } else { uint64_t field_type_bytes = LLVMABISizeOfType(g->target_data_ref, get_llvm_type(g, payload_value->type)); uint64_t pad_bytes = type_entry->data.unionation.union_abi_size - field_type_bytes; LLVMValueRef correctly_typed_value = gen_const_val(g, payload_value, ""); make_unnamed_struct = is_llvm_value_unnamed_type(g, payload_value->type, correctly_typed_value) || payload_value->type != type_entry->data.unionation.most_aligned_union_member->type_entry; { if (pad_bytes == 0) { union_value_ref = correctly_typed_value; } else { LLVMValueRef fields[2]; fields[0] = correctly_typed_value; fields[1] = LLVMGetUndef(LLVMArrayType(LLVMInt8Type(), (unsigned)pad_bytes)); if (make_unnamed_struct || type_entry->data.unionation.gen_tag_index != SIZE_MAX) { union_value_ref = LLVMConstStruct(fields, 2, false); } else { union_value_ref = LLVMConstNamedStruct(union_type_ref, fields, 2); } } } if (type_entry->data.unionation.gen_tag_index == SIZE_MAX) { return union_value_ref; } } LLVMValueRef tag_value = bigint_to_llvm_const( get_llvm_type(g, type_entry->data.unionation.tag_type), &const_val->data.x_union.tag); LLVMValueRef fields[3]; fields[type_entry->data.unionation.gen_union_index] = union_value_ref; fields[type_entry->data.unionation.gen_tag_index] = tag_value; if (make_unnamed_struct) { LLVMValueRef result = LLVMConstStruct(fields, 2, false); uint64_t last_field_offset = LLVMOffsetOfElement(g->target_data_ref, LLVMTypeOf(result), 1); uint64_t end_offset = last_field_offset + LLVMStoreSizeOfType(g->target_data_ref, LLVMTypeOf(fields[1])); uint64_t expected_sz = LLVMABISizeOfType(g->target_data_ref, get_llvm_type(g, type_entry)); unsigned pad_sz = expected_sz - end_offset; if (pad_sz != 0) { fields[2] = LLVMGetUndef(LLVMArrayType(LLVMInt8Type(), pad_sz)); result = LLVMConstStruct(fields, 3, false); } uint64_t actual_sz = LLVMStoreSizeOfType(g->target_data_ref, LLVMTypeOf(result)); assert(actual_sz == expected_sz); return result; } else { return LLVMConstNamedStruct(get_llvm_type(g, type_entry), fields, 2); } } case ZigTypeIdEnum: return bigint_to_llvm_const(get_llvm_type(g, type_entry), &const_val->data.x_enum_tag); case ZigTypeIdFn: if (const_val->data.x_ptr.special == ConstPtrSpecialFunction && const_val->data.x_ptr.mut != ConstPtrMutComptimeConst) { zig_unreachable(); } // Treat it the same as we do for pointers return gen_const_val_ptr(g, const_val, name); case ZigTypeIdPointer: return gen_const_val_ptr(g, const_val, name); case ZigTypeIdErrorUnion: { ZigType *payload_type = type_entry->data.error_union.payload_type; ZigType *err_set_type = type_entry->data.error_union.err_set_type; if (!type_has_bits(g, payload_type)) { assert(type_has_bits(g, err_set_type)); ErrorTableEntry *err_set = const_val->data.x_err_union.error_set->data.x_err_set; uint64_t value = (err_set == nullptr) ? 0 : err_set->value; return LLVMConstInt(get_llvm_type(g, g->err_tag_type), value, false); } else if (!type_has_bits(g, err_set_type)) { assert(type_has_bits(g, payload_type)); return gen_const_val(g, const_val->data.x_err_union.payload, ""); } else { LLVMValueRef err_tag_value; LLVMValueRef err_payload_value; bool make_unnamed_struct; ErrorTableEntry *err_set = const_val->data.x_err_union.error_set->data.x_err_set; if (err_set != nullptr) { err_tag_value = LLVMConstInt(get_llvm_type(g, g->err_tag_type), err_set->value, false); err_payload_value = LLVMConstNull(get_llvm_type(g, payload_type)); make_unnamed_struct = false; } else { err_tag_value = LLVMConstNull(get_llvm_type(g, g->err_tag_type)); ZigValue *payload_val = const_val->data.x_err_union.payload; err_payload_value = gen_const_val(g, payload_val, ""); make_unnamed_struct = is_llvm_value_unnamed_type(g, payload_val->type, err_payload_value); } LLVMValueRef fields[3]; fields[err_union_err_index] = err_tag_value; fields[err_union_payload_index] = err_payload_value; size_t field_count = 2; if (type_entry->data.error_union.pad_llvm_type != nullptr) { fields[2] = LLVMGetUndef(type_entry->data.error_union.pad_llvm_type); field_count = 3; } if (make_unnamed_struct) { return LLVMConstStruct(fields, field_count, false); } else { return LLVMConstNamedStruct(get_llvm_type(g, type_entry), fields, field_count); } } } case ZigTypeIdVoid: return nullptr; case ZigTypeIdInvalid: case ZigTypeIdMetaType: case ZigTypeIdUnreachable: case ZigTypeIdComptimeFloat: case ZigTypeIdComptimeInt: case ZigTypeIdEnumLiteral: case ZigTypeIdUndefined: case ZigTypeIdNull: case ZigTypeIdBoundFn: case ZigTypeIdOpaque: zig_unreachable(); case ZigTypeIdFnFrame: zig_panic("TODO"); case ZigTypeIdAnyFrame: zig_panic("TODO"); } zig_unreachable(); } static void render_const_val(CodeGen *g, ZigValue *const_val, const char *name) { if (!const_val->llvm_value) const_val->llvm_value = gen_const_val(g, const_val, name); if (const_val->llvm_global) LLVMSetInitializer(const_val->llvm_global, const_val->llvm_value); } static void render_const_val_global(CodeGen *g, ZigValue *const_val, const char *name) { if (!const_val->llvm_global) { LLVMTypeRef type_ref = const_val->llvm_value ? LLVMTypeOf(const_val->llvm_value) : get_llvm_type(g, const_val->type); LLVMValueRef global_value = LLVMAddGlobal(g->module, type_ref, name); LLVMSetLinkage(global_value, (name == nullptr) ? LLVMPrivateLinkage : LLVMInternalLinkage); LLVMSetGlobalConstant(global_value, true); LLVMSetUnnamedAddr(global_value, true); LLVMSetAlignment(global_value, (const_val->llvm_align == 0) ? get_abi_alignment(g, const_val->type) : const_val->llvm_align); const_val->llvm_global = global_value; } if (const_val->llvm_value) LLVMSetInitializer(const_val->llvm_global, const_val->llvm_value); } static void generate_error_name_table(CodeGen *g) { if (g->err_name_table != nullptr || !g->generate_error_name_table || g->errors_by_index.length == 1) { return; } assert(g->errors_by_index.length > 0); ZigType *u8_ptr_type = get_pointer_to_type_extra(g, g->builtin_types.entry_u8, true, false, PtrLenUnknown, get_abi_alignment(g, g->builtin_types.entry_u8), 0, 0, false); ZigType *str_type = get_slice_type(g, u8_ptr_type); LLVMValueRef *values = heap::c_allocator.allocate(g->errors_by_index.length); values[0] = LLVMGetUndef(get_llvm_type(g, str_type)); for (size_t i = 1; i < g->errors_by_index.length; i += 1) { ErrorTableEntry *err_entry = g->errors_by_index.at(i); Buf *name = &err_entry->name; g->largest_err_name_len = max(g->largest_err_name_len, buf_len(name)); LLVMValueRef str_init = LLVMConstString(buf_ptr(name), (unsigned)buf_len(name), true); LLVMValueRef str_global = LLVMAddGlobal(g->module, LLVMTypeOf(str_init), ""); LLVMSetInitializer(str_global, str_init); LLVMSetLinkage(str_global, LLVMPrivateLinkage); LLVMSetGlobalConstant(str_global, true); LLVMSetUnnamedAddr(str_global, true); LLVMSetAlignment(str_global, LLVMABIAlignmentOfType(g->target_data_ref, LLVMTypeOf(str_init))); LLVMValueRef fields[] = { LLVMConstBitCast(str_global, get_llvm_type(g, u8_ptr_type)), LLVMConstInt(g->builtin_types.entry_usize->llvm_type, buf_len(name), false), }; values[i] = LLVMConstNamedStruct(get_llvm_type(g, str_type), fields, 2); } LLVMValueRef err_name_table_init = LLVMConstArray(get_llvm_type(g, str_type), values, (unsigned)g->errors_by_index.length); heap::c_allocator.deallocate(values, g->errors_by_index.length); g->err_name_table = LLVMAddGlobal(g->module, LLVMTypeOf(err_name_table_init), get_mangled_name(g, buf_ptr(buf_create_from_str("__zig_err_name_table")))); LLVMSetInitializer(g->err_name_table, err_name_table_init); LLVMSetLinkage(g->err_name_table, LLVMPrivateLinkage); LLVMSetGlobalConstant(g->err_name_table, true); LLVMSetUnnamedAddr(g->err_name_table, true); LLVMSetAlignment(g->err_name_table, LLVMABIAlignmentOfType(g->target_data_ref, LLVMTypeOf(err_name_table_init))); } static void build_all_basic_blocks(CodeGen *g, ZigFn *fn) { IrExecutableGen *executable = &fn->analyzed_executable; assert(executable->basic_block_list.length > 0); LLVMValueRef fn_val = fn_llvm_value(g, fn); LLVMBasicBlockRef first_bb = nullptr; if (fn_is_async(fn)) { first_bb = LLVMAppendBasicBlock(fn_val, "AsyncSwitch"); g->cur_preamble_llvm_block = first_bb; } for (size_t block_i = 0; block_i < executable->basic_block_list.length; block_i += 1) { IrBasicBlockGen *bb = executable->basic_block_list.at(block_i); bb->llvm_block = LLVMAppendBasicBlock(fn_val, bb->name_hint); } if (first_bb == nullptr) { first_bb = executable->basic_block_list.at(0)->llvm_block; } LLVMPositionBuilderAtEnd(g->builder, first_bb); } static void gen_global_var(CodeGen *g, ZigVar *var, LLVMValueRef init_val, ZigType *type_entry) { if (g->strip_debug_symbols) { return; } assert(var->gen_is_const); assert(type_entry); ZigType *import = get_scope_import(var->parent_scope); assert(import); bool is_local_to_unit = true; ZigLLVMCreateGlobalVariable(g->dbuilder, get_di_scope(g, var->parent_scope), var->name, var->name, import->data.structure.root_struct->di_file, (unsigned)(var->decl_node->line + 1), get_llvm_di_type(g, type_entry), is_local_to_unit); // TODO ^^ make an actual global variable } static void set_global_tls(CodeGen *g, ZigVar *var, LLVMValueRef global_value) { bool is_extern = var->decl_node->data.variable_declaration.is_extern; bool is_export = var->decl_node->data.variable_declaration.is_export; bool is_internal_linkage = !is_extern && !is_export; if (var->is_thread_local && (!g->is_single_threaded || !is_internal_linkage)) { LLVMSetThreadLocalMode(global_value, LLVMGeneralDynamicTLSModel); } } static void do_code_gen(CodeGen *g) { Error err; assert(!g->errors.length); generate_error_name_table(g); // Generate module level variables for (size_t i = 0; i < g->global_vars.length; i += 1) { TldVar *tld_var = g->global_vars.at(i); ZigVar *var = tld_var->var; if (var->var_type->id == ZigTypeIdComptimeFloat) { // Generate debug info for it but that's it. ZigValue *const_val = var->const_value; assert(const_val->special != ConstValSpecialRuntime); if ((err = ir_resolve_lazy(g, var->decl_node, const_val))) zig_unreachable(); if (const_val->type != var->var_type) { zig_panic("TODO debug info for var with ptr casted value"); } ZigType *var_type = g->builtin_types.entry_f128; ZigValue coerced_value = {}; coerced_value.special = ConstValSpecialStatic; coerced_value.type = var_type; coerced_value.data.x_f128 = bigfloat_to_f128(&const_val->data.x_bigfloat); LLVMValueRef init_val = gen_const_val(g, &coerced_value, ""); gen_global_var(g, var, init_val, var_type); continue; } if (var->var_type->id == ZigTypeIdComptimeInt) { // Generate debug info for it but that's it. ZigValue *const_val = var->const_value; assert(const_val->special != ConstValSpecialRuntime); if ((err = ir_resolve_lazy(g, var->decl_node, const_val))) zig_unreachable(); if (const_val->type != var->var_type) { zig_panic("TODO debug info for var with ptr casted value"); } size_t bits_needed = bigint_bits_needed(&const_val->data.x_bigint); if (bits_needed < 8) { bits_needed = 8; } ZigType *var_type = get_int_type(g, const_val->data.x_bigint.is_negative, bits_needed); LLVMValueRef init_val = bigint_to_llvm_const(get_llvm_type(g, var_type), &const_val->data.x_bigint); gen_global_var(g, var, init_val, var_type); continue; } if (!type_has_bits(g, var->var_type)) continue; assert(var->decl_node); GlobalLinkageId linkage; const char *unmangled_name = var->name; const char *symbol_name; if (var->export_list.length == 0) { if (var->decl_node->data.variable_declaration.is_extern) { symbol_name = unmangled_name; linkage = GlobalLinkageIdStrong; } else { symbol_name = get_mangled_name(g, unmangled_name); linkage = GlobalLinkageIdInternal; } } else { GlobalExport *global_export = &var->export_list.items[0]; symbol_name = buf_ptr(&global_export->name); linkage = global_export->linkage; } LLVMValueRef global_value; bool externally_initialized = var->decl_node->data.variable_declaration.expr == nullptr; if (externally_initialized) { LLVMValueRef existing_llvm_var = LLVMGetNamedGlobal(g->module, symbol_name); if (existing_llvm_var) { global_value = LLVMConstBitCast(existing_llvm_var, LLVMPointerType(get_llvm_type(g, var->var_type), 0)); } else { global_value = LLVMAddGlobal(g->module, get_llvm_type(g, var->var_type), symbol_name); // TODO debug info for the extern variable LLVMSetLinkage(global_value, to_llvm_linkage(linkage, true)); maybe_import_dll(g, global_value, GlobalLinkageIdStrong); LLVMSetAlignment(global_value, var->align_bytes); LLVMSetGlobalConstant(global_value, var->gen_is_const); set_global_tls(g, var, global_value); } } else { bool exported = (linkage != GlobalLinkageIdInternal); render_const_val(g, var->const_value, symbol_name); render_const_val_global(g, var->const_value, symbol_name); global_value = var->const_value->llvm_global; if (exported) { LLVMSetLinkage(global_value, to_llvm_linkage(linkage, false)); maybe_export_dll(g, global_value, GlobalLinkageIdStrong); } if (var->section_name) { LLVMSetSection(global_value, buf_ptr(var->section_name)); } LLVMSetAlignment(global_value, var->align_bytes); // TODO debug info for function pointers // Here we use const_value->type because that's the type of the llvm global, // which we const ptr cast upon use to whatever it needs to be. if (var->gen_is_const && var->const_value->type->id != ZigTypeIdFn) { gen_global_var(g, var, var->const_value->llvm_value, var->const_value->type); } LLVMSetGlobalConstant(global_value, var->gen_is_const); set_global_tls(g, var, global_value); } var->value_ref = global_value; for (size_t export_i = 1; export_i < var->export_list.length; export_i += 1) { GlobalExport *global_export = &var->export_list.items[export_i]; LLVMAddAlias(g->module, LLVMTypeOf(var->value_ref), var->value_ref, buf_ptr(&global_export->name)); } } // Generate function definitions. stage2_progress_update_node(g->sub_progress_node, 0, g->fn_defs.length); for (size_t fn_i = 0; fn_i < g->fn_defs.length; fn_i += 1) { ZigFn *fn_table_entry = g->fn_defs.at(fn_i); Stage2ProgressNode *fn_prog_node = stage2_progress_start(g->sub_progress_node, buf_ptr(&fn_table_entry->symbol_name), buf_len(&fn_table_entry->symbol_name), 0); FnTypeId *fn_type_id = &fn_table_entry->type_entry->data.fn.fn_type_id; CallingConvention cc = fn_type_id->cc; bool is_c_abi = !calling_convention_allows_zig_types(cc); bool want_sret = want_first_arg_sret(g, fn_type_id); LLVMValueRef fn = fn_llvm_value(g, fn_table_entry); g->cur_fn = fn_table_entry; g->cur_fn_val = fn; build_all_basic_blocks(g, fn_table_entry); clear_debug_source_node(g); bool is_async = fn_is_async(fn_table_entry); if (is_async) { g->cur_frame_ptr = LLVMGetParam(fn, 0); } else { if (want_sret) { g->cur_ret_ptr = LLVMGetParam(fn, 0); } else if (type_has_bits(g, fn_type_id->return_type)) { g->cur_ret_ptr = build_alloca(g, fn_type_id->return_type, "result", 0); // TODO add debug info variable for this } else { g->cur_ret_ptr = nullptr; } } uint32_t err_ret_trace_arg_index = get_err_ret_trace_arg_index(g, fn_table_entry); bool have_err_ret_trace_arg = err_ret_trace_arg_index != UINT32_MAX; if (have_err_ret_trace_arg) { g->cur_err_ret_trace_val_arg = LLVMGetParam(fn, err_ret_trace_arg_index); } else { g->cur_err_ret_trace_val_arg = nullptr; } // error return tracing setup bool have_err_ret_trace_stack = g->have_err_ret_tracing && fn_table_entry->calls_or_awaits_errorable_fn && !is_async && !have_err_ret_trace_arg; LLVMValueRef err_ret_array_val = nullptr; if (have_err_ret_trace_stack) { ZigType *array_type = get_array_type(g, g->builtin_types.entry_usize, stack_trace_ptr_count, nullptr); err_ret_array_val = build_alloca(g, array_type, "error_return_trace_addresses", get_abi_alignment(g, array_type)); (void)get_llvm_type(g, get_stack_trace_type(g)); g->cur_err_ret_trace_val_stack = build_alloca(g, get_stack_trace_type(g), "error_return_trace", get_abi_alignment(g, g->stack_trace_type)); } else { g->cur_err_ret_trace_val_stack = nullptr; } if (!is_async) { // allocate async frames for nosuspend calls & awaits to async functions ZigType *largest_call_frame_type = nullptr; IrInstGen *all_calls_alloca = ir_create_alloca(g, &fn_table_entry->fndef_scope->base, fn_table_entry->body_node, fn_table_entry, g->builtin_types.entry_void, "@async_call_frame"); for (size_t i = 0; i < fn_table_entry->call_list.length; i += 1) { IrInstGenCall *call = fn_table_entry->call_list.at(i); if (call->fn_entry == nullptr) continue; if (!fn_is_async(call->fn_entry)) continue; if (call->modifier != CallModifierNoSuspend) continue; if (call->frame_result_loc != nullptr) continue; ZigType *callee_frame_type = get_fn_frame_type(g, call->fn_entry); if (largest_call_frame_type == nullptr || callee_frame_type->abi_size > largest_call_frame_type->abi_size) { largest_call_frame_type = callee_frame_type; } call->frame_result_loc = all_calls_alloca; } if (largest_call_frame_type != nullptr) { all_calls_alloca->value->type = get_pointer_to_type(g, largest_call_frame_type, false); } // allocate temporary stack data for (size_t alloca_i = 0; alloca_i < fn_table_entry->alloca_gen_list.length; alloca_i += 1) { IrInstGenAlloca *instruction = fn_table_entry->alloca_gen_list.at(alloca_i); ZigType *ptr_type = instruction->base.value->type; assert(ptr_type->id == ZigTypeIdPointer); ZigType *child_type = ptr_type->data.pointer.child_type; if (type_resolve(g, child_type, ResolveStatusSizeKnown)) zig_unreachable(); if (!type_has_bits(g, child_type)) continue; if (instruction->base.base.ref_count == 0) continue; if (instruction->base.value->special != ConstValSpecialRuntime) { if (const_ptr_pointee(nullptr, g, instruction->base.value, nullptr)->special != ConstValSpecialRuntime) { continue; } } if (type_resolve(g, child_type, ResolveStatusLLVMFull)) zig_unreachable(); instruction->base.llvm_value = build_alloca(g, child_type, instruction->name_hint, get_ptr_align(g, ptr_type)); } } ZigType *import = get_scope_import(&fn_table_entry->fndef_scope->base); unsigned gen_i_init = want_sret ? 1 : 0; // create debug variable declarations for variables and allocate all local variables FnWalk fn_walk_var = {}; fn_walk_var.id = FnWalkIdVars; fn_walk_var.data.vars.import = import; fn_walk_var.data.vars.fn = fn_table_entry; fn_walk_var.data.vars.llvm_fn = fn; fn_walk_var.data.vars.gen_i = gen_i_init; for (size_t var_i = 0; var_i < fn_table_entry->variable_list.length; var_i += 1) { ZigVar *var = fn_table_entry->variable_list.at(var_i); if (!type_has_bits(g, var->var_type)) { continue; } if (ir_get_var_is_comptime(var)) continue; switch (type_requires_comptime(g, var->var_type)) { case ReqCompTimeInvalid: zig_unreachable(); case ReqCompTimeYes: continue; case ReqCompTimeNo: break; } if (var->src_arg_index == SIZE_MAX) { var->di_loc_var = ZigLLVMCreateAutoVariable(g->dbuilder, get_di_scope(g, var->parent_scope), var->name, import->data.structure.root_struct->di_file, (unsigned)(var->decl_node->line + 1), get_llvm_di_type(g, var->var_type), !g->strip_debug_symbols, 0); } else if (is_c_abi) { fn_walk_var.data.vars.var = var; iter_function_params_c_abi(g, fn_table_entry->type_entry, &fn_walk_var, var->src_arg_index); } else if (!is_async) { ZigType *gen_type; FnGenParamInfo *gen_info = &fn_table_entry->type_entry->data.fn.gen_param_info[var->src_arg_index]; assert(gen_info->gen_index != SIZE_MAX); if (handle_is_ptr(g, var->var_type)) { if (gen_info->is_byval) { gen_type = var->var_type; } else { gen_type = gen_info->type; } var->value_ref = LLVMGetParam(fn, gen_info->gen_index); } else { gen_type = var->var_type; var->value_ref = build_alloca(g, var->var_type, var->name, var->align_bytes); } if (var->decl_node) { var->di_loc_var = ZigLLVMCreateParameterVariable(g->dbuilder, get_di_scope(g, var->parent_scope), var->name, import->data.structure.root_struct->di_file, (unsigned)(var->decl_node->line + 1), get_llvm_di_type(g, gen_type), !g->strip_debug_symbols, 0, (unsigned)(gen_info->gen_index+1)); } } } // finishing error return trace setup. we have to do this after all the allocas. if (have_err_ret_trace_stack) { ZigType *usize = g->builtin_types.entry_usize; size_t index_field_index = g->stack_trace_type->data.structure.fields[0]->gen_index; LLVMValueRef index_field_ptr = LLVMBuildStructGEP(g->builder, g->cur_err_ret_trace_val_stack, (unsigned)index_field_index, ""); gen_store_untyped(g, LLVMConstNull(usize->llvm_type), index_field_ptr, 0, false); size_t addresses_field_index = g->stack_trace_type->data.structure.fields[1]->gen_index; LLVMValueRef addresses_field_ptr = LLVMBuildStructGEP(g->builder, g->cur_err_ret_trace_val_stack, (unsigned)addresses_field_index, ""); ZigType *slice_type = g->stack_trace_type->data.structure.fields[1]->type_entry; size_t ptr_field_index = slice_type->data.structure.fields[slice_ptr_index]->gen_index; LLVMValueRef ptr_field_ptr = LLVMBuildStructGEP(g->builder, addresses_field_ptr, (unsigned)ptr_field_index, ""); LLVMValueRef zero = LLVMConstNull(usize->llvm_type); LLVMValueRef indices[] = {zero, zero}; LLVMValueRef err_ret_array_val_elem0_ptr = LLVMBuildInBoundsGEP(g->builder, err_ret_array_val, indices, 2, ""); ZigType *ptr_ptr_usize_type = get_pointer_to_type(g, get_pointer_to_type(g, usize, false), false); gen_store(g, err_ret_array_val_elem0_ptr, ptr_field_ptr, ptr_ptr_usize_type); size_t len_field_index = slice_type->data.structure.fields[slice_len_index]->gen_index; LLVMValueRef len_field_ptr = LLVMBuildStructGEP(g->builder, addresses_field_ptr, (unsigned)len_field_index, ""); gen_store(g, LLVMConstInt(usize->llvm_type, stack_trace_ptr_count, false), len_field_ptr, get_pointer_to_type(g, usize, false)); } if (is_async) { (void)get_llvm_type(g, fn_table_entry->frame_type); g->cur_resume_block_count = 0; LLVMTypeRef usize_type_ref = g->builtin_types.entry_usize->llvm_type; LLVMValueRef size_val = LLVMConstInt(usize_type_ref, fn_table_entry->frame_type->abi_size, false); if (g->need_frame_size_prefix_data) { ZigLLVMFunctionSetPrefixData(fn_table_entry->llvm_value, size_val); } if (!g->strip_debug_symbols) { AstNode *source_node = fn_table_entry->proto_node; ZigLLVMSetCurrentDebugLocation(g->builder, (int)source_node->line + 1, (int)source_node->column + 1, get_di_scope(g, fn_table_entry->child_scope)); } IrExecutableGen *executable = &fn_table_entry->analyzed_executable; LLVMBasicBlockRef bad_resume_block = LLVMAppendBasicBlock(g->cur_fn_val, "BadResume"); LLVMPositionBuilderAtEnd(g->builder, bad_resume_block); gen_assertion_scope(g, PanicMsgIdBadResume, fn_table_entry->child_scope); LLVMPositionBuilderAtEnd(g->builder, g->cur_preamble_llvm_block); render_async_spills(g); g->cur_async_awaiter_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, frame_awaiter_index, ""); LLVMValueRef resume_index_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, frame_resume_index, ""); g->cur_async_resume_index_ptr = resume_index_ptr; if (type_has_bits(g, fn_type_id->return_type)) { LLVMValueRef cur_ret_ptr_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, frame_ret_start, ""); g->cur_ret_ptr = LLVMBuildLoad(g->builder, cur_ret_ptr_ptr, ""); } uint32_t trace_field_index_stack = UINT32_MAX; if (codegen_fn_has_err_ret_tracing_stack(g, fn_table_entry, true)) { trace_field_index_stack = frame_index_trace_stack(g, fn_table_entry); g->cur_err_ret_trace_val_stack = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, trace_field_index_stack, ""); } LLVMValueRef resume_index = LLVMBuildLoad(g->builder, resume_index_ptr, ""); LLVMValueRef switch_instr = LLVMBuildSwitch(g->builder, resume_index, bad_resume_block, 4); g->cur_async_switch_instr = switch_instr; LLVMValueRef zero = LLVMConstNull(usize_type_ref); IrBasicBlockGen *entry_block = executable->basic_block_list.at(0); LLVMAddCase(switch_instr, zero, entry_block->llvm_block); g->cur_resume_block_count += 1; { LLVMBasicBlockRef bad_not_suspended_bb = LLVMAppendBasicBlock(g->cur_fn_val, "NotSuspended"); size_t new_block_index = g->cur_resume_block_count; g->cur_resume_block_count += 1; g->cur_bad_not_suspended_index = LLVMConstInt(usize_type_ref, new_block_index, false); LLVMAddCase(g->cur_async_switch_instr, g->cur_bad_not_suspended_index, bad_not_suspended_bb); LLVMPositionBuilderAtEnd(g->builder, bad_not_suspended_bb); gen_assertion_scope(g, PanicMsgIdResumeNotSuspendedFn, fn_table_entry->child_scope); } LLVMPositionBuilderAtEnd(g->builder, entry_block->llvm_block); LLVMBuildStore(g->builder, g->cur_bad_not_suspended_index, g->cur_async_resume_index_ptr); if (trace_field_index_stack != UINT32_MAX) { if (codegen_fn_has_err_ret_tracing_arg(g, fn_type_id->return_type)) { LLVMValueRef trace_ptr_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, frame_index_trace_arg(g, fn_type_id->return_type), ""); LLVMValueRef zero_ptr = LLVMConstNull(LLVMGetElementType(LLVMTypeOf(trace_ptr_ptr))); LLVMBuildStore(g->builder, zero_ptr, trace_ptr_ptr); } LLVMValueRef trace_field_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, trace_field_index_stack, ""); LLVMValueRef addrs_field_ptr = LLVMBuildStructGEP(g->builder, g->cur_frame_ptr, trace_field_index_stack + 1, ""); gen_init_stack_trace(g, trace_field_ptr, addrs_field_ptr); } render_async_var_decls(g, entry_block->instruction_list.at(0)->base.scope); } else { // create debug variable declarations for parameters // rely on the first variables in the variable_list being parameters. FnWalk fn_walk_init = {}; fn_walk_init.id = FnWalkIdInits; fn_walk_init.data.inits.fn = fn_table_entry; fn_walk_init.data.inits.llvm_fn = fn; fn_walk_init.data.inits.gen_i = gen_i_init; walk_function_params(g, fn_table_entry->type_entry, &fn_walk_init); } ir_render(g, fn_table_entry); stage2_progress_end(fn_prog_node); } assert(!g->errors.length); if (buf_len(&g->global_asm) != 0) { LLVMSetModuleInlineAsm(g->module, buf_ptr(&g->global_asm)); } while (g->type_resolve_stack.length != 0) { ZigType *ty = g->type_resolve_stack.last(); if (type_resolve(g, ty, ResolveStatusLLVMFull)) zig_unreachable(); } ZigLLVMDIBuilderFinalize(g->dbuilder); if (g->verbose_llvm_ir) { fflush(stderr); LLVMDumpModule(g->module); } char *error = nullptr; if (LLVMVerifyModule(g->module, LLVMReturnStatusAction, &error)) { zig_panic("broken LLVM module found: %s\nThis is a bug in the Zig compiler.", error); } } static void zig_llvm_emit_output(CodeGen *g) { g->pass1_arena->destruct(&heap::c_allocator); g->pass1_arena = nullptr; bool is_small = g->build_mode == BuildModeSmallRelease; char *err_msg = nullptr; const char *asm_filename = nullptr; const char *bin_filename = nullptr; const char *llvm_ir_filename = nullptr; if (buf_len(&g->o_file_output_path) != 0) bin_filename = buf_ptr(&g->o_file_output_path); if (buf_len(&g->asm_file_output_path) != 0) asm_filename = buf_ptr(&g->asm_file_output_path); if (buf_len(&g->llvm_ir_file_output_path) != 0) llvm_ir_filename = buf_ptr(&g->llvm_ir_file_output_path); // Unfortunately, LLVM shits the bed when we ask for both binary and assembly. So we call the entire // pipeline multiple times if this is requested. if (asm_filename != nullptr && bin_filename != nullptr) { if (ZigLLVMTargetMachineEmitToFile(g->target_machine, g->module, &err_msg, g->build_mode == BuildModeDebug, is_small, g->enable_time_report, nullptr, bin_filename, llvm_ir_filename)) { fprintf(stderr, "LLVM failed to emit file: %s\n", err_msg); exit(1); } bin_filename = nullptr; llvm_ir_filename = nullptr; } if (ZigLLVMTargetMachineEmitToFile(g->target_machine, g->module, &err_msg, g->build_mode == BuildModeDebug, is_small, g->enable_time_report, asm_filename, bin_filename, llvm_ir_filename)) { fprintf(stderr, "LLVM failed to emit file: %s\n", err_msg); exit(1); } LLVMDisposeModule(g->module); g->module = nullptr; LLVMDisposeTargetData(g->target_data_ref); g->target_data_ref = nullptr; LLVMDisposeTargetMachine(g->target_machine); g->target_machine = nullptr; } struct CIntTypeInfo { CIntType id; const char *name; bool is_signed; }; static const CIntTypeInfo c_int_type_infos[] = { {CIntTypeShort, "c_short", true}, {CIntTypeUShort, "c_ushort", false}, {CIntTypeInt, "c_int", true}, {CIntTypeUInt, "c_uint", false}, {CIntTypeLong, "c_long", true}, {CIntTypeULong, "c_ulong", false}, {CIntTypeLongLong, "c_longlong", true}, {CIntTypeULongLong, "c_ulonglong", false}, }; static const bool is_signed_list[] = { false, true, }; struct GlobalLinkageValue { GlobalLinkageId id; const char *name; }; static void add_fp_entry(CodeGen *g, const char *name, uint32_t bit_count, LLVMTypeRef type_ref, ZigType **field) { ZigType *entry = new_type_table_entry(ZigTypeIdFloat); entry->llvm_type = type_ref; entry->size_in_bits = 8*LLVMStoreSizeOfType(g->target_data_ref, entry->llvm_type); entry->abi_size = LLVMABISizeOfType(g->target_data_ref, entry->llvm_type); entry->abi_align = LLVMABIAlignmentOfType(g->target_data_ref, entry->llvm_type); buf_init_from_str(&entry->name, name); entry->data.floating.bit_count = bit_count; entry->llvm_di_type = ZigLLVMCreateDebugBasicType(g->dbuilder, buf_ptr(&entry->name), entry->size_in_bits, ZigLLVMEncoding_DW_ATE_float()); *field = entry; g->primitive_type_table.put(&entry->name, entry); } static void define_builtin_types(CodeGen *g) { { // if this type is anywhere in the AST, we should never hit codegen. ZigType *entry = new_type_table_entry(ZigTypeIdInvalid); buf_init_from_str(&entry->name, "(invalid)"); g->builtin_types.entry_invalid = entry; } { ZigType *entry = new_type_table_entry(ZigTypeIdComptimeFloat); buf_init_from_str(&entry->name, "comptime_float"); g->builtin_types.entry_num_lit_float = entry; g->primitive_type_table.put(&entry->name, entry); } { ZigType *entry = new_type_table_entry(ZigTypeIdComptimeInt); buf_init_from_str(&entry->name, "comptime_int"); g->builtin_types.entry_num_lit_int = entry; g->primitive_type_table.put(&entry->name, entry); } { ZigType *entry = new_type_table_entry(ZigTypeIdEnumLiteral); buf_init_from_str(&entry->name, "(enum literal)"); g->builtin_types.entry_enum_literal = entry; } { ZigType *entry = new_type_table_entry(ZigTypeIdUndefined); buf_init_from_str(&entry->name, "(undefined)"); g->builtin_types.entry_undef = entry; } { ZigType *entry = new_type_table_entry(ZigTypeIdNull); buf_init_from_str(&entry->name, "(null)"); g->builtin_types.entry_null = entry; } { ZigType *entry = new_type_table_entry(ZigTypeIdOpaque); buf_init_from_str(&entry->name, "(anytype)"); g->builtin_types.entry_anytype = entry; } for (size_t i = 0; i < array_length(c_int_type_infos); i += 1) { const CIntTypeInfo *info = &c_int_type_infos[i]; uint32_t size_in_bits = target_c_type_size_in_bits(g->zig_target, info->id); bool is_signed = info->is_signed; ZigType *entry = new_type_table_entry(ZigTypeIdInt); entry->llvm_type = LLVMIntType(size_in_bits); entry->size_in_bits = size_in_bits; entry->abi_size = LLVMABISizeOfType(g->target_data_ref, entry->llvm_type); entry->abi_align = LLVMABIAlignmentOfType(g->target_data_ref, entry->llvm_type); buf_init_from_str(&entry->name, info->name); entry->llvm_di_type = ZigLLVMCreateDebugBasicType(g->dbuilder, buf_ptr(&entry->name), 8*LLVMStoreSizeOfType(g->target_data_ref, entry->llvm_type), is_signed ? ZigLLVMEncoding_DW_ATE_signed() : ZigLLVMEncoding_DW_ATE_unsigned()); entry->data.integral.is_signed = is_signed; entry->data.integral.bit_count = size_in_bits; g->primitive_type_table.put(&entry->name, entry); get_c_int_type_ptr(g, info->id)[0] = entry; } { ZigType *entry = new_type_table_entry(ZigTypeIdBool); entry->llvm_type = LLVMInt1Type(); entry->size_in_bits = 1; entry->abi_size = LLVMABISizeOfType(g->target_data_ref, entry->llvm_type); entry->abi_align = LLVMABIAlignmentOfType(g->target_data_ref, entry->llvm_type); buf_init_from_str(&entry->name, "bool"); entry->llvm_di_type = ZigLLVMCreateDebugBasicType(g->dbuilder, buf_ptr(&entry->name), 8*LLVMStoreSizeOfType(g->target_data_ref, entry->llvm_type), ZigLLVMEncoding_DW_ATE_boolean()); g->builtin_types.entry_bool = entry; g->primitive_type_table.put(&entry->name, entry); } for (size_t sign_i = 0; sign_i < array_length(is_signed_list); sign_i += 1) { bool is_signed = is_signed_list[sign_i]; ZigType *entry = new_type_table_entry(ZigTypeIdInt); entry->llvm_type = LLVMIntType(g->pointer_size_bytes * 8); entry->size_in_bits = g->pointer_size_bytes * 8; entry->abi_size = LLVMABISizeOfType(g->target_data_ref, entry->llvm_type); entry->abi_align = LLVMABIAlignmentOfType(g->target_data_ref, entry->llvm_type); const char u_or_i = is_signed ? 'i' : 'u'; buf_resize(&entry->name, 0); buf_appendf(&entry->name, "%csize", u_or_i); entry->data.integral.is_signed = is_signed; entry->data.integral.bit_count = g->pointer_size_bytes * 8; entry->llvm_di_type = ZigLLVMCreateDebugBasicType(g->dbuilder, buf_ptr(&entry->name), 8*LLVMStoreSizeOfType(g->target_data_ref, entry->llvm_type), is_signed ? ZigLLVMEncoding_DW_ATE_signed() : ZigLLVMEncoding_DW_ATE_unsigned()); g->primitive_type_table.put(&entry->name, entry); if (is_signed) { g->builtin_types.entry_isize = entry; } else { g->builtin_types.entry_usize = entry; } } add_fp_entry(g, "f16", 16, LLVMHalfType(), &g->builtin_types.entry_f16); add_fp_entry(g, "f32", 32, LLVMFloatType(), &g->builtin_types.entry_f32); add_fp_entry(g, "f64", 64, LLVMDoubleType(), &g->builtin_types.entry_f64); add_fp_entry(g, "f128", 128, LLVMFP128Type(), &g->builtin_types.entry_f128); switch (g->zig_target->arch) { case ZigLLVM_x86: case ZigLLVM_x86_64: if (g->zig_target->abi != ZigLLVM_MSVC) add_fp_entry(g, "c_longdouble", 80, LLVMX86FP80Type(), &g->builtin_types.entry_c_longdouble); else add_fp_entry(g, "c_longdouble", 64, LLVMDoubleType(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_arm: case ZigLLVM_armeb: case ZigLLVM_thumb: case ZigLLVM_thumbeb: add_fp_entry(g, "c_longdouble", 64, LLVMDoubleType(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_aarch64: case ZigLLVM_aarch64_be: if (g->zig_target->os == OsWindows || target_os_is_darwin(g->zig_target->os)) add_fp_entry(g, "c_longdouble", 64, LLVMDoubleType(), &g->builtin_types.entry_c_longdouble); else add_fp_entry(g, "c_longdouble", 128, LLVMFP128Type(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_riscv32: case ZigLLVM_riscv64: add_fp_entry(g, "c_longdouble", 128, LLVMFP128Type(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_wasm32: case ZigLLVM_wasm64: add_fp_entry(g, "c_longdouble", 128, LLVMFP128Type(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_mips: case ZigLLVM_mipsel: // Assume o32 ABI add_fp_entry(g, "c_longdouble", 64, LLVMDoubleType(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_mips64: case ZigLLVM_mips64el: add_fp_entry(g, "c_longdouble", 128, LLVMFP128Type(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_ppc: case ZigLLVM_ppc64: case ZigLLVM_ppc64le: add_fp_entry(g, "c_longdouble", 128, LLVMFP128Type(), &g->builtin_types.entry_c_longdouble); break; case ZigLLVM_avr: // It's either a float or a double, depending on a toolchain switch add_fp_entry(g, "c_longdouble", 64, LLVMDoubleType(), &g->builtin_types.entry_c_longdouble); break; default: zig_panic("TODO implement mapping for c_longdouble"); } { ZigType *entry = new_type_table_entry(ZigTypeIdVoid); entry->llvm_type = LLVMVoidType(); buf_init_from_str(&entry->name, "void"); entry->llvm_di_type = ZigLLVMCreateDebugBasicType(g->dbuilder, buf_ptr(&entry->name), 0, ZigLLVMEncoding_DW_ATE_signed()); g->builtin_types.entry_void = entry; g->primitive_type_table.put(&entry->name, entry); } { ZigType *entry = new_type_table_entry(ZigTypeIdUnreachable); entry->llvm_type = LLVMVoidType(); buf_init_from_str(&entry->name, "noreturn"); entry->llvm_di_type = g->builtin_types.entry_void->llvm_di_type; g->builtin_types.entry_unreachable = entry; g->primitive_type_table.put(&entry->name, entry); } { ZigType *entry = new_type_table_entry(ZigTypeIdMetaType); buf_init_from_str(&entry->name, "type"); g->builtin_types.entry_type = entry; g->primitive_type_table.put(&entry->name, entry); } g->builtin_types.entry_u8 = get_int_type(g, false, 8); g->builtin_types.entry_u16 = get_int_type(g, false, 16); g->builtin_types.entry_u29 = get_int_type(g, false, 29); g->builtin_types.entry_u32 = get_int_type(g, false, 32); g->builtin_types.entry_u64 = get_int_type(g, false, 64); g->builtin_types.entry_i8 = get_int_type(g, true, 8); g->builtin_types.entry_i32 = get_int_type(g, true, 32); g->builtin_types.entry_i64 = get_int_type(g, true, 64); { g->builtin_types.entry_c_void = get_opaque_type(g, nullptr, nullptr, "c_void", buf_create_from_str("c_void")); g->primitive_type_table.put(&g->builtin_types.entry_c_void->name, g->builtin_types.entry_c_void); } { ZigType *entry = new_type_table_entry(ZigTypeIdErrorSet); buf_init_from_str(&entry->name, "anyerror"); entry->data.error_set.err_count = UINT32_MAX; // TODO https://github.com/ziglang/zig/issues/786 g->err_tag_type = g->builtin_types.entry_u16; entry->size_in_bits = g->err_tag_type->size_in_bits; entry->abi_align = g->err_tag_type->abi_align; entry->abi_size = g->err_tag_type->abi_size; g->builtin_types.entry_global_error_set = entry; g->errors_by_index.append(nullptr); g->primitive_type_table.put(&entry->name, entry); } } static void define_intern_values(CodeGen *g) { { auto& value = g->intern.x_undefined; value.type = g->builtin_types.entry_undef; value.special = ConstValSpecialStatic; } { auto& value = g->intern.x_void; value.type = g->builtin_types.entry_void; value.special = ConstValSpecialStatic; } { auto& value = g->intern.x_null; value.type = g->builtin_types.entry_null; value.special = ConstValSpecialStatic; } { auto& value = g->intern.x_unreachable; value.type = g->builtin_types.entry_unreachable; value.special = ConstValSpecialStatic; } { auto& value = g->intern.zero_byte; value.type = g->builtin_types.entry_u8; value.special = ConstValSpecialStatic; bigint_init_unsigned(&value.data.x_bigint, 0); } } static BuiltinFnEntry *create_builtin_fn(CodeGen *g, BuiltinFnId id, const char *name, size_t count) { BuiltinFnEntry *builtin_fn = heap::c_allocator.create(); buf_init_from_str(&builtin_fn->name, name); builtin_fn->id = id; builtin_fn->param_count = count; g->builtin_fn_table.put(&builtin_fn->name, builtin_fn); return builtin_fn; } static void define_builtin_fns(CodeGen *g) { create_builtin_fn(g, BuiltinFnIdBreakpoint, "breakpoint", 0); create_builtin_fn(g, BuiltinFnIdReturnAddress, "returnAddress", 0); create_builtin_fn(g, BuiltinFnIdMemcpy, "memcpy", 3); create_builtin_fn(g, BuiltinFnIdMemset, "memset", 3); create_builtin_fn(g, BuiltinFnIdSizeof, "sizeOf", 1); create_builtin_fn(g, BuiltinFnIdAlignOf, "alignOf", 1); create_builtin_fn(g, BuiltinFnIdField, "field", 2); create_builtin_fn(g, BuiltinFnIdTypeInfo, "typeInfo", 1); create_builtin_fn(g, BuiltinFnIdType, "Type", 1); create_builtin_fn(g, BuiltinFnIdHasField, "hasField", 2); create_builtin_fn(g, BuiltinFnIdTypeof, "TypeOf", SIZE_MAX); create_builtin_fn(g, BuiltinFnIdAddWithOverflow, "addWithOverflow", 4); create_builtin_fn(g, BuiltinFnIdSubWithOverflow, "subWithOverflow", 4); create_builtin_fn(g, BuiltinFnIdMulWithOverflow, "mulWithOverflow", 4); create_builtin_fn(g, BuiltinFnIdShlWithOverflow, "shlWithOverflow", 4); create_builtin_fn(g, BuiltinFnIdCInclude, "cInclude", 1); create_builtin_fn(g, BuiltinFnIdCDefine, "cDefine", 2); create_builtin_fn(g, BuiltinFnIdCUndef, "cUndef", 1); create_builtin_fn(g, BuiltinFnIdCtz, "ctz", 2); create_builtin_fn(g, BuiltinFnIdClz, "clz", 2); create_builtin_fn(g, BuiltinFnIdPopCount, "popCount", 2); create_builtin_fn(g, BuiltinFnIdBswap, "byteSwap", 2); create_builtin_fn(g, BuiltinFnIdBitReverse, "bitReverse", 2); create_builtin_fn(g, BuiltinFnIdImport, "import", 1); create_builtin_fn(g, BuiltinFnIdCImport, "cImport", 1); create_builtin_fn(g, BuiltinFnIdErrName, "errorName", 1); create_builtin_fn(g, BuiltinFnIdTypeName, "typeName", 1); create_builtin_fn(g, BuiltinFnIdEmbedFile, "embedFile", 1); create_builtin_fn(g, BuiltinFnIdCmpxchgWeak, "cmpxchgWeak", 6); create_builtin_fn(g, BuiltinFnIdCmpxchgStrong, "cmpxchgStrong", 6); create_builtin_fn(g, BuiltinFnIdFence, "fence", 1); create_builtin_fn(g, BuiltinFnIdTruncate, "truncate", 2); create_builtin_fn(g, BuiltinFnIdIntCast, "intCast", 2); create_builtin_fn(g, BuiltinFnIdFloatCast, "floatCast", 2); create_builtin_fn(g, BuiltinFnIdIntToFloat, "intToFloat", 2); create_builtin_fn(g, BuiltinFnIdFloatToInt, "floatToInt", 2); create_builtin_fn(g, BuiltinFnIdBoolToInt, "boolToInt", 1); create_builtin_fn(g, BuiltinFnIdErrToInt, "errorToInt", 1); create_builtin_fn(g, BuiltinFnIdIntToErr, "intToError", 1); create_builtin_fn(g, BuiltinFnIdEnumToInt, "enumToInt", 1); create_builtin_fn(g, BuiltinFnIdIntToEnum, "intToEnum", 2); create_builtin_fn(g, BuiltinFnIdCompileErr, "compileError", 1); create_builtin_fn(g, BuiltinFnIdCompileLog, "compileLog", SIZE_MAX); create_builtin_fn(g, BuiltinFnIdVectorType, "Vector", 2); create_builtin_fn(g, BuiltinFnIdShuffle, "shuffle", 4); create_builtin_fn(g, BuiltinFnIdSplat, "splat", 2); create_builtin_fn(g, BuiltinFnIdSetCold, "setCold", 1); create_builtin_fn(g, BuiltinFnIdSetRuntimeSafety, "setRuntimeSafety", 1); create_builtin_fn(g, BuiltinFnIdSetFloatMode, "setFloatMode", 1); create_builtin_fn(g, BuiltinFnIdPanic, "panic", 1); create_builtin_fn(g, BuiltinFnIdPtrCast, "ptrCast", 2); create_builtin_fn(g, BuiltinFnIdBitCast, "bitCast", 2); create_builtin_fn(g, BuiltinFnIdIntToPtr, "intToPtr", 2); create_builtin_fn(g, BuiltinFnIdPtrToInt, "ptrToInt", 1); create_builtin_fn(g, BuiltinFnIdTagName, "tagName", 1); create_builtin_fn(g, BuiltinFnIdTagType, "TagType", 1); create_builtin_fn(g, BuiltinFnIdFieldParentPtr, "fieldParentPtr", 3); create_builtin_fn(g, BuiltinFnIdByteOffsetOf, "byteOffsetOf", 2); create_builtin_fn(g, BuiltinFnIdBitOffsetOf, "bitOffsetOf", 2); create_builtin_fn(g, BuiltinFnIdDivExact, "divExact", 2); create_builtin_fn(g, BuiltinFnIdDivTrunc, "divTrunc", 2); create_builtin_fn(g, BuiltinFnIdDivFloor, "divFloor", 2); create_builtin_fn(g, BuiltinFnIdRem, "rem", 2); create_builtin_fn(g, BuiltinFnIdMod, "mod", 2); create_builtin_fn(g, BuiltinFnIdSqrt, "sqrt", 1); create_builtin_fn(g, BuiltinFnIdSin, "sin", 1); create_builtin_fn(g, BuiltinFnIdCos, "cos", 1); create_builtin_fn(g, BuiltinFnIdExp, "exp", 1); create_builtin_fn(g, BuiltinFnIdExp2, "exp2", 1); create_builtin_fn(g, BuiltinFnIdLog, "log", 1); create_builtin_fn(g, BuiltinFnIdLog2, "log2", 1); create_builtin_fn(g, BuiltinFnIdLog10, "log10", 1); create_builtin_fn(g, BuiltinFnIdFabs, "fabs", 1); create_builtin_fn(g, BuiltinFnIdFloor, "floor", 1); create_builtin_fn(g, BuiltinFnIdCeil, "ceil", 1); create_builtin_fn(g, BuiltinFnIdTrunc, "trunc", 1); create_builtin_fn(g, BuiltinFnIdNearbyInt, "nearbyInt", 1); create_builtin_fn(g, BuiltinFnIdRound, "round", 1); create_builtin_fn(g, BuiltinFnIdMulAdd, "mulAdd", 4); create_builtin_fn(g, BuiltinFnIdAsyncCall, "asyncCall", SIZE_MAX); create_builtin_fn(g, BuiltinFnIdShlExact, "shlExact", 2); create_builtin_fn(g, BuiltinFnIdShrExact, "shrExact", 2); create_builtin_fn(g, BuiltinFnIdSetEvalBranchQuota, "setEvalBranchQuota", 1); create_builtin_fn(g, BuiltinFnIdAlignCast, "alignCast", 2); create_builtin_fn(g, BuiltinFnIdSetAlignStack, "setAlignStack", 1); create_builtin_fn(g, BuiltinFnIdExport, "export", 2); create_builtin_fn(g, BuiltinFnIdExtern, "extern1", 2); create_builtin_fn(g, BuiltinFnIdErrorReturnTrace, "errorReturnTrace", 0); create_builtin_fn(g, BuiltinFnIdAtomicRmw, "atomicRmw", 5); create_builtin_fn(g, BuiltinFnIdAtomicLoad, "atomicLoad", 3); create_builtin_fn(g, BuiltinFnIdAtomicStore, "atomicStore", 4); create_builtin_fn(g, BuiltinFnIdErrSetCast, "errSetCast", 2); create_builtin_fn(g, BuiltinFnIdThis, "This", 0); create_builtin_fn(g, BuiltinFnIdHasDecl, "hasDecl", 2); create_builtin_fn(g, BuiltinFnIdUnionInit, "unionInit", 3); create_builtin_fn(g, BuiltinFnIdFrameHandle, "frame", 0); create_builtin_fn(g, BuiltinFnIdFrameType, "Frame", 1); create_builtin_fn(g, BuiltinFnIdFrameAddress, "frameAddress", 0); create_builtin_fn(g, BuiltinFnIdFrameSize, "frameSize", 1); create_builtin_fn(g, BuiltinFnIdAs, "as", 2); create_builtin_fn(g, BuiltinFnIdCall, "call", 3); create_builtin_fn(g, BuiltinFnIdBitSizeof, "bitSizeOf", 1); create_builtin_fn(g, BuiltinFnIdWasmMemorySize, "wasmMemorySize", 1); create_builtin_fn(g, BuiltinFnIdWasmMemoryGrow, "wasmMemoryGrow", 2); create_builtin_fn(g, BuiltinFnIdSrc, "src", 0); create_builtin_fn(g, BuiltinFnIdReduce, "reduce", 2); } static const char *bool_to_str(bool b) { return b ? "true" : "false"; } static const char *build_mode_to_str(BuildMode build_mode) { switch (build_mode) { case BuildModeDebug: return "Mode.Debug"; case BuildModeSafeRelease: return "Mode.ReleaseSafe"; case BuildModeFastRelease: return "Mode.ReleaseFast"; case BuildModeSmallRelease: return "Mode.ReleaseSmall"; } zig_unreachable(); } static const char *subsystem_to_str(TargetSubsystem subsystem) { switch (subsystem) { case TargetSubsystemConsole: return "Console"; case TargetSubsystemWindows: return "Windows"; case TargetSubsystemPosix: return "Posix"; case TargetSubsystemNative: return "Native"; case TargetSubsystemEfiApplication: return "EfiApplication"; case TargetSubsystemEfiBootServiceDriver: return "EfiBootServiceDriver"; case TargetSubsystemEfiRom: return "EfiRom"; case TargetSubsystemEfiRuntimeDriver: return "EfiRuntimeDriver"; case TargetSubsystemAuto: zig_unreachable(); } zig_unreachable(); } // Returns TargetSubsystemAuto to mean "no subsystem" TargetSubsystem detect_subsystem(CodeGen *g) { if (g->subsystem != TargetSubsystemAuto) return g->subsystem; if (g->zig_target->os == OsWindows) { if (g->stage1.have_dllmain_crt_startup) return TargetSubsystemAuto; if (g->stage1.have_c_main || g->is_test_build || g->stage1.have_winmain_crt_startup || g->stage1.have_wwinmain_crt_startup) return TargetSubsystemConsole; if (g->stage1.have_winmain || g->stage1.have_wwinmain) return TargetSubsystemWindows; } else if (g->zig_target->os == OsUefi) { return TargetSubsystemEfiApplication; } return TargetSubsystemAuto; } static bool detect_err_ret_tracing(CodeGen *g) { return !g->strip_debug_symbols && g->build_mode != BuildModeFastRelease && g->build_mode != BuildModeSmallRelease; } static LLVMCodeModel to_llvm_code_model(CodeGen *g) { switch (g->code_model) { case CodeModelDefault: return LLVMCodeModelDefault; case CodeModelTiny: return LLVMCodeModelTiny; case CodeModelSmall: return LLVMCodeModelSmall; case CodeModelKernel: return LLVMCodeModelKernel; case CodeModelMedium: return LLVMCodeModelMedium; case CodeModelLarge: return LLVMCodeModelLarge; } zig_unreachable(); } Buf *codegen_generate_builtin_source(CodeGen *g) { // Note that this only runs when zig0 is building the self-hosted zig compiler code, // so it makes a few assumption that are always true for that case. Once we have // built the stage2 zig components then zig is in charge of generating the builtin.zig // file. g->have_err_ret_tracing = detect_err_ret_tracing(g); Buf *contents = buf_alloc(); buf_appendf(contents, "usingnamespace @import(\"std\").builtin;\n\n"); const char *cur_os = nullptr; { uint32_t field_count = (uint32_t)target_os_count(); for (uint32_t i = 0; i < field_count; i += 1) { Os os_type = target_os_enum(i); const char *name = target_os_name(os_type); if (os_type == g->zig_target->os) { cur_os = name; } } } assert(cur_os != nullptr); const char *cur_arch = nullptr; { uint32_t field_count = (uint32_t)target_arch_count(); for (uint32_t arch_i = 0; arch_i < field_count; arch_i += 1) { ZigLLVM_ArchType arch = target_arch_enum(arch_i); const char *arch_name = target_arch_name(arch); if (arch == g->zig_target->arch) { cur_arch = arch_name; } } } assert(cur_arch != nullptr); const char *cur_abi = nullptr; { uint32_t field_count = (uint32_t)target_abi_count(); for (uint32_t i = 0; i < field_count; i += 1) { ZigLLVM_EnvironmentType abi = target_abi_enum(i); const char *name = target_abi_name(abi); if (abi == g->zig_target->abi) { cur_abi = name; } } } assert(cur_abi != nullptr); const char *cur_obj_fmt = nullptr; { uint32_t field_count = (uint32_t)target_oformat_count(); for (uint32_t i = 0; i < field_count; i += 1) { ZigLLVM_ObjectFormatType oformat = target_oformat_enum(i); const char *name = target_oformat_name(oformat); ZigLLVM_ObjectFormatType target_oformat = target_object_format(g->zig_target); if (oformat == target_oformat) { cur_obj_fmt = name; } } } assert(cur_obj_fmt != nullptr); // If any of these asserts trip then you need to either fix the internal compiler enum // or the corresponding one in std.Target or std.builtin. static_assert(ContainerLayoutAuto == 0, ""); static_assert(ContainerLayoutExtern == 1, ""); static_assert(ContainerLayoutPacked == 2, ""); static_assert(CallingConventionUnspecified == 0, ""); static_assert(CallingConventionC == 1, ""); static_assert(CallingConventionNaked == 2, ""); static_assert(CallingConventionAsync == 3, ""); static_assert(CallingConventionInterrupt == 4, ""); static_assert(CallingConventionSignal == 5, ""); static_assert(CallingConventionStdcall == 6, ""); static_assert(CallingConventionFastcall == 7, ""); static_assert(CallingConventionVectorcall == 8, ""); static_assert(CallingConventionThiscall == 9, ""); static_assert(CallingConventionAPCS == 10, ""); static_assert(CallingConventionAAPCS == 11, ""); static_assert(CallingConventionAAPCSVFP == 12, ""); static_assert(FnInlineAuto == 0, ""); static_assert(FnInlineAlways == 1, ""); static_assert(FnInlineNever == 2, ""); static_assert(BuiltinPtrSizeOne == 0, ""); static_assert(BuiltinPtrSizeMany == 1, ""); static_assert(BuiltinPtrSizeSlice == 2, ""); static_assert(BuiltinPtrSizeC == 3, ""); static_assert(TargetSubsystemConsole == 0, ""); static_assert(TargetSubsystemWindows == 1, ""); static_assert(TargetSubsystemPosix == 2, ""); static_assert(TargetSubsystemNative == 3, ""); static_assert(TargetSubsystemEfiApplication == 4, ""); static_assert(TargetSubsystemEfiBootServiceDriver == 5, ""); static_assert(TargetSubsystemEfiRom == 6, ""); static_assert(TargetSubsystemEfiRuntimeDriver == 7, ""); buf_append_str(contents, "/// Deprecated: use `std.Target.current.cpu.arch`\n"); buf_append_str(contents, "pub const arch = Target.current.cpu.arch;\n"); buf_append_str(contents, "/// Deprecated: use `std.Target.current.cpu.arch.endian()`\n"); buf_append_str(contents, "pub const endian = Target.current.cpu.arch.endian();\n"); buf_appendf(contents, "pub const output_mode = OutputMode.Obj;\n"); buf_appendf(contents, "pub const link_mode = LinkMode.%s;\n", ZIG_LINK_MODE); buf_appendf(contents, "pub const is_test = false;\n"); buf_appendf(contents, "pub const single_threaded = %s;\n", bool_to_str(g->is_single_threaded)); buf_appendf(contents, "pub const abi = Abi.%s;\n", cur_abi); buf_appendf(contents, "pub const cpu: Cpu = Target.Cpu.baseline(.%s);\n", cur_arch); buf_appendf(contents, "pub const os = Target.Os.Tag.defaultVersionRange(.%s);\n", cur_os); buf_appendf(contents, "pub const object_format = ObjectFormat.%s;\n", cur_obj_fmt); buf_appendf(contents, "pub const mode = %s;\n", build_mode_to_str(g->build_mode)); buf_appendf(contents, "pub const link_libc = %s;\n", bool_to_str(g->link_libc)); buf_appendf(contents, "pub const link_libcpp = %s;\n", bool_to_str(g->link_libcpp)); buf_appendf(contents, "pub const have_error_return_tracing = %s;\n", bool_to_str(g->have_err_ret_tracing)); buf_appendf(contents, "pub const valgrind_support = false;\n"); buf_appendf(contents, "pub const position_independent_code = %s;\n", bool_to_str(g->have_pic)); buf_appendf(contents, "pub const strip_debug_info = %s;\n", bool_to_str(g->strip_debug_symbols)); buf_appendf(contents, "pub const code_model = CodeModel.default;\n"); { TargetSubsystem detected_subsystem = detect_subsystem(g); if (detected_subsystem != TargetSubsystemAuto) { buf_appendf(contents, "pub const explicit_subsystem = SubSystem.%s;\n", subsystem_to_str(detected_subsystem)); } } return contents; } static ZigPackage *create_test_runner_pkg(CodeGen *g) { return codegen_create_package(g, buf_ptr(g->zig_std_special_dir), "test_runner.zig", "std.special"); } static Error define_builtin_compile_vars(CodeGen *g) { Error err; if (g->std_package == nullptr) return ErrorNone; assert(g->main_pkg); const char *builtin_zig_basename = "builtin.zig"; Buf *contents; if (g->builtin_zig_path == nullptr) { // Then this is zig0 building stage2. We can make many assumptions about the compilation. Buf *out_dir = buf_alloc(); os_path_split(&g->o_file_output_path, out_dir, nullptr); g->builtin_zig_path = buf_alloc(); os_path_join(out_dir, buf_create_from_str(builtin_zig_basename), g->builtin_zig_path); Buf *resolve_paths[] = { g->builtin_zig_path, }; *g->builtin_zig_path = os_path_resolve(resolve_paths, 1); contents = codegen_generate_builtin_source(g); if ((err = os_write_file(g->builtin_zig_path, contents))) { fprintf(stderr, "Unable to write file '%s': %s\n", buf_ptr(g->builtin_zig_path), err_str(err)); exit(1); } g->compile_var_package = new_package(buf_ptr(out_dir), builtin_zig_basename, "builtin"); } else { Buf *resolve_paths[] = { g->builtin_zig_path, }; *g->builtin_zig_path = os_path_resolve(resolve_paths, 1); contents = buf_alloc(); if ((err = os_fetch_file_path(g->builtin_zig_path, contents))) { fprintf(stderr, "unable to open '%s': %s\n", buf_ptr(g->builtin_zig_path), err_str(err)); exit(1); } Buf builtin_dirname = BUF_INIT; os_path_dirname(g->builtin_zig_path, &builtin_dirname); g->compile_var_package = new_package(buf_ptr(&builtin_dirname), builtin_zig_basename, "builtin"); } if (g->is_test_build) { if (g->test_runner_package == nullptr) { g->test_runner_package = create_test_runner_pkg(g); } g->root_pkg = g->test_runner_package; } else { g->root_pkg = g->main_pkg; } g->compile_var_package->package_table.put(buf_create_from_str("std"), g->std_package); g->main_pkg->package_table.put(buf_create_from_str("builtin"), g->compile_var_package); g->main_pkg->package_table.put(buf_create_from_str("root"), g->root_pkg); g->std_package->package_table.put(buf_create_from_str("builtin"), g->compile_var_package); g->std_package->package_table.put(buf_create_from_str("std"), g->std_package); g->std_package->package_table.put(buf_create_from_str("root"), g->root_pkg); g->compile_var_import = add_source_file(g, g->compile_var_package, g->builtin_zig_path, contents, SourceKindPkgMain); return ErrorNone; } static void init(CodeGen *g) { if (g->module) return; codegen_add_time_event(g, "Initialize"); { const char *progress_name = "Initialize"; codegen_switch_sub_prog_node(g, stage2_progress_start(g->main_progress_node, progress_name, strlen(progress_name), 0)); } g->have_err_ret_tracing = detect_err_ret_tracing(g); assert(g->root_out_name); g->module = LLVMModuleCreateWithName(buf_ptr(g->root_out_name)); LLVMSetTarget(g->module, buf_ptr(&g->llvm_triple_str)); if (target_object_format(g->zig_target) == ZigLLVM_COFF) { ZigLLVMAddModuleCodeViewFlag(g->module); } else { ZigLLVMAddModuleDebugInfoFlag(g->module); } LLVMTargetRef target_ref; char *err_msg = nullptr; if (LLVMGetTargetFromTriple(buf_ptr(&g->llvm_triple_str), &target_ref, &err_msg)) { fprintf(stderr, "Zig is expecting LLVM to understand this target: '%s'\n" "However LLVM responded with: \"%s\"\n" "Zig is unable to continue. This is a bug in Zig:\n" "https://github.com/ziglang/zig/issues/438\n" , buf_ptr(&g->llvm_triple_str), err_msg); exit(1); } bool is_optimized = g->build_mode != BuildModeDebug; LLVMCodeGenOptLevel opt_level = is_optimized ? LLVMCodeGenLevelAggressive : LLVMCodeGenLevelNone; LLVMRelocMode reloc_mode; if (g->have_pic) { reloc_mode = LLVMRelocPIC; } else if (g->link_mode_dynamic) { reloc_mode = LLVMRelocDynamicNoPic; } else { reloc_mode = LLVMRelocStatic; } const char *target_specific_cpu_args = ""; const char *target_specific_features = ""; if (g->zig_target->is_native_cpu) { target_specific_cpu_args = ZigLLVMGetHostCPUName(); target_specific_features = ZigLLVMGetNativeFeatures(); } // Override CPU and features if defined by user. if (g->zig_target->llvm_cpu_name != nullptr) { target_specific_cpu_args = g->zig_target->llvm_cpu_name; } if (g->zig_target->llvm_cpu_features != nullptr) { target_specific_features = g->zig_target->llvm_cpu_features; } if (g->verbose_llvm_cpu_features) { fprintf(stderr, "name=%s triple=%s\n", buf_ptr(g->root_out_name), buf_ptr(&g->llvm_triple_str)); fprintf(stderr, "name=%s target_specific_cpu_args=%s\n", buf_ptr(g->root_out_name), target_specific_cpu_args); fprintf(stderr, "name=%s target_specific_features=%s\n", buf_ptr(g->root_out_name), target_specific_features); } // TODO handle float ABI better- it should depend on the ABI portion of std.Target ZigLLVMABIType float_abi = ZigLLVMABITypeDefault; // TODO a way to override this as part of std.Target ABI? const char *abi_name = nullptr; if (target_is_riscv(g->zig_target)) { // RISC-V Linux defaults to ilp32d/lp64d if (g->zig_target->os == OsLinux) { abi_name = (g->zig_target->arch == ZigLLVM_riscv32) ? "ilp32d" : "lp64d"; } else { abi_name = (g->zig_target->arch == ZigLLVM_riscv32) ? "ilp32" : "lp64"; } } g->target_machine = ZigLLVMCreateTargetMachine(target_ref, buf_ptr(&g->llvm_triple_str), target_specific_cpu_args, target_specific_features, opt_level, reloc_mode, to_llvm_code_model(g), g->function_sections, float_abi, abi_name); g->target_data_ref = LLVMCreateTargetDataLayout(g->target_machine); char *layout_str = LLVMCopyStringRepOfTargetData(g->target_data_ref); LLVMSetDataLayout(g->module, layout_str); assert(g->pointer_size_bytes == LLVMPointerSize(g->target_data_ref)); g->is_big_endian = (LLVMByteOrder(g->target_data_ref) == LLVMBigEndian); g->builder = LLVMCreateBuilder(); g->dbuilder = ZigLLVMCreateDIBuilder(g->module, true); // Don't use ZIG_VERSION_STRING here, llvm misparses it when it includes // the git revision. Buf *producer = buf_sprintf("zig %d.%d.%d", ZIG_VERSION_MAJOR, ZIG_VERSION_MINOR, ZIG_VERSION_PATCH); const char *flags = ""; unsigned runtime_version = 0; // For macOS stack traces, we want to avoid having to parse the compilation unit debug // info. As long as each debug info file has a path independent of the compilation unit // directory (DW_AT_comp_dir), then we never have to look at the compilation unit debug // info. If we provide an absolute path to LLVM here for the compilation unit debug info, // LLVM will emit DWARF info that depends on DW_AT_comp_dir. To avoid this, we pass "." // for the compilation unit directory. This forces each debug file to have a directory // rather than be relative to DW_AT_comp_dir. According to DWARF 5, debug files will // no longer reference DW_AT_comp_dir, for the purpose of being able to support the // common practice of stripping all but the line number sections from an executable. const char *compile_unit_dir = target_os_is_darwin(g->zig_target->os) ? "." : buf_ptr(&g->main_pkg->root_src_dir); ZigLLVMDIFile *compile_unit_file = ZigLLVMCreateFile(g->dbuilder, buf_ptr(g->root_out_name), compile_unit_dir); g->compile_unit = ZigLLVMCreateCompileUnit(g->dbuilder, ZigLLVMLang_DW_LANG_C99(), compile_unit_file, buf_ptr(producer), is_optimized, flags, runtime_version, "", 0, !g->strip_debug_symbols); // This is for debug stuff that doesn't have a real file. g->dummy_di_file = nullptr; define_builtin_types(g); define_intern_values(g); IrInstGen *sentinel_instructions = heap::c_allocator.allocate(2); g->invalid_inst_gen = &sentinel_instructions[0]; g->invalid_inst_gen->value = g->pass1_arena->create(); g->invalid_inst_gen->value->type = g->builtin_types.entry_invalid; g->unreach_instruction = &sentinel_instructions[1]; g->unreach_instruction->value = g->pass1_arena->create(); g->unreach_instruction->value->type = g->builtin_types.entry_unreachable; g->invalid_inst_src = heap::c_allocator.create(); define_builtin_fns(g); Error err; if ((err = define_builtin_compile_vars(g))) { fprintf(stderr, "Unable to create builtin.zig: %s\n", err_str(err)); exit(1); } } static void update_test_functions_builtin_decl(CodeGen *g) { Error err; assert(g->is_test_build); if (g->test_fns.length == 0) { fprintf(stderr, "No tests to run.\n"); exit(0); } ZigType *fn_type = get_test_fn_type(g); ZigValue *test_fn_type_val = get_builtin_value(g, "TestFn"); assert(test_fn_type_val->type->id == ZigTypeIdMetaType); ZigType *struct_type = test_fn_type_val->data.x_type; if ((err = type_resolve(g, struct_type, ResolveStatusSizeKnown))) zig_unreachable(); ZigValue *test_fn_array = g->pass1_arena->create(); test_fn_array->type = get_array_type(g, struct_type, g->test_fns.length, nullptr); test_fn_array->special = ConstValSpecialStatic; test_fn_array->data.x_array.data.s_none.elements = g->pass1_arena->allocate(g->test_fns.length); for (size_t i = 0; i < g->test_fns.length; i += 1) { ZigFn *test_fn_entry = g->test_fns.at(i); ZigValue *this_val = &test_fn_array->data.x_array.data.s_none.elements[i]; this_val->special = ConstValSpecialStatic; this_val->type = struct_type; this_val->parent.id = ConstParentIdArray; this_val->parent.data.p_array.array_val = test_fn_array; this_val->parent.data.p_array.elem_index = i; this_val->data.x_struct.fields = alloc_const_vals_ptrs(g, 3); ZigValue *name_field = this_val->data.x_struct.fields[0]; ZigValue *name_array_val = create_const_str_lit(g, &test_fn_entry->symbol_name)->data.x_ptr.data.ref.pointee; init_const_slice(g, name_field, name_array_val, 0, buf_len(&test_fn_entry->symbol_name), true); ZigValue *fn_field = this_val->data.x_struct.fields[1]; fn_field->type = fn_type; fn_field->special = ConstValSpecialStatic; fn_field->data.x_ptr.special = ConstPtrSpecialFunction; fn_field->data.x_ptr.mut = ConstPtrMutComptimeConst; fn_field->data.x_ptr.data.fn.fn_entry = test_fn_entry; ZigValue *frame_size_field = this_val->data.x_struct.fields[2]; frame_size_field->type = get_optional_type(g, g->builtin_types.entry_usize); frame_size_field->special = ConstValSpecialStatic; frame_size_field->data.x_optional = nullptr; if (fn_is_async(test_fn_entry)) { frame_size_field->data.x_optional = g->pass1_arena->create(); frame_size_field->data.x_optional->special = ConstValSpecialStatic; frame_size_field->data.x_optional->type = g->builtin_types.entry_usize; bigint_init_unsigned(&frame_size_field->data.x_optional->data.x_bigint, test_fn_entry->frame_type->abi_size); } } report_errors_and_maybe_exit(g); ZigValue *test_fn_slice = create_const_slice(g, test_fn_array, 0, g->test_fns.length, true); update_compile_var(g, buf_create_from_str("test_functions"), test_fn_slice); assert(g->test_runner_package != nullptr); } static Buf *get_resolved_root_src_path(CodeGen *g) { // TODO memoize if (buf_len(&g->main_pkg->root_src_path) == 0) return nullptr; Buf rel_full_path = BUF_INIT; os_path_join(&g->main_pkg->root_src_dir, &g->main_pkg->root_src_path, &rel_full_path); Buf *resolved_path = buf_alloc(); Buf *resolve_paths[] = {&rel_full_path}; *resolved_path = os_path_resolve(resolve_paths, 1); return resolved_path; } static void gen_root_source(CodeGen *g) { Buf *resolved_path = get_resolved_root_src_path(g); if (resolved_path == nullptr) return; Buf *source_code = buf_alloc(); Error err; // No need for using the caching system for this file fetch because it is handled // separately. if ((err = os_fetch_file_path(resolved_path, source_code))) { fprintf(stderr, "unable to open '%s': %s\n", buf_ptr(resolved_path), err_str(err)); exit(1); } ZigType *root_import_alias = add_source_file(g, g->main_pkg, resolved_path, source_code, SourceKindRoot); assert(root_import_alias == g->root_import); assert(g->root_out_name); // Zig has lazy top level definitions. Here we semantically analyze the panic function. Buf *import_target_path; Buf full_path = BUF_INIT; ZigType *std_import; if ((err = analyze_import(g, g->root_import, buf_create_from_str("std"), &std_import, &import_target_path, &full_path))) { if (err == ErrorFileNotFound) { fprintf(stderr, "unable to find '%s'", buf_ptr(import_target_path)); } else { fprintf(stderr, "unable to open '%s': %s\n", buf_ptr(&full_path), err_str(err)); } exit(1); } Tld *builtin_tld = find_decl(g, &get_container_scope(std_import)->base, buf_create_from_str("builtin")); assert(builtin_tld != nullptr); resolve_top_level_decl(g, builtin_tld, nullptr, false); report_errors_and_maybe_exit(g); assert(builtin_tld->id == TldIdVar); TldVar *builtin_tld_var = (TldVar*)builtin_tld; ZigValue *builtin_val = builtin_tld_var->var->const_value; assert(builtin_val->type->id == ZigTypeIdMetaType); ZigType *builtin_type = builtin_val->data.x_type; Tld *panic_tld = find_decl(g, &get_container_scope(builtin_type)->base, buf_create_from_str("panic")); assert(panic_tld != nullptr); resolve_top_level_decl(g, panic_tld, nullptr, false); report_errors_and_maybe_exit(g); assert(panic_tld->id == TldIdVar); TldVar *panic_tld_var = (TldVar*)panic_tld; ZigValue *panic_fn_val = panic_tld_var->var->const_value; assert(panic_fn_val->type->id == ZigTypeIdFn); assert(panic_fn_val->data.x_ptr.special == ConstPtrSpecialFunction); g->panic_fn = panic_fn_val->data.x_ptr.data.fn.fn_entry; assert(g->panic_fn != nullptr); if (!g->error_during_imports) { semantic_analyze(g); } report_errors_and_maybe_exit(g); if (g->is_test_build) { update_test_functions_builtin_decl(g); if (!g->error_during_imports) { semantic_analyze(g); } } report_errors_and_maybe_exit(g); } void codegen_print_timing_report(CodeGen *g, FILE *f) { double start_time = g->timing_events.at(0).time; double end_time = g->timing_events.last().time; double total = end_time - start_time; fprintf(f, "%20s%12s%12s%12s%12s\n", "Name", "Start", "End", "Duration", "Percent"); for (size_t i = 0; i < g->timing_events.length - 1; i += 1) { TimeEvent *te = &g->timing_events.at(i); TimeEvent *next_te = &g->timing_events.at(i + 1); fprintf(f, "%20s%12.4f%12.4f%12.4f%12.4f\n", te->name, te->time - start_time, next_te->time - start_time, next_te->time - te->time, (next_te->time - te->time) / total); } fprintf(f, "%20s%12.4f%12.4f%12.4f%12.4f\n", "Total", 0.0, total, total, 1.0); } void codegen_add_time_event(CodeGen *g, const char *name) { OsTimeStamp timestamp = os_timestamp_monotonic(); double seconds = (double)timestamp.sec; seconds += ((double)timestamp.nsec) / 1000000000.0; g->timing_events.append({seconds, name}); } void codegen_build_object(CodeGen *g) { g->have_err_ret_tracing = detect_err_ret_tracing(g); init(g); codegen_add_time_event(g, "Semantic Analysis"); const char *progress_name = "Semantic Analysis"; codegen_switch_sub_prog_node(g, stage2_progress_start(g->main_progress_node, progress_name, strlen(progress_name), 0)); gen_root_source(g); if (buf_len(&g->analysis_json_output_path) != 0) { const char *analysis_json_filename = buf_ptr(&g->analysis_json_output_path); FILE *f = fopen(analysis_json_filename, "wb"); if (f == nullptr) { fprintf(stderr, "Unable to open '%s': %s\n", analysis_json_filename, strerror(errno)); exit(1); } zig_print_analysis_dump(g, f, " ", "\n"); if (fclose(f) != 0) { fprintf(stderr, "Unable to write '%s': %s\n", analysis_json_filename, strerror(errno)); exit(1); } } if (buf_len(&g->docs_output_path) != 0) { Error err; Buf *doc_dir_path = &g->docs_output_path; if ((err = os_make_path(doc_dir_path))) { fprintf(stderr, "Unable to create directory %s: %s\n", buf_ptr(doc_dir_path), err_str(err)); exit(1); } Buf *index_html_src_path = buf_sprintf("%s" OS_SEP "special" OS_SEP "docs" OS_SEP "index.html", buf_ptr(g->zig_std_dir)); Buf *index_html_dest_path = buf_sprintf("%s" OS_SEP "index.html", buf_ptr(doc_dir_path)); Buf *main_js_src_path = buf_sprintf("%s" OS_SEP "special" OS_SEP "docs" OS_SEP "main.js", buf_ptr(g->zig_std_dir)); Buf *main_js_dest_path = buf_sprintf("%s" OS_SEP "main.js", buf_ptr(doc_dir_path)); if ((err = os_copy_file(index_html_src_path, index_html_dest_path))) { fprintf(stderr, "Unable to copy %s to %s: %s\n", buf_ptr(index_html_src_path), buf_ptr(index_html_dest_path), err_str(err)); exit(1); } if ((err = os_copy_file(main_js_src_path, main_js_dest_path))) { fprintf(stderr, "Unable to copy %s to %s: %s\n", buf_ptr(main_js_src_path), buf_ptr(main_js_dest_path), err_str(err)); exit(1); } const char *data_js_filename = buf_ptr(buf_sprintf("%s" OS_SEP "data.js", buf_ptr(doc_dir_path))); FILE *f = fopen(data_js_filename, "wb"); if (f == nullptr) { fprintf(stderr, "Unable to open '%s': %s\n", data_js_filename, strerror(errno)); exit(1); } fprintf(f, "zigAnalysis="); zig_print_analysis_dump(g, f, "", ""); fprintf(f, ";"); if (fclose(f) != 0) { fprintf(stderr, "Unable to write '%s': %s\n", data_js_filename, strerror(errno)); exit(1); } } codegen_add_time_event(g, "Code Generation"); { const char *progress_name = "Code Generation"; codegen_switch_sub_prog_node(g, stage2_progress_start(g->main_progress_node, progress_name, strlen(progress_name), 0)); } do_code_gen(g); codegen_add_time_event(g, "LLVM Emit Output"); { const char *progress_name = "LLVM Emit Output"; codegen_switch_sub_prog_node(g, stage2_progress_start(g->main_progress_node, progress_name, strlen(progress_name), 0)); } zig_llvm_emit_output(g); codegen_add_time_event(g, "Done"); codegen_switch_sub_prog_node(g, nullptr); } ZigPackage *codegen_create_package(CodeGen *g, const char *root_src_dir, const char *root_src_path, const char *pkg_path) { init(g); ZigPackage *pkg = new_package(root_src_dir, root_src_path, pkg_path); if (g->std_package != nullptr) { assert(g->compile_var_package != nullptr); pkg->package_table.put(buf_create_from_str("std"), g->std_package); pkg->package_table.put(buf_create_from_str("root"), g->root_pkg); pkg->package_table.put(buf_create_from_str("builtin"), g->compile_var_package); } return pkg; } void codegen_destroy(CodeGen *g) { if (g->pass1_arena != nullptr) { g->pass1_arena->destruct(&heap::c_allocator); g->pass1_arena = nullptr; } heap::c_allocator.destroy(g); } CodeGen *codegen_create(Buf *main_pkg_path, Buf *root_src_path, const ZigTarget *target, BuildMode build_mode, Buf *override_lib_dir, bool is_test_build) { CodeGen *g = heap::c_allocator.create(); g->pass1_arena = heap::ArenaAllocator::construct(&heap::c_allocator, &heap::c_allocator, "pass1"); g->subsystem = TargetSubsystemAuto; g->zig_target = target; assert(override_lib_dir != nullptr); g->zig_lib_dir = override_lib_dir; g->zig_std_dir = buf_alloc(); os_path_join(g->zig_lib_dir, buf_create_from_str("std"), g->zig_std_dir); g->build_mode = build_mode; g->import_table.init(32); g->builtin_fn_table.init(32); g->primitive_type_table.init(32); g->type_table.init(32); g->fn_type_table.init(32); g->error_table.init(16); g->generic_table.init(16); g->llvm_fn_table.init(16); g->memoized_fn_eval_table.init(16); g->exported_symbol_names.init(8); g->external_symbol_names.init(8); g->string_literals_table.init(16); g->type_info_cache.init(32); g->one_possible_values.init(32); g->is_test_build = is_test_build; g->is_single_threaded = false; g->code_model = CodeModelDefault; buf_resize(&g->global_asm, 0); for (size_t i = 0; i < array_length(symbols_that_llvm_depends_on); i += 1) { g->external_symbol_names.put(buf_create_from_str(symbols_that_llvm_depends_on[i]), nullptr); } if (root_src_path) { Buf *root_pkg_path; Buf *rel_root_src_path; if (main_pkg_path == nullptr) { Buf *src_basename = buf_alloc(); Buf *src_dir = buf_alloc(); os_path_split(root_src_path, src_dir, src_basename); if (buf_len(src_basename) == 0) { fprintf(stderr, "Invalid root source path: %s\n", buf_ptr(root_src_path)); exit(1); } root_pkg_path = src_dir; rel_root_src_path = src_basename; } else { Buf resolved_root_src_path = os_path_resolve(&root_src_path, 1); Buf resolved_main_pkg_path = os_path_resolve(&main_pkg_path, 1); if (!buf_starts_with_buf(&resolved_root_src_path, &resolved_main_pkg_path)) { fprintf(stderr, "Root source path '%s' outside main package path '%s'\n", buf_ptr(root_src_path), buf_ptr(main_pkg_path)); exit(1); } root_pkg_path = main_pkg_path; rel_root_src_path = buf_create_from_mem( buf_ptr(&resolved_root_src_path) + buf_len(&resolved_main_pkg_path) + 1, buf_len(&resolved_root_src_path) - buf_len(&resolved_main_pkg_path) - 1); } g->main_pkg = new_package(buf_ptr(root_pkg_path), buf_ptr(rel_root_src_path), ""); g->std_package = new_package(buf_ptr(g->zig_std_dir), "std.zig", "std"); g->main_pkg->package_table.put(buf_create_from_str("std"), g->std_package); } else { g->main_pkg = new_package(".", "", ""); } g->zig_std_special_dir = buf_alloc(); os_path_join(g->zig_std_dir, buf_sprintf("special"), g->zig_std_special_dir); target_triple_llvm(&g->llvm_triple_str, g->zig_target); g->pointer_size_bytes = target_arch_pointer_bit_width(g->zig_target->arch) / 8; if (!target_has_debug_info(g->zig_target)) { g->strip_debug_symbols = true; } return g; } bool codegen_fn_has_err_ret_tracing_arg(CodeGen *g, ZigType *return_type) { return g->have_err_ret_tracing && (return_type->id == ZigTypeIdErrorUnion || return_type->id == ZigTypeIdErrorSet); } bool codegen_fn_has_err_ret_tracing_stack(CodeGen *g, ZigFn *fn, bool is_async) { if (is_async) { return g->have_err_ret_tracing && (fn->calls_or_awaits_errorable_fn || codegen_fn_has_err_ret_tracing_arg(g, fn->type_entry->data.fn.fn_type_id.return_type)); } else { return g->have_err_ret_tracing && fn->calls_or_awaits_errorable_fn && !codegen_fn_has_err_ret_tracing_arg(g, fn->type_entry->data.fn.fn_type_id.return_type); } } void codegen_switch_sub_prog_node(CodeGen *g, Stage2ProgressNode *node) { if (g->sub_progress_node != nullptr) { stage2_progress_end(g->sub_progress_node); } g->sub_progress_node = node; } ZigValue *CodeGen::Intern::for_undefined() { return &this->x_undefined; } ZigValue *CodeGen::Intern::for_void() { return &this->x_void; } ZigValue *CodeGen::Intern::for_null() { return &this->x_null; } ZigValue *CodeGen::Intern::for_unreachable() { return &this->x_unreachable; } ZigValue *CodeGen::Intern::for_zero_byte() { return &this->zero_byte; }