This commit adds enough Mach-O linker implementation to write out simple
Mach-O object file. Be warned however, the object file is largely incomplete:
misses relocation info, debug symbols, etc. However, it seemed like a
good starting to get the basic understanding right.
Signed-off-by: Jakub Konka <kubkon@jakubkonka.com>
* Move branch-local register and stack allocation metadata to the
function-local struct. Conditional branches clone this data in order
to restore it after generating machine code for a branch.
Branch-local data is now only the instruction table mapping *ir.Inst
to MCValue.
* Implement conditional branching
- Process operand deaths
- Handle register and stack allocation metadata
* Avoid storing unreferenced or void typed instructions into
the branch-local instruction table.
* Fix integer types reporting the wrong value for hasCodeGenBits.
* Remove the codegen optimization for eliding length-0 jumps. I need to
reexamine how this works because it was causing invalid jumps to be
emitted.
While we try to work out what the correlation between the OS and runtime
versions is, this commit hardcodes the latter to the minimum (compat)
version of 1.0.0.
Signed-off-by: Jakub Konka <kubkon@jakubkonka.com>
* move SPU code from std to self hosted compiler
* change std lib comments to be descriptive rather than prescriptive
* avoid usingnamespace
* fix case style of error codes
* remove duplication of producer_string
* generalize handling of less than 64 bit arch pointers
* clean up SPU II related test harness code
See #6113 for an alternate way of doing this that we didn't end up
following.
Closes#6079.
I also took the opportunity here to extract C.zig and Elf.zig from
link.zig.
According to the Mach-O file format reference, the first
load command should be a `__PAGEZERO` segment command. The
segment is located at virtual memory location 0, has no protection
rights, and causes acccesses to NULL to immediately crash.
Signed-off-by: Jakub Konka <kubkon@jakubkonka.com>
During codegen we do not yet know the indexes that will be used for
called functions. Therefore, we store the offset into the in-memory
code where the index is needed with a pointer to the Decl and use this
data to insert the proper indexes while writing the binary in the flush
function.
Before this commit the wasm backend worked similarly to elf. As
functions were generated they were written directly to the output file
and existing code was shifted around in the file as necessary. This
approach had several disadvantages:
- Large amounts of padding in the output were necessary to avoid
expensive copying of data within the file.
- Function/type/global/etc indexes were required to be known at the time
of preforming codegen, which severely limited the flexibility of where
code could be placed in the binary
- Significant complexity to track the state of the output file through
incremental updates
This commit takes things in a different direction. Code is incrementally
compiled into in-memory buffers and the entire binary is rewritten using
these buffers on flush. This has several advantages:
- Significantly smaller resulting binaries
- More performant resulting binaries due to lack of indirection
- Significantly simpler compiler code
- Indexes no longer need to be known before codegen. We can track where
Decls must be referenced by index insert the proper indexes while
writing the code in the flush() function. This is not yet implemented
but is planned for the next commit.
The main disadvantage is of course increased memory usage in order to
store these buffers of generated code.
This commit write out Mach-O header in the linker's `flush`
method. The header currently only populates the magic number,
filetype, and cpu info.
Signed-off-by: Jakub Konka <kubkon@jakubkonka.com>
Functions which are free'd are not immediately removed from the binary
as this would cause a shifting of function indexes. Instead, they hang
around until they can be overwritten by a new function. This means that
the types associated with these dead functions must also remain until
the function is overwritten to avoid a type mismatch.
Exports now have a dirty flag and are rewritten on flush if this flag
has been set.
A couple other minor changes have been made based on Andrew's review.
Thus far, we only generate the type, function, export, and code
sections. These are sufficient to generate and export simple functions.
Codegen is currently hardcoded to `i32.const 42`, the main goal of this
commit is to create infrastructure for the container format which will
work with incremental compilation.
