diff --git a/lib/common/compiler.h b/lib/common/compiler.h
index 31eb1ccd..07f875e4 100644
--- a/lib/common/compiler.h
+++ b/lib/common/compiler.h
@@ -89,7 +89,13 @@
 #endif
 
 /* prefetch
- * can be disabled, by declaring NO_PREFETCH macro */
+ * can be disabled, by declaring NO_PREFETCH macro
+ * All prefetch invocations use a single default locality 2,
+ * generating instruction prefetcht1,
+ * which, according to Intel, means "load data into L2 cache".
+ * This is a good enough "middle ground" for the time being,
+ * though in theory, it would be better to specialize locality depending on data being prefetched.
+ * Tests could not determine any sensible difference based on locality value. */
 #if defined(NO_PREFETCH)
 #  define PREFETCH(ptr)     (void)(ptr)  /* disabled */
 #else
diff --git a/lib/decompress/zstd_decompress.c b/lib/decompress/zstd_decompress.c
index 9d740833..1382c9c7 100644
--- a/lib/decompress/zstd_decompress.c
+++ b/lib/decompress/zstd_decompress.c
@@ -56,7 +56,8 @@
 *  Dependencies
 *********************************************************/
 #include <string.h>      /* memcpy, memmove, memset */
-#include "cpu.h"         /* prefetch */
+#include "compiler.h"    /* prefetch */
+#include "cpu.h"         /* bmi2 */
 #include "mem.h"         /* low level memory routines */
 #define FSE_STATIC_LINKING_ONLY
 #include "fse.h"