From 9f0a7ac0be869f14b3e3b629c775fe907c6ec775 Mon Sep 17 00:00:00 2001
From: David Carlier <devnexen@gmail.com>
Date: Sat, 16 Nov 2024 09:38:26 +0000
Subject: [PATCH] _mi_memcpy/_mi_memzero: tighten criteria for intrinsics for
 windows.

FSRM is better used for buffer <= 128 bytes and ERMS, if supported for larger
chunks.
---
 include/mimalloc/internal.h | 5 +++--
 src/init.c                  | 5 ++++-
 2 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/include/mimalloc/internal.h b/include/mimalloc/internal.h
index 82e8c766..71de25ad 100644
--- a/include/mimalloc/internal.h
+++ b/include/mimalloc/internal.h
@@ -911,8 +911,9 @@ static inline size_t mi_bsr(uintptr_t x) {
 #if !MI_TRACK_ENABLED && defined(_WIN32) && (defined(_M_IX86) || defined(_M_X64))
 #include <intrin.h>
 extern bool _mi_cpu_has_fsrm;
+extern bool _mi_cpu_has_erms;
 static inline void _mi_memcpy(void* dst, const void* src, size_t n) {
-  if (_mi_cpu_has_fsrm) {
+  if ((_mi_cpu_has_fsrm && n <= 128) || (_mi_cpu_has_erms && n > 128)) {
     __movsb((unsigned char*)dst, (const unsigned char*)src, n);
   }
   else {
@@ -920,7 +921,7 @@ static inline void _mi_memcpy(void* dst, const void* src, size_t n) {
   }
 }
 static inline void _mi_memzero(void* dst, size_t n) {
-  if (_mi_cpu_has_fsrm) {
+  if ((_mi_cpu_has_fsrm && n <= 128) || (_mi_cpu_has_erms && n > 128)) {
     __stosb((unsigned char*)dst, 0, n);
   }
   else {
diff --git a/src/init.c b/src/init.c
index 75458a1f..e25abd98 100644
--- a/src/init.c
+++ b/src/init.c
@@ -543,12 +543,15 @@ void _mi_process_load(void) {
 #if defined(_WIN32) && (defined(_M_IX86) || defined(_M_X64))
 #include <intrin.h>
 mi_decl_cache_align bool _mi_cpu_has_fsrm = false;
+mi_decl_cache_align bool _mi_cpu_has_erms = false;
 
 static void mi_detect_cpu_features(void) {
-  // FSRM for fast rep movsb support (AMD Zen3+ (~2020) or Intel Ice Lake+ (~2017))
+  // FSRM for fast short rep movsb/stosb support (AMD Zen3+ (~2020) or Intel Ice Lake+ (~2017))
+  // EMRS for fast enhanced rep movsb/stosb support
   int32_t cpu_info[4];
   __cpuid(cpu_info, 7);
   _mi_cpu_has_fsrm = ((cpu_info[3] & (1 << 4)) != 0); // bit 4 of EDX : see <https://en.wikipedia.org/wiki/CPUID#EAX=7,_ECX=0:_Extended_Features>
+  _mi_cpu_has_erms = ((cpu_info[2] & (1 << 9)) != 0); // bit 9 of ECX : see <https://en.wikipedia.org/wiki/CPUID#EAX=7,_ECX=0:_Extended_Features>
 }
 #else
 static void mi_detect_cpu_features(void) {