mirrors/kitty
1
0
Fork 0
mirror of https://github.com/kovidgoyal/kitty synced 2026-07-11 18:32:12 +02:00
Code Issues Packages Projects Releases Wiki Activity

Turns out that using 256 bit code on ARM is slightly faster even though it is emulated with 128 bit registers

Browse Source
This commit is contained in:
Kovid Goyal
2024-01-22 16:47:42 +05:30
parent 76c6630084
commit 4b846e0106
1 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
kitty/simd-string.c
View File

@@ -124,8 +124,8 @@ init_simd(void *x) {
#ifdef __APPLE__
#ifdef __arm64__
// simde takes care of NEON on Apple Silicon
// ARM has only 128 bit registers.
has_sse4_2 = true; has_avx2 = false;
// ARM has only 128 bit registers buy using the avx2 code is still slightly faster
has_sse4_2 = true; has_avx2 = true;
#else
do_check();
// On GitHub actions there are some weird macOS machines which report avx2 not available but sse4.2 is available and then
@@ -139,8 +139,8 @@ init_simd(void *x) {
#ifdef __aarch64__
// no idea how to probe ARM cpu for NEON support. This file uses pretty
// basic AVX2 and SSE4.2 intrinsics, so hopefully they work on ARM
// ARM has only 128 bit registers.
has_sse4_2 = true; has_avx2 = false;
// ARM has only 128 bit registers buy using the avx2 code is still slightly faster
has_sse4_2 = true; has_avx2 = true;
#else
do_check();
#endif