mirrors/Odin
1
0
Fork 0
mirror of https://github.com/odin-lang/Odin.git synced 2026-06-06 02:34:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

Skip bounds checking on the inner accumulate loop.

Browse Source 
This helps performance with SSE (somewhat) and AVX-512 (quite a bit),
but not AVX2 for some reason.
This commit is contained in:
Barinzaya
2025-07-31 16:51:42 -04:00
parent f61dc7d071
commit 4ef7ed1cbd
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
core/hash/xxhash/xxhash_3.odin
View File

@@ -64,7 +64,7 @@ XXH3_INTERNAL_BUFFER_SIZE :: 256
IMPORTANT: This structure has a strict alignment requirement of 64 bytes!! **
Default allocators will align it correctly if created via `new`, as will
placing this struct on the cache, but if using a custom allocator make sure
placing this struct on the stack, but if using a custom allocator make sure
that it handles the alignment correctly!
*/
XXH3_state :: struct #align(64) {
@@ -870,7 +870,7 @@ XXH_PREFETCH_DIST :: 320
XXH3_accumulate :: #force_inline proc(
acc: []xxh_u64, input: []u8, secret: []u8, nbStripes: uint, f_acc512: XXH3_accumulate_512_f) {
for n := uint(0); n < nbStripes; n += 1 {
#no_bounds_check for n := uint(0); n < nbStripes; n += 1 {
when !XXH_DISABLE_PREFETCH {
in_ptr := &input[n * XXH_STRIPE_LEN]
prefetch(in_ptr, XXH_PREFETCH_DIST)