ver4a/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

ggml : add SSE 4.2 and x64 base variant for CPUs without AVX (#12871)

Browse source 
* ggml : add SSE 4.2 variant for CPUs without AVX

* ggml : add x64 base ABI variant
This commit is contained in:
Diego Devesa 2025-04-21 18:13:51 +02:00 • committed by GitHub
parent 5368ddda7a
commit 1d735c0b4f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
4 changed files with 16 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

8
ggml/src/ggml-cpu/CMakeLists.txt
View file

@ -222,7 +222,7 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
elseif (GGML_AVX)
list(APPEND ARCH_FLAGS /arch:AVX)
list(APPEND ARCH_DEFINITIONS GGML_AVX)
else ()
elseif (GGML_SSE42)
list(APPEND ARCH_FLAGS /arch:SSE4.2)
list(APPEND ARCH_DEFINITIONS GGML_SSE42)
endif()
@ -237,8 +237,10 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
if (GGML_NATIVE)
list(APPEND ARCH_FLAGS -march=native)
else ()
list(APPEND ARCH_FLAGS -msse4.2)
list(APPEND ARCH_DEFINITIONS GGML_SSE42)
if (GGML_SSE42)
list(APPEND ARCH_FLAGS -msse4.2)
list(APPEND ARCH_DEFINITIONS GGML_SSE42)
endif()
if (GGML_F16C)
list(APPEND ARCH_FLAGS -mf16c)
list(APPEND ARCH_DEFINITIONS GGML_F16C)

2
ggml/src/ggml-cpu/cpu-feats-x86.cpp
View file

@ -263,7 +263,7 @@ void test_x86_is() {
static int ggml_backend_cpu_x86_score() {
// FIXME: this does not check for OS support
int score = 0;
int score = 1;
cpuid_x86 is;
#ifdef GGML_FMA