extern/whisper.cpp
1
0
Fork 1
mirror of https://github.com/ggerganov/whisper.cpp.git synced 2025-06-19 17:28:09 +02:00
Code Issues Packages Projects Releases Wiki Activity

ggml : add SSE 4.2 and x64 base variant for CPUs without AVX (llama/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 Georgi Gerganov
parent 33c89ade7d
commit 11218294db
4 changed files with 16 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
ggml/CMakeLists.txt
View File

@ -107,6 +107,7 @@ message(DEBUG "INS_ENB : ${INS_ENB}")
option(GGML_CPU_HBM "ggml: use memkind for CPU HBM" OFF) option(GGML_CPU_HBM "ggml: use memkind for CPU HBM" OFF)
option(GGML_CPU_AARCH64 "ggml: use runtime weight conversion of Q4_0 to Q4_X_X" ON) option(GGML_CPU_AARCH64 "ggml: use runtime weight conversion of Q4_0 to Q4_X_X" ON)
option(GGML_CPU_KLEIDIAI "ggml: use KleidiAI optimized kernels if applicable" OFF) option(GGML_CPU_KLEIDIAI "ggml: use KleidiAI optimized kernels if applicable" OFF)
option(GGML_SSE42 "ggml: enable SSE 4.2" ${INS_ENB})
option(GGML_AVX "ggml: enable AVX" ${INS_ENB}) option(GGML_AVX "ggml: enable AVX" ${INS_ENB})
option(GGML_AVX_VNNI "ggml: enable AVX-VNNI" OFF) option(GGML_AVX_VNNI "ggml: enable AVX-VNNI" OFF)
option(GGML_AVX2 "ggml: enable AVX2" ${INS_ENB}) option(GGML_AVX2 "ggml: enable AVX2" ${INS_ENB})

15
ggml/src/CMakeLists.txt
View File

@ -267,6 +267,7 @@ function(ggml_add_cpu_backend_variant tag_name)
set(GGML_CPU_TAG_NAME ${tag_name}) set(GGML_CPU_TAG_NAME ${tag_name})
# other: OPENMP LLAMAFILE CPU_HBM # other: OPENMP LLAMAFILE CPU_HBM
foreach (feat NATIVE foreach (feat NATIVE
SSE42
AVX AVX2 BMI2 AVX_VNNI FMA F16C AVX AVX2 BMI2 AVX_VNNI FMA F16C
AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16 AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16
AMX_TILE AMX_INT8 AMX_BF16) AMX_TILE AMX_INT8 AMX_BF16)
@ -286,14 +287,16 @@ if (GGML_CPU_ALL_VARIANTS)
if (NOT GGML_BACKEND_DL) if (NOT GGML_BACKEND_DL)
message(FATAL_ERROR "GGML_CPU_ALL_VARIANTS requires GGML_BACKEND_DL") message(FATAL_ERROR "GGML_CPU_ALL_VARIANTS requires GGML_BACKEND_DL")
endif() endif()
ggml_add_cpu_backend_variant(sandybridge AVX) ggml_add_cpu_backend_variant(x64)
ggml_add_cpu_backend_variant(haswell AVX F16C AVX2 BMI2 FMA) ggml_add_cpu_backend_variant(sse42 SSE42)
ggml_add_cpu_backend_variant(skylakex AVX F16C AVX2 BMI2 FMA AVX512) ggml_add_cpu_backend_variant(sandybridge SSE42 AVX)
ggml_add_cpu_backend_variant(icelake AVX F16C AVX2 BMI2 FMA AVX512 AVX512_VBMI AVX512_VNNI) ggml_add_cpu_backend_variant(haswell SSE42 AVX F16C AVX2 BMI2 FMA)
ggml_add_cpu_backend_variant(alderlake AVX F16C AVX2 BMI2 FMA AVX_VNNI) ggml_add_cpu_backend_variant(skylakex SSE42 AVX F16C AVX2 BMI2 FMA AVX512)
ggml_add_cpu_backend_variant(icelake SSE42 AVX F16C AVX2 BMI2 FMA AVX512 AVX512_VBMI AVX512_VNNI)
ggml_add_cpu_backend_variant(alderlake SSE42 AVX F16C AVX2 BMI2 FMA AVX_VNNI)
if (NOT MSVC) if (NOT MSVC)
# MSVC doesn't support AMX # MSVC doesn't support AMX
ggml_add_cpu_backend_variant(sapphirerapids AVX F16C AVX2 BMI2 FMA AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16 AMX_TILE AMX_INT8) ggml_add_cpu_backend_variant(sapphirerapids SSE42 AVX F16C AVX2 BMI2 FMA AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16 AMX_TILE AMX_INT8)
endif() endif()
elseif (GGML_CPU) elseif (GGML_CPU)
ggml_add_cpu_backend_variant_impl("") ggml_add_cpu_backend_variant_impl("")

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) elseif (GGML_AVX)
list(APPEND ARCH_FLAGS /arch:AVX) list(APPEND ARCH_FLAGS /arch:AVX)
list(APPEND ARCH_DEFINITIONS GGML_AVX) list(APPEND ARCH_DEFINITIONS GGML_AVX)
else () elseif (GGML_SSE42)
list(APPEND ARCH_FLAGS /arch:SSE4.2) list(APPEND ARCH_FLAGS /arch:SSE4.2)
list(APPEND ARCH_DEFINITIONS GGML_SSE42) list(APPEND ARCH_DEFINITIONS GGML_SSE42)
endif() endif()
@ -237,8 +237,10 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
if (GGML_NATIVE) if (GGML_NATIVE)
list(APPEND ARCH_FLAGS -march=native) list(APPEND ARCH_FLAGS -march=native)
else () else ()
list(APPEND ARCH_FLAGS -msse4.2) if (GGML_SSE42)
list(APPEND ARCH_DEFINITIONS GGML_SSE42) list(APPEND ARCH_FLAGS -msse4.2)
list(APPEND ARCH_DEFINITIONS GGML_SSE42)
endif()
if (GGML_F16C) if (GGML_F16C)
list(APPEND ARCH_FLAGS -mf16c) list(APPEND ARCH_FLAGS -mf16c)
list(APPEND ARCH_DEFINITIONS GGML_F16C) 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() { static int ggml_backend_cpu_x86_score() {
// FIXME: this does not check for OS support // FIXME: this does not check for OS support
int score = 0; int score = 1;
cpuid_x86 is; cpuid_x86 is;
#ifdef GGML_FMA #ifdef GGML_FMA