Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ggml : add predefined list of CPU backend variants to build #10626

Merged
merged 4 commits into from
Dec 4, 2024
Merged

ggml : add predefined list of CPU backend variants to build #10626

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
c4e9c7c
ggml : add predefined list of CPU backend variants to build
slaren Dec 2, 2024
674dec9
fixes
slaren Dec 3, 2024
478194b
update CPU dockerfiles
slaren Dec 4, 2024
52e69b5
Merge remote-tracking branch 'origin/master' into sl/dl-backend-6
slaren Dec 4, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
Next Next commit
fixes
  • Loading branch information
@slaren
slaren committed Dec 3, 2024
commit 674dec9b36ab92986ced89994bbebaf1b2631b37
49 changes: 24 additions & 25 deletions ggml/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -92,30 +92,33 @@ else()
set(INS_ENB ON)
endif()

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_AVX "ggml: enable AVX" ${INS_ENB})
option(GGML_AVX_VNNI "ggml: enable AVX-VNNI" OFF)
option(GGML_AVX2 "ggml: enable AVX2" ${INS_ENB})
option(GGML_AVX512 "ggml: enable AVX512" OFF)
option(GGML_AVX512_VBMI "ggml: enable AVX512-VBMI" OFF)
option(GGML_AVX512_VNNI "ggml: enable AVX512-VNNI" OFF)
option(GGML_AVX512_BF16 "ggml: enable AVX512-BF16" OFF)
option(GGML_AMX_TILE "ggml: enable AMX-TILE" OFF)
option(GGML_AMX_INT8 "ggml: enable AMX-INT8" OFF)
option(GGML_AMX_BF16 "ggml: enable AMX-BF16" OFF)
option(GGML_FMA "ggml: enable FMA" ${INS_ENB})
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_AVX "ggml: enable AVX" ${INS_ENB})
option(GGML_AVX_VNNI "ggml: enable AVX-VNNI" OFF)
option(GGML_AVX2 "ggml: enable AVX2" ${INS_ENB})
option(GGML_AVX512 "ggml: enable AVX512F" OFF)
option(GGML_AVX512_VBMI "ggml: enable AVX512-VBMI" OFF)
option(GGML_AVX512_VNNI "ggml: enable AVX512-VNNI" OFF)
option(GGML_AVX512_BF16 "ggml: enable AVX512-BF16" OFF)
if (NOT MSVC)
option(GGML_F16C "ggml: enable F16C" ${INS_ENB}) # in MSVC F16C is implied with AVX2/AVX512
# in MSVC F16C and FMA is implied with AVX2/AVX512
option(GGML_FMA "ggml: enable FMA" ${INS_ENB})
option(GGML_F16C "ggml: enable F16C" ${INS_ENB})
# MSVC does not seem to support AMX
option(GGML_AMX_TILE "ggml: enable AMX-TILE" OFF)
option(GGML_AMX_INT8 "ggml: enable AMX-INT8" OFF)
option(GGML_AMX_BF16 "ggml: enable AMX-BF16" OFF)
endif()
option(GGML_LASX "ggml: enable lasx" ON)
option(GGML_LSX "ggml: enable lsx" ON)
option(GGML_RVV "ggml: enable rvv" ON)
option(GGML_SVE "ggml: enable SVE" OFF)
option(GGML_LASX "ggml: enable lasx" ON)
option(GGML_LSX "ggml: enable lsx" ON)
option(GGML_RVV "ggml: enable rvv" ON)
option(GGML_SVE "ggml: enable SVE" OFF)
option(GGML_CPU_ALL_VARIANTS "ggml: build all variants of the CPU backend (requires GGML_BACKEND_DL)" OFF)


if (WIN32)
set(GGML_WIN_VER "0x602" CACHE STRING "ggml: Windows Version")
set(GGML_WIN_VER "0x602" CACHE STRING "ggml: Windows version")
endif()

# ggml core
Expand Down Expand Up @@ -180,11 +183,7 @@ option(GGML_BUILD_EXAMPLES "ggml: build examples" ${GGML_STANDALONE})
set(CMAKE_C_STANDARD 11)
set(CMAKE_C_STANDARD_REQUIRED true)

if (GGML_SYCL)
set(CMAKE_CXX_STANDARD 17)
else()
set(CMAKE_CXX_STANDARD 11)
endif()
set(CMAKE_CXX_STANDARD 17)
set(CMAKE_CXX_STANDARD_REQUIRED true)

set(THREADS_PREFER_PTHREAD_FLAG ON)
Expand Down
9 changes: 6 additions & 3 deletions ggml/src/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -294,12 +294,15 @@ if (GGML_CPU_ALL_VARIANTS)
endif()
ggml_add_cpu_backend_variant(sandybridge AVX)
ggml_add_cpu_backend_variant(haswell AVX F16C AVX2 FMA)
ggml_add_cpu_backend_variant(alderlake AVX F16C AVX2 FMA AVX_VNNI)
ggml_add_cpu_backend_variant(skylakex AVX F16C AVX2 FMA AVX512)
ggml_add_cpu_backend_variant(icelake AVX F16C AVX2 FMA AVX512 AVX512_VBMI AVX512_VNNI)
ggml_add_cpu_backend_variant(sapphirerapids AVX F16C AVX2 FMA AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16 AMX_TILE AMX_INT8)
if (NOT MSVC)
# MSVC doesn't support AVX-VNNI or AMX
ggml_add_cpu_backend_variant(alderlake AVX F16C AVX2 FMA AVX_VNNI)
ggml_add_cpu_backend_variant(sapphirerapids AVX F16C AVX2 FMA AVX512 AVX512_VBMI AVX512_VNNI AVX512_BF16 AMX_TILE AMX_INT8)
endif()
else ()
ggml_add_cpu_backend_variant_impl()
ggml_add_cpu_backend_variant_impl("")
endif()

ggml_add_backend(BLAS)
Expand Down
25 changes: 12 additions & 13 deletions ggml/src/ggml-cpu/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -179,7 +179,7 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
if (MSVC)
# instruction set detection for MSVC only
if (GGML_NATIVE)
include(cmake/FindSIMD.cmake)
include(ggml-cpu/cmake/FindSIMD.cmake)
endif ()
if (GGML_AVX512)
list(APPEND ARCH_FLAGS /arch:AVX512)
Expand Down Expand Up @@ -227,12 +227,10 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
list(APPEND ARCH_DEFINITIONS GGML_SSE42)
endif()
if (GGML_AVX_VNNI)
list(APPEND ARCH_DEFINITIONS __AVXVNNI__ GGML_AVX_VNNI)
if (CMAKE_C_COMPILER_ID STREQUAL "Clang")
list(APPEND ARCH_FLAGS -mavxvnni)
endif()
# MSVC generates AVX512 with AVX-VNNI intrinsics even with /arch:AVX2
#list(APPEND ARCH_DEFINITIONS __AVXVNNI__ GGML_AVX_VNNI)
endif()
else()
else ()
if (GGML_NATIVE)
list(APPEND ARCH_FLAGS -march=native)
else ()
Expand Down Expand Up @@ -340,13 +338,14 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
# The feature detection code is compiled as a separate target so that
# it can be built without the architecture flags
# Since multiple variants of the CPU backend may be included in the same
# build, using set_source_files_properties() is not possible
add_library(${GGML_CPU_NAME}_feats OBJECT ggml-cpu/cpu-feats-x86.cpp)
target_include_directories(${GGML_CPU_NAME}_feats PRIVATE . .. ../include)
target_compile_definitions(${GGML_CPU_NAME}_feats PRIVATE ${ARCH_DEFINITIONS})
target_compile_definitions(${GGML_CPU_NAME}_feats PRIVATE GGML_BACKEND_DL GGML_BACKEND_BUILD GGML_BACKEND_SHARED)
set_target_properties(${GGML_CPU_NAME}_feats PROPERTIES POSITION_INDEPENDENT_CODE ON)
target_link_libraries(${GGML_CPU_NAME} PRIVATE ${GGML_CPU_NAME}_feats)
# build, using set_source_files_properties() to set the arch flags is not possible
set(GGML_CPU_FEATS_NAME ${GGML_CPU_NAME}-feats)
add_library(${GGML_CPU_FEATS_NAME} OBJECT ggml-cpu/cpu-feats-x86.cpp)
target_include_directories(${GGML_CPU_FEATS_NAME} PRIVATE . .. ../include)
target_compile_definitions(${GGML_CPU_FEATS_NAME} PRIVATE ${ARCH_DEFINITIONS})
target_compile_definitions(${GGML_CPU_FEATS_NAME} PRIVATE GGML_BACKEND_DL GGML_BACKEND_BUILD GGML_BACKEND_SHARED)
set_target_properties(${GGML_CPU_FEATS_NAME} PROPERTIES POSITION_INDEPENDENT_CODE ON)
target_link_libraries(${GGML_CPU_NAME} PRIVATE ${GGML_CPU_FEATS_NAME})
endif()

if (EMSCRIPTEN)
Expand Down
Loading