Skip to content

Commit 017efe8

Browse files
netrunnerevenetrunnereve
authored
cmake : make LLAMA_NATIVE flag actually use the instructions supported by the processor (#3273)
* fix LLAMA_NATIVE * syntax * alternate implementation * my eyes must be getting bad... * set cmake LLAMA_NATIVE=ON by default * march=native doesn't work for ios/tvos, so disable for those targets. also see what happens if we use it on msvc * revert 8283237 and only allow LLAMA_NATIVE on x86 like the Makefile * remove -DLLAMA_MPI=ON --------- Co-authored-by: netrunnereve <[email protected]>
1 parent ff5a3f0 commit 017efe8

File tree

3 files changed

+26
-20
lines changed

3 files changed

+26
-20
lines changed

‎.github/workflows/build.yml‎

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -188,7 +188,7 @@ jobs:
188188
sysctl -a
189189
mkdir build
190190
cd build
191-
cmake -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF ..
191+
cmake ..
192192
cmake --build . --config Release -j $(sysctl -n hw.logicalcpu)
193193
194194
- name: Test
@@ -265,17 +265,17 @@ jobs:
265265
matrix:
266266
include:
267267
- build: 'noavx'
268-
defines: '-DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX=OFF -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF -DBUILD_SHARED_LIBS=ON'
268+
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX=OFF -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF -DBUILD_SHARED_LIBS=ON'
269269
- build: 'avx2'
270-
defines: '-DLLAMA_BUILD_SERVER=ON -DBUILD_SHARED_LIBS=ON'
270+
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DBUILD_SHARED_LIBS=ON'
271271
- build: 'avx'
272-
defines: '-DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX2=OFF -DBUILD_SHARED_LIBS=ON'
272+
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX2=OFF -DBUILD_SHARED_LIBS=ON'
273273
- build: 'avx512'
274-
defines: '-DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX512=ON -DBUILD_SHARED_LIBS=ON'
274+
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_AVX512=ON -DBUILD_SHARED_LIBS=ON'
275275
- build: 'clblast'
276-
defines: '-DLLAMA_BUILD_SERVER=ON -DLLAMA_CLBLAST=ON -DBUILD_SHARED_LIBS=ON -DCMAKE_PREFIX_PATH="$env:RUNNER_TEMP/clblast"'
276+
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_CLBLAST=ON -DBUILD_SHARED_LIBS=ON -DCMAKE_PREFIX_PATH="$env:RUNNER_TEMP/clblast"'
277277
- build: 'openblas'
278-
defines: '-DLLAMA_BUILD_SERVER=ON -DLLAMA_BLAS=ON -DBUILD_SHARED_LIBS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS -DBLAS_INCLUDE_DIRS="$env:RUNNER_TEMP/openblas/include" -DBLAS_LIBRARIES="$env:RUNNER_TEMP/openblas/lib/openblas.lib"'
278+
defines: '-DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_BLAS=ON -DBUILD_SHARED_LIBS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS -DBLAS_INCLUDE_DIRS="$env:RUNNER_TEMP/openblas/include" -DBLAS_LIBRARIES="$env:RUNNER_TEMP/openblas/lib/openblas.lib"'
279279

280280
steps:
281281
- name: Clone
@@ -414,7 +414,7 @@ jobs:
414414
run: |
415415
mkdir build
416416
cd build
417-
cmake .. -DLLAMA_BUILD_SERVER=ON -DLLAMA_CUBLAS=ON -DBUILD_SHARED_LIBS=ON
417+
cmake .. -DLLAMA_NATIVE=OFF -DLLAMA_BUILD_SERVER=ON -DLLAMA_CUBLAS=ON -DBUILD_SHARED_LIBS=ON
418418
cmake --build . --config Release -j ${env:NUMBER_OF_PROCESSORS}
419419
420420
- name: Determine tag name

‎CMakeLists.txt‎

Lines changed: 17 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -44,7 +44,7 @@ endif()
4444

4545
# general
4646
option(LLAMA_STATIC "llama: static link libraries" OFF)
47-
option(LLAMA_NATIVE "llama: enable -march=native flag" OFF)
47+
option(LLAMA_NATIVE "llama: enable -march=native flag" ON)
4848
option(LLAMA_LTO "llama: enable link time optimization" OFF)
4949

5050
# debug
@@ -58,15 +58,21 @@ option(LLAMA_SANITIZE_ADDRESS "llama: enable address sanitizer"
5858
option(LLAMA_SANITIZE_UNDEFINED "llama: enable undefined sanitizer" OFF)
5959

6060
# instruction set specific
61-
option(LLAMA_AVX "llama: enable AVX" ON)
62-
option(LLAMA_AVX2 "llama: enable AVX2" ON)
63-
option(LLAMA_AVX512 "llama: enable AVX512" OFF)
64-
option(LLAMA_AVX512_VBMI "llama: enable AVX512-VBMI" OFF)
65-
option(LLAMA_AVX512_VNNI "llama: enable AVX512-VNNI" OFF)
66-
option(LLAMA_FMA "llama: enable FMA" ON)
61+
if (LLAMA_NATIVE)
62+
set(INS_ENB OFF)
63+
else()
64+
set(INS_ENB ON)
65+
endif()
66+
67+
option(LLAMA_AVX "llama: enable AVX" ${INS_ENB})
68+
option(LLAMA_AVX2 "llama: enable AVX2" ${INS_ENB})
69+
option(LLAMA_AVX512 "llama: enable AVX512" OFF)
70+
option(LLAMA_AVX512_VBMI "llama: enable AVX512-VBMI" OFF)
71+
option(LLAMA_AVX512_VNNI "llama: enable AVX512-VNNI" OFF)
72+
option(LLAMA_FMA "llama: enable FMA" ${INS_ENB})
6773
# in MSVC F16C is implied with AVX2/AVX512
6874
if (NOT MSVC)
69-
option(LLAMA_F16C "llama: enable F16C" ON)
75+
option(LLAMA_F16C "llama: enable F16C" ${INS_ENB})
7076
endif()
7177

7278
# 3rd party libs
@@ -504,9 +510,6 @@ if (NOT MSVC)
504510
if (LLAMA_GPROF)
505511
add_compile_options(-pg)
506512
endif()
507-
if (LLAMA_NATIVE)
508-
add_compile_options(-march=native)
509-
endif()
510513
endif()
511514

512515
if ((${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm") OR (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64") OR ("${CMAKE_GENERATOR_PLATFORM_LWR}" MATCHES "arm64"))
@@ -561,6 +564,9 @@ elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$" OR "${CMAKE_GE
561564
add_compile_options($<$<COMPILE_LANGUAGE:CXX>:/arch:AVX>)
562565
endif()
563566
else()
567+
if (LLAMA_NATIVE)
568+
add_compile_options(-march=native)
569+
endif()
564570
if (LLAMA_F16C)
565571
add_compile_options(-mf16c)
566572
endif()

‎flake.nix‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -62,7 +62,7 @@
6262
mkdir -p $out/include
6363
cp ${src}/llama.h $out/include/
6464
'';
65-
cmakeFlags = [ "-DLLAMA_BUILD_SERVER=ON" "-DLLAMA_MPI=ON" "-DBUILD_SHARED_LIBS=ON" "-DCMAKE_SKIP_BUILD_RPATH=ON" ];
65+
cmakeFlags = [ "-DLLAMA_NATIVE=OFF" "-DLLAMA_BUILD_SERVER=ON" "-DBUILD_SHARED_LIBS=ON" "-DCMAKE_SKIP_BUILD_RPATH=ON" ];
6666
in
6767
{
6868
packages.default = pkgs.stdenv.mkDerivation {

0 commit comments

Comments
 (0)