Skip to content

cmake : fix ARM feature verification #17170

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

Open
wants to merge 3 commits into
base: master
Choose a base branch
from
+30 −36
Open

cmake : fix ARM feature verification #17170

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
bd13d06
cmake : fix ARM feature verification
angt Nov 11, 2025
e3604f6
cmake : unset __ARM_FEATURE when feature is disabled
angt Nov 11, 2025
01ac1a6
cmake : fix scope, this is really a macro
angt Nov 11, 2025
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
66 changes: 30 additions & 36 deletions ggml/src/ggml-cpu/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -145,26 +145,27 @@ function(ggml_add_cpu_backend_variant_impl tag_name)

include(CheckCXXSourceRuns)

function(check_arm_feature tag code)
macro(check_arm_feature tag feature code)
set(CMAKE_REQUIRED_FLAGS_SAVE ${CMAKE_REQUIRED_FLAGS})
set(CMAKE_REQUIRED_FLAGS "${ARM_NATIVE_FLAG}+${tag}")
check_cxx_source_runs("${code}" GGML_MACHINE_SUPPORTS_${tag})
if (GGML_MACHINE_SUPPORTS_${tag})
set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+${tag}" PARENT_SCOPE)
set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+${tag}")
else()
set(CMAKE_REQUIRED_FLAGS "${ARM_NATIVE_FLAG}+no${tag}")
check_cxx_source_compiles("int main() { return 0; }" GGML_MACHINE_SUPPORTS_no${tag})
if (GGML_MACHINE_SUPPORTS_no${tag})
set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+no${tag}" PARENT_SCOPE)
set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+no${tag}")
list(APPEND ARCH_FLAGS -U__ARM_FEATURE_${feature})
endif()
endif()
set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_SAVE})
endfunction()
endmacro()

check_arm_feature(dotprod "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vdotq_s32(_s, _a, _b); return 0; }")
check_arm_feature(i8mm "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vmmlaq_s32(_s, _a, _b); return 0; }")
check_arm_feature(sve "#include <arm_sve.h>\nint main() { svfloat32_t _a, _b; volatile svfloat32_t _c = svadd_f32_z(svptrue_b8(), _a, _b); return 0; }")
check_arm_feature(sme "#include <arm_sme.h>\n__arm_locally_streaming int main() { __asm__ volatile(\"smstart; smstop;\"); return 0; }")
check_arm_feature(dotprod DOTPROD "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vdotq_s32(_s, _a, _b); return 0; }")
check_arm_feature(i8mm MATMUL_INT8 "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vmmlaq_s32(_s, _a, _b); return 0; }")
check_arm_feature(sve SVE "#include <arm_sve.h>\nint main() { svfloat32_t _a, _b; volatile svfloat32_t _c = svadd_f32_z(svptrue_b8(), _a, _b); return 0; }")
check_arm_feature(sme SME "#include <arm_sme.h>\n__arm_locally_streaming int main() { __asm__ volatile(\"smstart; smstop;\"); return 0; }")

list(APPEND ARCH_FLAGS "${ARM_NATIVE_FLAG}${ARM_NATIVE_FLAG_FIX}")
else()
Expand Down Expand Up @@ -216,35 +217,28 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
endif()
endif()

# show enabled features
if (CMAKE_HOST_SYSTEM_NAME STREQUAL "Windows")
set(FEAT_INPUT_FILE "NUL")
else()
set(FEAT_INPUT_FILE "/dev/null")
endif()
message(STATUS "Checking for ARM features using flags:")
foreach(flag IN LISTS ARCH_FLAGS)
message(STATUS " ${flag}")
endforeach()

execute_process(
COMMAND ${CMAKE_C_COMPILER} ${ARCH_FLAGS} -dM -E -
INPUT_FILE ${FEAT_INPUT_FILE}
OUTPUT_VARIABLE ARM_FEATURE
RESULT_VARIABLE ARM_FEATURE_RESULT
)
if (ARM_FEATURE_RESULT)
message(WARNING "Failed to get ARM features")
else()
foreach(feature DOTPROD SVE MATMUL_INT8 FMA FP16_VECTOR_ARITHMETIC SME)
string(FIND "${ARM_FEATURE}" "__ARM_FEATURE_${feature} 1" feature_pos)
if (NOT ${feature_pos} EQUAL -1)
# Special handling for MATMUL_INT8 when machine doesn't support i8mm
if ("${feature}" STREQUAL "MATMUL_INT8" AND GGML_MACHINE_SUPPORTS_noi8mm)
message(STATUS "ARM feature ${feature} detected but unsetting due to machine not supporting i8mm")
list(APPEND ARCH_FLAGS -U__ARM_FEATURE_MATMUL_INT8)
else()
message(STATUS "ARM feature ${feature} enabled")
endif()
endif()
endforeach()
endif()
include(CheckCXXSourceCompiles)
set(CMAKE_REQUIRED_FLAGS_SAVE ${CMAKE_REQUIRED_FLAGS})
set(CMAKE_REQUIRED_FLAGS "${ARCH_FLAGS}")
foreach(feature DOTPROD SVE MATMUL_INT8 FMA FP16_VECTOR_ARITHMETIC SME)
set(ARM_FEATURE "HAVE_${feature}")
check_cxx_source_compiles(
"
#include <arm_neon.h>
#if !defined(__ARM_FEATURE_${feature})
# error \"Feature ${feature} is not defined\"
#endif
int main() { return 0; }
"
${ARM_FEATURE}
)
endforeach()
set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_SAVE})
endif()
elseif (GGML_SYSTEM_ARCH STREQUAL "x86")
message(STATUS "x86 detected")
Expand Down
Loading