ggml : sync/merge cmake,riscv,powerpc, add common.cmake (ggml/0)

This commit is contained in:
Georgi Gerganov
2025-03-27 09:12:54 +02:00
parent fc6d343e76
commit 8ca67df291
5 changed files with 822 additions and 404 deletions

View File

@ -287,18 +287,26 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
endif()
endif()
endif()
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64")
elseif ("${CMAKE_SYSTEM_PROCESSOR} " STREQUAL "ppc64le " OR "${CMAKE_SYSTEM_PROCESSOR} " STREQUAL "powerpc ")
message(STATUS "PowerPC detected")
if (GGML_NATIVE)
execute_process(COMMAND bash -c "grep POWER /proc/cpuinfo | head -n 1" OUTPUT_VARIABLE POWER_M)
if (${POWER_M} MATCHES "POWER10")
list(APPEND ARCH_FLAGS -mcpu=power10)
elseif (${POWER_M} MATCHES "POWER9")
list(APPEND ARCH_FLAGS -mcpu=power9)
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64")
file(READ "/proc/cpuinfo" POWER10_M)
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "powerpc")
execute_process(COMMAND bash -c "prtconf |grep 'Implementation' | head -n 1" OUTPUT_VARIABLE POWER10_M)
endif()
string(REGEX MATCHALL "POWER *([0-9]+)" MATCHED_STRING "${POWER10_M}")
string(REGEX REPLACE "POWER *([0-9]+)" "\\1" EXTRACTED_NUMBER "${MATCHED_STRING}")
if (EXTRACTED_NUMBER GREATER_EQUAL 10)
list(APPEND ARCH_FLAGS -mcpu=power10 -mpowerpc64)
elseif (EXTRACTED_NUMBER EQUAL 9)
list(APPEND ARCH_FLAGS -mcpu=power9 -mpowerpc64)
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64le")
list(APPEND ARCH_FLAGS -mcpu=powerpc64le -mtune=native)
else()
list(APPEND ARCH_FLAGS -mcpu=powerpc64 -mtune=native)
list(APPEND ARCH_FLAGS -mcpu=native -mtune=native -mpowerpc64)
endif()
else()
if (GGML_CPU_POWERPC_CPUTYPE)
@ -318,7 +326,11 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "riscv64")
message(STATUS "RISC-V detected")
if (GGML_RVV)
list(APPEND ARCH_FLAGS -march=rv64gcv -mabi=lp64d)
if (GGML_RV_ZFH)
list(APPEND ARCH_FLAGS -march=rv64gcv_zfhmin -DGGML_RV_ZFH -mabi=lp64d)
else()
list(APPEND ARCH_FLAGS -march=rv64gcv -mabi=lp64d)
endif()
endif()
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "s390x")
message(STATUS "s390x detected")

File diff suppressed because it is too large Load Diff

View File

@ -381,6 +381,35 @@ GGML_API void ggml_aligned_free(void * ptr, size_t size);
return r;
}
#elif defined(__riscv) && defined(GGML_RV_ZFH)
static inline float ggml_compute_fp16_to_fp32(ggml_fp16_t h) {
float f;
__asm__(
"fmv.h.x %[f], %[h]\n\t"
"fcvt.s.h %[f], %[f]"
: [f] "=&f" (f)
: [h] "r" (h)
);
return f;
}
static inline ggml_fp16_t ggml_compute_fp32_to_fp16(float f) {
ggml_fp16_t res;
__asm__(
"fcvt.h.s %[f], %[f]\n\t"
"fmv.x.h %[h], %[f]"
: [h] "=&r" (res)
: [f] "f" (f)
);
return res;
}
#define GGML_COMPUTE_FP16_TO_FP32(x) ggml_compute_fp16_to_fp32(x)
#define GGML_COMPUTE_FP32_TO_FP16(x) ggml_compute_fp32_to_fp16(x)
#define GGML_FP16_TO_FP32(x) GGML_COMPUTE_FP16_TO_FP32(x)
#define GGML_FP32_TO_FP16(x) GGML_COMPUTE_FP32_TO_FP16(x)
#else
// FP16 <-> FP32