vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

fix debug link error. fix windows crash

Browse source
This commit is contained in:
luoyu-intel 2024-06-19 07:43:51 +00:00
parent 079dd3f592
commit 4488134edf
5 changed files with 817 additions and 820 deletions
Download patch file Download diff file Expand all files Collapse all files

7
CMakeLists.txt
View file

@ -665,6 +665,7 @@ if (LLAMA_SYCL)
#todo: AOT #todo: AOT
find_package(IntelSYCL REQUIRED) find_package(IntelSYCL REQUIRED)
find_package(MKL REQUIRED)
message(STATUS "SYCL found") message(STATUS "SYCL found")
@ -679,11 +680,9 @@ if (LLAMA_SYCL)
endif() endif()
add_compile_options(-I./) #include DPCT add_compile_options(-I./) #include DPCT
add_compile_options(-I/${SYCL_INCLUDE_DIR})
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-narrowing") set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-narrowing")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3") set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O3")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl -L${MKLROOT}/lib")
if (LLAMA_SYCL_TARGET STREQUAL "NVIDIA") if (LLAMA_SYCL_TARGET STREQUAL "NVIDIA")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=nvptx64-nvidia-cuda") set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=nvptx64-nvidia-cuda")
endif() endif()
@ -693,8 +692,10 @@ if (LLAMA_SYCL)
list(APPEND GGML_SOURCES_SYCL "ggml-sycl.cpp") list(APPEND GGML_SOURCES_SYCL "ggml-sycl.cpp")
if (WIN32) if (WIN32)
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} -fsycl sycl7 OpenCL mkl_sycl_blas_dll.lib mkl_intel_ilp64_dll.lib mkl_sequential_dll.lib mkl_core_dll.lib) set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} IntelSYCL::SYCL_CXX MKL::MKL MKL::MKL_SYCL)
else() else()
add_compile_options(-I/${SYCL_INCLUDE_DIR})
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl -L${MKLROOT}/lib")
if (LLAMA_SYCL_TARGET STREQUAL "INTEL") if (LLAMA_SYCL_TARGET STREQUAL "INTEL")
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} -fsycl OpenCL mkl_core pthread m dl mkl_sycl_blas mkl_intel_ilp64 mkl_tbb_thread) set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} -fsycl OpenCL mkl_core pthread m dl mkl_sycl_blas mkl_intel_ilp64 mkl_tbb_thread)
elseif (LLAMA_SYCL_TARGET STREQUAL "NVIDIA") elseif (LLAMA_SYCL_TARGET STREQUAL "NVIDIA")

1
CMakePresets.json
View file

@ -19,7 +19,6 @@
"cacheVariables": { "cacheVariables": {
"CMAKE_EXPORT_COMPILE_COMMANDS": "ON", "CMAKE_EXPORT_COMPILE_COMMANDS": "ON",
"CMAKE_CXX_COMPILER": "icx", "CMAKE_CXX_COMPILER": "icx",
"CMAKE_C_COMPILER": "icx",
"LLAMA_SYCL": "ON", "LLAMA_SYCL": "ON",
"CMAKE_INSTALL_RPATH": "$ORIGIN;$ORIGIN/.." "CMAKE_INSTALL_RPATH": "$ORIGIN;$ORIGIN/.."
} }

2
ggml-sycl.cpp
View file

@ -4911,7 +4911,7 @@ static void ggml_sycl_cpy(ggml_backend_sycl_context & ctx, const ggml_tensor *sr
GGML_ASSERT(ggml_nbytes(src0) <= INT_MAX); GGML_ASSERT(ggml_nbytes(src0) <= INT_MAX);
GGML_ASSERT(ggml_nbytes(src1) <= INT_MAX); GGML_ASSERT(ggml_nbytes(src1) <= INT_MAX);
GGML_TENSOR_BINARY_OP_LOCALS; GGML_TENSOR_BINARY_OP_LOCALS01;
SYCL_CHECK(ggml_sycl_set_device(ctx.device)); SYCL_CHECK(ggml_sycl_set_device(ctx.device));
queue_ptr main_stream = ctx.stream(); queue_ptr main_stream = ctx.stream();

1621
ggml-sycl/dpct/helper.hpp
View file

File diff suppressed because it is too large Load diff

6
ggml.h
View file

@ -312,6 +312,12 @@
GGML_TENSOR_LOCALS(int64_t, ne, dst, ne) \ GGML_TENSOR_LOCALS(int64_t, ne, dst, ne) \
GGML_TENSOR_LOCALS(size_t, nb, dst, nb) GGML_TENSOR_LOCALS(size_t, nb, dst, nb)
#define GGML_TENSOR_BINARY_OP_LOCALS01 \
GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne) \
GGML_TENSOR_LOCALS(size_t, nb0, src0, nb) \
GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne) \
GGML_TENSOR_LOCALS(size_t, nb1, src1, nb)
#ifdef __cplusplus #ifdef __cplusplus
extern "C" { extern "C" {
#endif #endif