mirror of
https://github.com/LostRuins/koboldcpp.git
synced 2026-04-28 03:30:20 +00:00
7ca5991d2b
* Faster tensors (#8) Add fast matrix and matrix/vector multiplication. * Use map for shader replacements instead of pair of strings * Wasm (#9) * webgpu : fix build on emscripten * more debugging stuff * test-backend-ops: force single thread on wasm * fix single-thread case for init_tensor_uniform * use jspi * add pthread * test: remember to set n_thread for cpu backend * Add buffer label and enable dawn-specific toggles to turn off some checks * Intermediate state * Fast working f16/f32 vec4 * Working float fast mul mat * Clean up naming of mul_mat to match logical model, start work on q mul_mat * Setup for subgroup matrix mat mul * Basic working subgroup matrix * Working subgroup matrix tiling * Handle weirder sg matrix sizes (but still % sg matrix size) * Working start to gemv * working f16 accumulation with shared memory staging * Print out available subgroup matrix configurations * Vectorize dst stores for sg matrix shader * Gemv working scalar * Minor set_rows optimization (#4) * updated optimization, fixed errors * non vectorized version now dispatches one thread per element * Simplify * Change logic for set_rows pipelines --------- Co-authored-by: Neha Abbas <nehaabbas@macbookpro.lan> Co-authored-by: Neha Abbas <nehaabbas@ReeseLevines-MacBook-Pro.local> Co-authored-by: Reese Levine <reeselevine1@gmail.com> * Comment on dawn toggles * Working subgroup matrix code for (semi)generic sizes * Remove some comments * Cleanup code * Update dawn version and move to portable subgroup size * Try to fix new dawn release * Update subgroup size comment * Only check for subgroup matrix configs if they are supported * Add toggles for subgroup matrix/f16 support on nvidia+vulkan * Make row/col naming consistent * Refactor shared memory loading * Move sg matrix stores to correct file * Working q4_0 * Formatting * Work with emscripten builds * Fix test-backend-ops emscripten for f16/quantized types * Use emscripten memory64 to support get_memory * Add build flags and try ci --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co> * Remove extra whitespace * Move wasm single-thread logic out of test-backend-ops for cpu backend * Disable multiple threads for emscripten single-thread builds in ggml_graph_plan * Fix .gitignore * Add memory64 option and remove unneeded macros for setting threads to 1 --------- Co-authored-by: Xuan Son Nguyen <son@huggingface.co>
540 lines
24 KiB
CMake
540 lines
24 KiB
CMake
# THIS FILE IS ONLY INTENDED CUBLAS BUILD PURPOSES ON WINDOWS VISUAL STUDIO.
|
|
# YOU'RE NOT RECOMMENDED TO USE IT
|
|
|
|
message(STATUS "============== ============== ==============")
|
|
message(STATUS "WARNING! Recommend NOT to use this file. It is UNSUPPORTED for normal users. Use MAKE instead.")
|
|
message(STATUS "It is ONLY for CUBLAS builds on windows visual studio. IT WILL OVERWRITE YOUR EXISTING MAKEFILE !!!")
|
|
message(STATUS "IF YOU ARE SEEING THIS, you MUST ONLY be building CUBLAS BUILDS! NOTHING ELSE WILL BE SUPPORTED !!!")
|
|
message(STATUS "============== ============== ==============")
|
|
|
|
cmake_minimum_required(VERSION 3.18) # Don't bump this version for no reason
|
|
project("llama.cpp" C CXX)
|
|
|
|
set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
|
|
set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS 1)
|
|
set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
|
|
set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Release")
|
|
set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
|
|
set(LLAMA_STANDALONE ON)
|
|
set(BUILD_SHARED_LIBS_DEFAULT ON)
|
|
set(LLAMA_STATIC OFF)
|
|
set(LLAMA_NATIVE OFF)
|
|
set(LLAMA_LTO OFF)
|
|
set(LLAMA_GPROF OFF)
|
|
|
|
|
|
# instruction set specific
|
|
option(LLAMA_AVX "llama: enable AVX" ON)
|
|
option(LLAMA_AVX2 "llama: enable AVX2" ON)
|
|
option(LLAMA_AVX512 "llama: enable AVX512" OFF)
|
|
option(LLAMA_AVX512_VBMI "llama: enable AVX512-VBMI" OFF)
|
|
option(LLAMA_AVX512_VNNI "llama: enable AVX512-VNNI" OFF)
|
|
option(LLAMA_FMA "llama: enable FMA" ON)
|
|
# in MSVC F16C is implied with AVX2/AVX512
|
|
if (NOT MSVC)
|
|
option(LLAMA_F16C "llama: enable F16C" ON)
|
|
endif()
|
|
|
|
# 3rd party libs
|
|
option(LLAMA_CUBLAS "llama: use CUDA" ON)
|
|
set(LLAMA_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
|
|
"llama: max. batch size for using peer access")
|
|
|
|
option(LLAMA_HIPBLAS "llama: use hipBLAS" OFF)
|
|
|
|
# Other
|
|
option(LLAMA_OPENMP "llama: use OpenMP" OFF)
|
|
|
|
#
|
|
# Compile flags
|
|
#
|
|
|
|
set(CMAKE_CXX_STANDARD 17)
|
|
set(CMAKE_CXX_STANDARD_REQUIRED true)
|
|
set(CMAKE_C_STANDARD 11)
|
|
set(CMAKE_C_STANDARD_REQUIRED true)
|
|
set(THREADS_PREFER_PTHREAD_FLAG ON)
|
|
find_package(Threads REQUIRED)
|
|
|
|
add_compile_definitions(LOG_DISABLE_LOGS)
|
|
add_compile_definitions(GGML_USE_CPU)
|
|
add_compile_definitions(GGML_USE_CPU_REPACK)
|
|
add_compile_definitions(NOMINMAX)
|
|
add_compile_definitions(_REGEX_MAX_STACK_COUNT=5000)
|
|
|
|
if (GGML_HIP_FORCE_ROCWMMA_FATTN_GFX12)
|
|
add_compile_definitions(GGML_HIP_ROCWMMA_FATTN_GFX12)
|
|
endif()
|
|
|
|
if (MSVC)
|
|
add_compile_options("$<$<COMPILE_LANGUAGE:C>:/utf-8>")
|
|
add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:/utf-8>")
|
|
add_compile_options("$<$<COMPILE_LANGUAGE:C>:/bigobj>")
|
|
add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:/bigobj>")
|
|
endif()
|
|
|
|
file(GLOB GGML_SOURCES_CUDA "ggml/src/ggml-cuda/*.cu")
|
|
list(APPEND GGML_SOURCES_CUDA "ggml/src/ggml-cuda/ggml-cuda.cu")
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-tile*.cu")
|
|
list(APPEND GGML_SOURCES_CUDA ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-mma*.cu")
|
|
list(APPEND GGML_SOURCES_CUDA ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/mmq*.cu")
|
|
list(APPEND GGML_SOURCES_CUDA ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/mmf*.cu")
|
|
list(APPEND GGML_SOURCES_CUDA ${SRCS})
|
|
set(GGML_V3_CUDA_SOURCES otherarch/ggml_v3-cuda.cu otherarch/ggml_v3-cuda.h)
|
|
set(GGML_V2_CUDA_SOURCES otherarch/ggml_v2-cuda.cu otherarch/ggml_v2-cuda.h)
|
|
set(GGML_V2_LEGACY_CUDA_SOURCES otherarch/ggml_v2-cuda-legacy.cu otherarch/ggml_v2-cuda-legacy.h)
|
|
|
|
|
|
if (LLAMA_CUBLAS)
|
|
cmake_minimum_required(VERSION 3.17)
|
|
|
|
find_package(CUDAToolkit)
|
|
if (CUDAToolkit_FOUND)
|
|
message(STATUS "cuBLAS found")
|
|
|
|
enable_language(CUDA)
|
|
|
|
add_compile_definitions(GGML_USE_LLAMAFILE)
|
|
add_compile_definitions(GGML_USE_CUDA)
|
|
add_compile_definitions(SD_USE_CUDA)
|
|
|
|
add_compile_definitions(GGML_CUDA_PEER_MAX_BATCH_SIZE=${LLAMA_CUDA_PEER_MAX_BATCH_SIZE})
|
|
|
|
# only build minimal quants required for fattn quant kv
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-vec*q4_0-q4_0.cu")
|
|
list(APPEND GGML_SOURCES_CUDA ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-vec*q8_0-q8_0.cu")
|
|
list(APPEND GGML_SOURCES_CUDA ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-vec*f16-f16.cu")
|
|
list(APPEND GGML_SOURCES_CUDA ${SRCS})
|
|
|
|
if (LLAMA_STATIC)
|
|
if (WIN32)
|
|
# As of 12.3.1 CUDA Tookit for Windows does not offer a static cublas library
|
|
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} CUDA::cudart_static CUDA::cublas CUDA::cublasLt)
|
|
else ()
|
|
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} CUDA::cudart_static CUDA::cublas_static CUDA::cublasLt_static)
|
|
endif()
|
|
else()
|
|
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} CUDA::cudart CUDA::cublas CUDA::cublasLt)
|
|
endif()
|
|
|
|
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} CUDA::cuda_driver)
|
|
|
|
if (NOT DEFINED CMAKE_CUDA_ARCHITECTURES)
|
|
# 50 == lowest CUDA 12 standard
|
|
# 60 == f16 CUDA intrinsics
|
|
# 61 == integer CUDA intrinsics
|
|
# 70 == (assumed) compute capability at which unrolling a loop in mul_mat_q kernels is faster
|
|
# 75 == int8 tensor cores
|
|
# 80 == Ampere, asynchronous data loading, faster tensor core instructions
|
|
message("CUDA Toolkit Version: ${CUDAToolkit_VERSION}")
|
|
if(CUDAToolkit_VERSION VERSION_GREATER_EQUAL 13)
|
|
add_compile_definitions(GGML_CUDA_USE_GRAPHS) #try enable cuda graphs on cu12 build
|
|
add_compile_definitions(KCPP_LIMIT_CUDA_MAX_ARCH=860)
|
|
set(CMAKE_CUDA_ARCHITECTURES "75-virtual;80-virtual;86-virtual") # lowest CUDA 13 standard
|
|
elseif(CUDAToolkit_VERSION VERSION_GREATER 12)
|
|
add_compile_definitions(GGML_CUDA_USE_GRAPHS) #try enable cuda graphs on cu12 build
|
|
add_compile_definitions(KCPP_LIMIT_CUDA_MAX_ARCH=800)
|
|
set(CMAKE_CUDA_ARCHITECTURES "50-virtual;61-virtual;70-virtual;75-virtual;80-virtual") # lowest CUDA 12 standard + lowest for integer intrinsics
|
|
else()
|
|
add_compile_definitions(KCPP_LIMIT_CUDA_MAX_ARCH=750) #will cause issues with ggml_cuda_highest_compiled_arch if removed
|
|
set(CMAKE_CUDA_ARCHITECTURES "35-virtual;50-virtual;61-virtual;70-virtual;75-virtual") # lowest CUDA 12 standard + lowest for integer intrinsics
|
|
endif()
|
|
endif()
|
|
message(STATUS "Using CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}")
|
|
|
|
else()
|
|
message(WARNING "cuBLAS not found")
|
|
endif()
|
|
endif()
|
|
|
|
if (LLAMA_HIPBLAS)
|
|
if (MSVC)
|
|
execute_process(COMMAND powershell.exe -Command "Get-ChildItem 'C:/Program Files/AMD/ROCm/' | Sort-Object {$_.Version} | Select-Object -Last 1 | ForEach-Object { Write-Output $_.FullName }"
|
|
OUTPUT_VARIABLE ROCM_PATH
|
|
OUTPUT_STRIP_TRAILING_WHITESPACE
|
|
)
|
|
list(APPEND CMAKE_PREFIX_PATH ${ROCM_PATH})
|
|
add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:-xhip>")
|
|
message(STATUS "ROCM_PATH: ${ROCM_PATH}")
|
|
else()
|
|
list(APPEND CMAKE_PREFIX_PATH /opt/rocm)
|
|
endif()
|
|
|
|
|
|
if (NOT ${CMAKE_C_COMPILER_ID} MATCHES "Clang")
|
|
message(WARNING "Only LLVM is supported for HIP, hint: CC=/opt/rocm/llvm/bin/clang")
|
|
endif()
|
|
if (NOT ${CMAKE_CXX_COMPILER_ID} MATCHES "Clang")
|
|
message(WARNING "Only LLVM is supported for HIP, hint: CXX=/opt/rocm/llvm/bin/clang++")
|
|
endif()
|
|
|
|
find_package(hip)
|
|
find_package(hipblas)
|
|
find_package(rocblas)
|
|
|
|
if (${hipblas_FOUND} AND ${hip_FOUND})
|
|
message(STATUS "HIP and hipBLAS found")
|
|
file(GLOB GGML_SOURCES_ROCM "ggml/src/ggml-cuda/*.cu")
|
|
list(APPEND GGML_SOURCES_ROCM "ggml/src/ggml-cuda/ggml-cuda.cu")
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-tile*.cu")
|
|
list(APPEND GGML_SOURCES_ROCM ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-mma*.cu")
|
|
list(APPEND GGML_SOURCES_ROCM ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/mmq*.cu")
|
|
list(APPEND GGML_SOURCES_ROCM ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/mmf*.cu")
|
|
list(APPEND GGML_SOURCES_ROCM ${SRCS})
|
|
add_compile_definitions(GGML_USE_HIP GGML_USE_CUDA SD_USE_CUDA GGML_HIP_NO_VMM)
|
|
add_library(ggml-rocm ${GGML_SOURCES_CUDA})
|
|
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-vec*q4_0-q4_0.cu")
|
|
list(APPEND GGML_SOURCES_ROCM ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-vec*q8_0-q8_0.cu")
|
|
list(APPEND GGML_SOURCES_ROCM ${SRCS})
|
|
file(GLOB SRCS "ggml/src/ggml-cuda/template-instances/fattn-vec*f16-f16.cu")
|
|
list(APPEND GGML_SOURCES_ROCM ${SRCS})
|
|
|
|
# only build minimal quants required for fattn quant kv
|
|
set_source_files_properties(${GGML_SOURCES_ROCM} PROPERTIES LANGUAGE CXX)
|
|
target_link_libraries(ggml-rocm PUBLIC hip::device hip::host roc::rocblas roc::hipblas)
|
|
|
|
add_library(ggml-v2-rocm ${GGML_V2_CUDA_SOURCES})
|
|
set_source_files_properties(otherarch/ggml_v2-cuda.cu PROPERTIES LANGUAGE CXX)
|
|
target_link_libraries(ggml-v2-rocm PUBLIC hip::device hip::host roc::rocblas roc::hipblas)
|
|
|
|
add_library(ggml-v3-rocm ${GGML_V3_CUDA_SOURCES})
|
|
set_source_files_properties(otherarch/ggml_v3-cuda.cu PROPERTIES LANGUAGE CXX)
|
|
target_link_libraries(ggml-v3-rocm PUBLIC hip::device hip::host roc::rocblas roc::hipblas)
|
|
|
|
add_library(ggml-v2-legacy-rocm ${GGML_V2_LEGACY_CUDA_SOURCES})
|
|
set_source_files_properties(otherarch/ggml_v2-cuda-legacy.cu PROPERTIES LANGUAGE CXX)
|
|
target_link_libraries(ggml-v2-legacy-rocm PUBLIC hip::device hip::host roc::rocblas roc::hipblas)
|
|
|
|
if (LLAMA_STATIC)
|
|
message(FATAL_ERROR "Static linking not supported for HIP/ROCm")
|
|
endif()
|
|
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} ggml-rocm ggml-v2-rocm ggml-v3-rocm ggml-v2-legacy-rocm)
|
|
else()
|
|
message(WARNING "hipBLAS or HIP not found. Try setting CMAKE_PREFIX_PATH=/opt/rocm")
|
|
endif()
|
|
endif()
|
|
|
|
|
|
if (WIN32)
|
|
add_compile_definitions(_CRT_SECURE_NO_WARNINGS)
|
|
|
|
if (BUILD_SHARED_LIBS)
|
|
set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON)
|
|
endif()
|
|
endif()
|
|
|
|
if (LLAMA_LTO)
|
|
include(CheckIPOSupported)
|
|
check_ipo_supported(RESULT result OUTPUT output)
|
|
if (result)
|
|
set(CMAKE_INTERPROCEDURAL_OPTIMIZATION TRUE)
|
|
else()
|
|
message(WARNING "IPO is not supported: ${output}")
|
|
endif()
|
|
endif()
|
|
|
|
if (LLAMA_OPENMP)
|
|
find_package(OpenMP)
|
|
if (OpenMP_FOUND)
|
|
message(STATUS "OpenMP found")
|
|
add_compile_definitions(GGML_USE_OPENMP)
|
|
set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} OpenMP::OpenMP_C OpenMP::OpenMP_CXX)
|
|
else()
|
|
message(WARNING "OpenMP not found")
|
|
endif()
|
|
endif()
|
|
|
|
# this version of Apple ld64 is buggy
|
|
execute_process(
|
|
COMMAND ${CMAKE_C_COMPILER} ${CMAKE_EXE_LINKER_FLAGS} -Wl,-v
|
|
ERROR_VARIABLE output
|
|
)
|
|
if (output MATCHES "dyld-1015\.7")
|
|
add_compile_definitions(HAVE_BUGGY_APPLE_LINKER)
|
|
endif()
|
|
|
|
# Architecture specific
|
|
# TODO: probably these flags need to be tweaked on some architectures
|
|
# feel free to update the Makefile for your architecture and send a pull request or issue
|
|
message(STATUS "CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
|
|
if (NOT MSVC)
|
|
if (LLAMA_STATIC)
|
|
add_link_options(-static)
|
|
if (MINGW)
|
|
add_link_options(-static-libgcc -static-libstdc++)
|
|
endif()
|
|
endif()
|
|
if (LLAMA_GPROF)
|
|
add_compile_options(-pg)
|
|
endif()
|
|
if (LLAMA_NATIVE)
|
|
add_compile_options(-march=native)
|
|
endif()
|
|
endif()
|
|
|
|
if ((${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm") OR (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64"))
|
|
message(STATUS "ARM detected")
|
|
if (MSVC)
|
|
# TODO: arm msvc?
|
|
else()
|
|
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv6")
|
|
# Raspberry Pi 1, Zero
|
|
add_compile_options(-mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access)
|
|
endif()
|
|
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv7")
|
|
# Raspberry Pi 2
|
|
add_compile_options(-mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations)
|
|
endif()
|
|
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv8")
|
|
# Raspberry Pi 3, 4, Zero 2 (32-bit)
|
|
add_compile_options(-mfp16-format=ieee -mno-unaligned-access)
|
|
endif()
|
|
endif()
|
|
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$")
|
|
message(STATUS "x86 detected")
|
|
if (MSVC)
|
|
if (LLAMA_AVX512)
|
|
add_compile_options($<$<COMPILE_LANGUAGE:C>:/arch:AVX512>)
|
|
add_compile_options($<$<COMPILE_LANGUAGE:CXX>:/arch:AVX512>)
|
|
# MSVC has no compile-time flags enabling specific
|
|
# AVX512 extensions, neither it defines the
|
|
# macros corresponding to the extensions.
|
|
# Do it manually.
|
|
if (LLAMA_AVX512_VBMI)
|
|
add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AVX512VBMI__>)
|
|
add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AVX512VBMI__>)
|
|
endif()
|
|
if (LLAMA_AVX512_VNNI)
|
|
add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AVX512VNNI__>)
|
|
add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AVX512VNNI__>)
|
|
endif()
|
|
elseif (LLAMA_AVX2)
|
|
add_compile_options($<$<COMPILE_LANGUAGE:C>:/arch:AVX2>)
|
|
add_compile_options($<$<COMPILE_LANGUAGE:CXX>:/arch:AVX2>)
|
|
elseif (LLAMA_AVX)
|
|
add_compile_options($<$<COMPILE_LANGUAGE:C>:/arch:AVX>)
|
|
add_compile_options($<$<COMPILE_LANGUAGE:CXX>:/arch:AVX>)
|
|
endif()
|
|
else()
|
|
if (LLAMA_F16C)
|
|
add_compile_options(-mf16c)
|
|
endif()
|
|
if (LLAMA_FMA)
|
|
add_compile_options(-mfma)
|
|
endif()
|
|
if (LLAMA_AVX)
|
|
add_compile_options(-mavx)
|
|
endif()
|
|
if (LLAMA_AVX2)
|
|
add_compile_options(-mavx2)
|
|
endif()
|
|
if (LLAMA_AVX512)
|
|
add_compile_options(-mavx512f)
|
|
add_compile_options(-mavx512bw)
|
|
endif()
|
|
if (LLAMA_AVX512_VBMI)
|
|
add_compile_options(-mavx512vbmi)
|
|
endif()
|
|
if (LLAMA_AVX512_VNNI)
|
|
add_compile_options(-mavx512vnni)
|
|
endif()
|
|
endif()
|
|
elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64")
|
|
message(STATUS "PowerPC detected")
|
|
if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64le")
|
|
add_compile_options(-mcpu=powerpc64le)
|
|
else()
|
|
add_compile_options(-mcpu=native -mtune=native)
|
|
#TODO: Add targets for Power8/Power9 (Altivec/VSX) and Power10(MMA) and query for big endian systems (ppc64/le/be)
|
|
endif()
|
|
else()
|
|
message(STATUS "Unknown architecture")
|
|
endif()
|
|
|
|
if (MINGW)
|
|
# Target Windows 8 for PrefetchVirtualMemory
|
|
add_compile_definitions(_WIN32_WINNT=0x602)
|
|
endif()
|
|
|
|
#
|
|
# Build libraries
|
|
#
|
|
|
|
add_library(ggml
|
|
ggml/src/ggml.c
|
|
ggml/include/ggml.h
|
|
ggml/src/ggml-cpu/ggml-cpu.c
|
|
ggml/include/ggml-cpu.h
|
|
ggml/src/ggml-alloc.c
|
|
ggml/include/ggml-alloc.h
|
|
ggml/src/ggml-backend.cpp
|
|
ggml/src/ggml-backend-impl.h
|
|
ggml/include/ggml-backend.h
|
|
ggml/include/ggml-cpp.h
|
|
ggml/src/ggml-quants.c
|
|
ggml/src/ggml-quants.h
|
|
ggml/src/ggml-cpu/llamafile/sgemm.cpp
|
|
ggml/src/ggml-cpu/llamafile/sgemm.h
|
|
ggml/src/ggml-cpu/traits.cpp
|
|
ggml/src/ggml-cpu/traits.h
|
|
ggml/src/ggml-threading.cpp
|
|
ggml/src/ggml-cpu/ggml-cpu.cpp
|
|
ggml/src/ggml-cpu/kcpp-quantmapper.c
|
|
ggml/src/ggml-cpu/kcpp-repackmapper.cpp
|
|
ggml/src/ggml-cpu/repack.cpp
|
|
ggml/src/ggml-cpu/repack.h
|
|
ggml/src/ggml-cpu/quants.c
|
|
ggml/src/ggml-cpu/quants.h
|
|
ggml/src/ggml-cpu/common.h
|
|
ggml/src/ggml-cpu/binary-ops.h
|
|
ggml/src/ggml-cpu/binary-ops.cpp
|
|
ggml/src/ggml-cpu/unary-ops.h
|
|
ggml/src/ggml-cpu/unary-ops.cpp
|
|
ggml/src/ggml-cpu/simd-mappings.h
|
|
ggml/src/ggml-cpu/vec.h
|
|
ggml/src/ggml-cpu/vec.cpp
|
|
ggml/src/ggml-cpu/ops.h
|
|
ggml/src/ggml-cpu/ops.cpp
|
|
ggml/src/ggml-backend-reg.cpp
|
|
ggml/include/gguf.h
|
|
ggml/src/gguf.cpp
|
|
${GGML_SOURCES_CUDA})
|
|
target_include_directories(ggml PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor)
|
|
target_compile_features(ggml PUBLIC c_std_11) # don't bump
|
|
target_link_libraries(ggml PUBLIC Threads::Threads ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(ggml PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
target_compile_options(ggml PRIVATE $<$<COMPILE_LANGUAGE:CUDA>:-use_fast_math -extended-lambda>)
|
|
|
|
add_library(ggml_v1
|
|
otherarch/ggml_v1.c
|
|
otherarch/ggml_v1.h)
|
|
target_include_directories(ggml_v1 PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor)
|
|
target_compile_features(ggml_v1 PUBLIC c_std_11) # don't bump
|
|
target_link_libraries(ggml_v1 PUBLIC Threads::Threads ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(ggml_v1 PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(ggml_v2
|
|
otherarch/ggml_v2.c
|
|
otherarch/ggml_v2.h
|
|
${GGML_V2_CUDA_SOURCES}
|
|
${GGML_V2_LEGACY_CUDA_SOURCES})
|
|
target_include_directories(ggml_v2 PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor)
|
|
target_compile_features(ggml_v2 PUBLIC c_std_11) # don't bump
|
|
target_link_libraries(ggml_v2 PUBLIC Threads::Threads ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(ggml_v2 PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(ggml_v3
|
|
otherarch/ggml_v3.c
|
|
otherarch/ggml_v3.h
|
|
${GGML_V3_CUDA_SOURCES})
|
|
target_include_directories(ggml_v3 PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor)
|
|
target_compile_features(ggml_v3 PUBLIC c_std_11) # don't bump
|
|
target_link_libraries(ggml_v3 PUBLIC Threads::Threads ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(ggml_v3 PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(common2
|
|
common/common.cpp
|
|
common/common.h
|
|
common/sampling.cpp
|
|
common/sampling.h
|
|
tools/mtmd/llava.cpp
|
|
tools/mtmd/llava.h
|
|
tools/mtmd/clip.cpp
|
|
tools/mtmd/clip.h
|
|
src/unicode.h
|
|
src/unicode.cpp
|
|
src/unicode-data.cpp
|
|
otherarch/utils.cpp
|
|
otherarch/utils.h
|
|
tools/mtmd/mtmd-audio.cpp
|
|
tools/mtmd/mtmd-audio.h)
|
|
target_include_directories(common2 PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./otherarch/sdcpp ./otherarch/sdcpp/thirdparty ./tools ./common)
|
|
target_compile_features(common2 PUBLIC cxx_std_17) # don't bump
|
|
target_link_libraries(common2 PRIVATE ggml ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(common2 PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(sdtype_adapter
|
|
otherarch/sdcpp/sdtype_adapter.cpp)
|
|
target_include_directories(sdtype_adapter PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./otherarch/sdcpp ./otherarch/sdcpp/thirdparty ./tools ./common)
|
|
target_compile_features(sdtype_adapter PUBLIC cxx_std_17) # don't bump
|
|
target_link_libraries(sdtype_adapter PRIVATE common2 ggml ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(sdtype_adapter PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(whisper_adapter
|
|
otherarch/whispercpp/whisper_adapter.cpp)
|
|
target_include_directories(whisper_adapter PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./otherarch/whispercpp ./tools ./common)
|
|
target_compile_features(whisper_adapter PUBLIC cxx_std_17) # don't bump
|
|
target_link_libraries(whisper_adapter PRIVATE common2 ggml ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(whisper_adapter PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(tts_adapter
|
|
otherarch/tts_adapter.cpp)
|
|
target_include_directories(tts_adapter PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./otherarch/ttscpp/include ./otherarch/ttscpp/src ./tools ./common)
|
|
target_compile_features(tts_adapter PUBLIC cxx_std_17) # don't bump
|
|
target_link_libraries(tts_adapter PRIVATE common2 ggml ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(tts_adapter PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(embeddings_adapter
|
|
otherarch/embeddings_adapter.cpp)
|
|
target_include_directories(embeddings_adapter PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./tools ./common)
|
|
target_compile_features(embeddings_adapter PUBLIC cxx_std_17) # don't bump
|
|
target_link_libraries(embeddings_adapter PRIVATE common2 ggml ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(embeddings_adapter PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
add_library(gpttype_adapter
|
|
gpttype_adapter.cpp)
|
|
target_include_directories(gpttype_adapter PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./otherarch/sdcpp ./otherarch/sdcpp/thirdparty ./tools ./common)
|
|
target_compile_features(gpttype_adapter PUBLIC cxx_std_17) # don't bump
|
|
target_link_libraries(gpttype_adapter PRIVATE common2 ggml ggml_v1 ggml_v2 ggml_v3 ${LLAMA_EXTRA_LIBS})
|
|
set_target_properties(gpttype_adapter PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
|
|
if (LLAMA_CUBLAS)
|
|
set(TARGET koboldcpp_cublas)
|
|
add_library(${TARGET} SHARED expose.cpp expose.h)
|
|
target_include_directories(${TARGET} PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./otherarch/sdcpp ./otherarch/sdcpp/thirdparty ./tools ./common)
|
|
target_compile_features(${TARGET} PUBLIC cxx_std_17) # don't bump
|
|
set_target_properties(${TARGET} PROPERTIES PREFIX "")
|
|
set_target_properties(${TARGET} PROPERTIES OUTPUT_NAME "koboldcpp_cublas")
|
|
set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
target_link_libraries(${TARGET} PUBLIC Threads::Threads ggml ggml_v1 ggml_v2 ggml_v3 common2 gpttype_adapter whisper_adapter tts_adapter embeddings_adapter sdtype_adapter ${LLAMA_EXTRA_LIBS})
|
|
target_compile_features(${TARGET} PRIVATE cxx_std_17)
|
|
|
|
add_custom_command(
|
|
TARGET koboldcpp_cublas POST_BUILD
|
|
COMMAND ${CMAKE_COMMAND} -E copy
|
|
$<TARGET_FILE:koboldcpp_cublas> # The generated DLL
|
|
${CMAKE_SOURCE_DIR}/ # Destination directory
|
|
COMMENT "Copying DLL to parent directory"
|
|
)
|
|
endif()
|
|
|
|
if (LLAMA_HIPBLAS)
|
|
set(TARGET koboldcpp_hipblas)
|
|
add_library(${TARGET} SHARED expose.cpp expose.h)
|
|
target_include_directories(${TARGET} PUBLIC . ./ggml/include ./ggml/src ./ggml/src/ggml-cpu ./include ./otherarch ./otherarch/tools ./vendor/stb ./vendor ./otherarch/sdcpp ./otherarch/sdcpp/thirdparty ./tools ./common)
|
|
target_compile_features(${TARGET} PUBLIC cxx_std_17) # don't bump
|
|
set_target_properties(${TARGET} PROPERTIES PREFIX "")
|
|
set_target_properties(${TARGET} PROPERTIES OUTPUT_NAME "koboldcpp_hipblas")
|
|
set_target_properties(${TARGET} PROPERTIES POSITION_INDEPENDENT_CODE ON)
|
|
target_link_libraries(${TARGET} PUBLIC Threads::Threads ggml ggml_v1 ggml_v2 ggml_v3 common2 gpttype_adapter whisper_adapter tts_adapter embeddings_adapter sdtype_adapter ${LLAMA_EXTRA_LIBS})
|
|
target_compile_features(${TARGET} PRIVATE cxx_std_17)
|
|
|
|
add_custom_command(
|
|
TARGET koboldcpp_hipblas POST_BUILD
|
|
COMMAND ${CMAKE_COMMAND} -E copy
|
|
$<TARGET_FILE:koboldcpp_hipblas> # The generated DLL
|
|
${CMAKE_SOURCE_DIR}/ # Destination directory
|
|
COMMENT "Copying DLL to parent directory"
|
|
)
|
|
endif()
|
|
|