files : reorganize + reuse CMake scripts [no ci]

.gitignore

@@ -1,13 +1,5 @@
 build/
-build-blas/
-build-debug/
-build-release/
-build-sanitize-addr/
-build-sanitize-thread/
-build-cov/
-build-ci-debug/
-build-ci-release/
-build-cublas/
+build-*/
 out/
 tmp/
 models/

CMakeLists.txt

@@ -1,20 +1,29 @@
-cmake_minimum_required (VERSION 3.12)
-project(ggml VERSION 0.1.0)
+cmake_minimum_required(VERSION 3.14) # for add_link_options and implicit target directories.
+project("ggml" C CXX)
+include(CheckIncludeFileCXX)
 
-set(CMAKE_EXPORT_COMPILE_COMMANDS "on")
-set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
-set(CMAKE_INSTALL_RPATH "${CMAKE_INSTALL_PREFIX}/lib")
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
 
-if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
+if (NOT XCODE AND NOT MSVC AND NOT CMAKE_BUILD_TYPE)
+ set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
+ set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "MinSizeRel" "RelWithDebInfo")
+endif()
+
+if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR)
  set(GGML_STANDALONE ON)
- include(cmake/GitVars.cmake)
- include(cmake/BuildTypes.cmake)
+
+ set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
+
+ # configure project version
+ # TODO
 else()
  set(GGML_STANDALONE OFF)
 endif()
 
 if (EMSCRIPTEN)
  set(BUILD_SHARED_LIBS_DEFAULT OFF)
+
+ option(GGML_WASM_SINGLE_FILE "ggml: embed WASM inside the generated ggml.js" ON)
 else()
  if (MINGW)
  set(BUILD_SHARED_LIBS_DEFAULT OFF)
@@ -23,7 +32,13 @@ else()
  endif()
 endif()
 
-# options
+option(BUILD_SHARED_LIBS "ggml: build shared libraries" ${BUILD_SHARED_LIBS_DEFAULT})
+
+#
+# option list
+#
+
+# TODO: mark all options as advanced when not GGML_STANDALONE
 
 if (APPLE)
  set(GGML_METAL_DEFAULT ON)
@@ -35,190 +50,165 @@ else()
  set(GGML_BLAS_VENDOR_DEFAULT "Generic")
 endif()
 
-option(BUILD_SHARED_LIBS "ggml: build shared libs" ${BUILD_SHARED_LIBS_DEFAULT})
-
-option(GGML_ALL_WARNINGS "ggml: enable all compiler warnings" ON)
-option(GGML_ALL_WARNINGS_3RD_PARTY "ggml: enable all compiler warnings in 3rd party libs" OFF)
-
-option(GGML_SANITIZE_THREAD "ggml: enable thread sanitizer" OFF)
-option(GGML_SANITIZE_ADDRESS "ggml: enable address sanitizer" OFF)
-option(GGML_SANITIZE_UNDEFINED "ggml: enable undefined sanitizer" OFF)
+# general
+option(GGML_STATIC "ggml: static link libraries" OFF)
+option(GGML_NATIVE "ggml: enable -march=native flag" ON)
+option(GGML_LTO "ggml: enable link time optimization" OFF)
+option(GGML_CCACHE "ggml: use ccache if available" ON)
 
-option(GGML_BUILD_TESTS "ggml: build tests" ${GGML_STANDALONE})
-option(GGML_BUILD_EXAMPLES "ggml: build examples" ${GGML_STANDALONE})
+# debug
+option(GGML_ALL_WARNINGS "ggml: enable all compiler warnings" ON)
+option(GGML_ALL_WARNINGS_3RD_PARTY "ggml: enable all compiler warnings in 3rd party libs" OFF)
+option(GGML_GPROF "ggml: enable gprof" OFF)
 
-option(GGML_TEST_COVERAGE "ggml: enable test coverage" OFF)
+# build
+option(GGML_FATAL_WARNINGS "ggml: enable -Werror flag" OFF)
 
-option(GGML_PERF "ggml: enable perf timings" OFF)
-option(GGML_NO_ACCELERATE "ggml: disable Accelerate framework" OFF)
-option(GGML_BLAS "ggml: use BLAS" ${GGML_BLAS_DEFAULT})
-set(GGML_BLAS_VENDOR ${GGML_BLAS_VENDOR_DEFAULT} CACHE STRING
- "ggml: BLAS library vendor")
-option(GGML_HIPBLAS "ggml: use hipBLAS" OFF)
-option(GGML_CUDA "ggml: use CUDA" OFF)
-option(GGML_CUBLAS "ggml: use CUDA (deprecated)" OFF)
-option(GGML_METAL "ggml: use Metal" ${GGML_METAL_DEFAULT})
-option(GGML_METAL_NDEBUG "ggml: disable Metal debugging" OFF)
-option(GGML_METAL_SHADER_DEBUG "ggml: compile Metal with -fno-fast-math" OFF)
-option(GGML_METAL_EMBED_LIBRARY "ggml: embed Metal library" OFF)
-option(GGML_RPC "ggml: use RPC" OFF)
-option(GGML_VULKAN "ggml: use Vulkan" OFF)
-
-option(GGML_CUDA_FORCE_DMMV "ggml: use dmmv instead of mmvq CUDA kernels" OFF)
-option(GGML_CUDA_FORCE_MMQ "ggml: use mmq kernels instead of cuBLAS" OFF)
-set(GGML_CUDA_DMMV_X "32" CACHE STRING "ggml: x stride for dmmv CUDA kernels")
-set(GGML_CUDA_MMV_Y "1" CACHE STRING "ggml: y block size for mmv CUDA kernels")
-option(GGML_CUDA_F16 "ggml: use 16 bit floats for some calculations" OFF)
-set(GGML_CUDA_KQUANTS_ITER "2" CACHE STRING "ggml: iters./thread per block for Q2_K/Q6_K")
-set(GGML_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
- "ggml: max. batch size for using peer access")
 # sanitizers
+option(GGML_SANITIZE_THREAD "ggml: enable thread sanitizer" OFF)
+option(GGML_SANITIZE_ADDRESS "ggml: enable address sanitizer" OFF)
+option(GGML_SANITIZE_UNDEFINED "ggml: enable undefined sanitizer" OFF)
 
-if (GGML_SANITIZE_THREAD)
- set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=thread")
- set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=thread")
-endif()
-
-if (GGML_SANITIZE_ADDRESS)
- set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=address -fno-omit-frame-pointer")
- set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=address -fno-omit-frame-pointer")
+# instruction set specific
+if (GGML_NATIVE)
+ set(INS_ENB OFF)
+else()
+ set(INS_ENB ON)
 endif()
 
-if (GGML_SANITIZE_UNDEFINED)
- set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fsanitize=undefined")
- set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsanitize=undefined")
-endif()
+option(GGML_CPU_HBM "ggml: use memkind for CPU HBM" OFF)
 
-# instruction set specific
-option(GGML_AVX "ggml: enable AVX" ON)
-option(GGML_AVX2 "ggml: enable AVX2" ON)
-option(GGML_AVX512 "ggml: enable AVX512" OFF)
-option(GGML_AVX512_VBMI "ggml: enable AVX512-VBMI" OFF)
-option(GGML_AVX512_VNNI "ggml: enable AVX512-VNNI" OFF)
-option(GGML_FMA "ggml: enable FMA" ON)
-# in MSVC F16C is implied with AVX2/AVX512
+option(GGML_AVX "ggml: enable AVX" ${INS_ENB})
+option(GGML_AVX2 "ggml: enable AVX2" ${INS_ENB})
+option(GGML_AVX512 "ggml: enable AVX512" OFF)
+option(GGML_AVX512_VBMI "ggml: enable AVX512-VBMI" OFF)
+option(GGML_AVX512_VNNI "ggml: enable AVX512-VNNI" OFF)
+option(GGML_AVX512_BF16 "ggml: enable AVX512-BF16" OFF)
+option(GGML_FMA "ggml: enable FMA" ${INS_ENB})
 if (NOT MSVC)
- option(GGML_F16C  "ggml: enable F16C"  ON)
+ option(GGML_F16C "ggml: enable F16C" ${INS_ENB}) # in MSVC F16C is implied with AVX2/AVX512
 endif()
+option(GGML_LASX "ggml: enable lasx" ON)
+option(GGML_LSX "ggml: enable lsx" ON)
+option(GGML_SVE "ggml: enable SVE" OFF)
 
-#set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -ffast-math")
-#set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -march=native")
-#set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -mcpu=native")
+if (WIN32)
+ set(GGML_WIN_VER "0x602" CACHE STRING "ggml: Windows Version")
+endif()
 
-# warning flags
+# ggml core
+set(GGML_SCHED_MAX_COPIES "4" CACHE STRING "ggml: max input copies for pipeline parallelism")
+option(GGML_PERF "ggml: enable perf" OFF)
 
-if (GGML_ALL_WARNINGS)
- if (NOT MSVC)
- set(c_flags -Wall -Wpedantic -Wformat=2 -Wno-unused -Wstrict-prototypes)
- set(cxx_flags -Wall -Wpedantic -Wformat=2)
- else()
- # todo : windows
- endif()
-
- add_compile_options(
- "$<$<COMPILE_LANGUAGE:C>:${c_flags}>"
- "$<$<COMPILE_LANGUAGE:CXX>:${cxx_flags}>"
- )
-endif()
+# 3rd party libs / backends
+option(GGML_ACCELERATE "ggml: enable Accelerate framework" ON)
+option(GGML_BLAS "ggml: use BLAS" ${GGML_BLAS_DEFAULT})
+set(GGML_BLAS_VENDOR ${GGML_BLAS_VENDOR_DEFAULT} CACHE STRING
+ "ggml: BLAS library vendor")
+option(GGML_LLAMAFILE "ggml: use ggml SGEMM" OFF)
 
-if (NOT MSVC)
- # TODO: temporary disabled until we figure out ggml-metal.m
- #add_compile_options(
- # "$<$<COMPILE_LANGUAGE:C>:-Werror=vla>"
- # "$<$<COMPILE_LANGUAGE:CXX>:-Werror=vla>"
- # "$<$<COMPILE_LANGUAGE:CUDA>:-Xcompiler;-Werror=vla>"
- #)
-endif()
+option(GGML_CUDA "ggml: use CUDA" OFF)
+option(GGML_CUDA_FORCE_DMMV "ggml: use dmmv instead of mmvq CUDA kernels" OFF)
+option(GGML_CUDA_FORCE_MMQ "ggml: use mmq kernels instead of cuBLAS" OFF)
+set (GGML_CUDA_DMMV_X "32" CACHE STRING "ggml: x stride for dmmv CUDA kernels")
+set (GGML_CUDA_MMV_Y "1" CACHE STRING "ggml: y block size for mmv CUDA kernels")
+option(GGML_CUDA_F16 "ggml: use 16 bit floats for some calculations" OFF)
+set (GGML_CUDA_KQUANTS_ITER "2" CACHE STRING
+ "ggml: iters./thread per block for Q2_K/Q6_K")
+set (GGML_CUDA_PEER_MAX_BATCH_SIZE "128" CACHE STRING
+ "ggml: max. batch size for using peer access")
+option(GGML_CUDA_NO_PEER_COPY "ggml: do not use peer to peer copies" OFF)
+option(GGML_CUDA_NO_VMM "ggml: do not try to use CUDA VMM" OFF)
+option(GGML_CUDA_FA_ALL_QUANTS "ggml: compile all quants for FlashAttention" OFF)
+
+option(GGML_CURL "ggml: use libcurl to download model from an URL" OFF)
+option(GGML_HIPBLAS "ggml: use hipBLAS" OFF)
+option(GGML_HIP_UMA "ggml: use HIP unified memory architecture" OFF)
+option(GGML_VULKAN "ggml: use Vulkan" OFF)
+option(GGML_VULKAN_CHECK_RESULTS "ggml: run Vulkan op checks" OFF)
+option(GGML_VULKAN_DEBUG "ggml: enable Vulkan debug output" OFF)
+option(GGML_VULKAN_MEMORY_DEBUG "ggml: enable Vulkan memory debug output" OFF)
+option(GGML_VULKAN_VALIDATE "ggml: enable Vulkan validation" OFF)
+option(GGML_VULKAN_RUN_TESTS "ggml: run Vulkan tests" OFF)
+option(GGML_KOMPUTE "ggml: use Kompute" OFF)
+option(GGML_METAL "ggml: use Metal" ${GGML_METAL_DEFAULT})
+option(GGML_METAL_NDEBUG "ggml: disable Metal debugging" OFF)
+option(GGML_METAL_SHADER_DEBUG "ggml: compile Metal with -fno-fast-math" OFF)
+option(GGML_METAL_EMBED_LIBRARY "ggml: embed Metal library" OFF)
+set (GGML_METAL_MACOSX_VERSION_MIN "" CACHE STRING
+ "ggml: metal minimum macOS version")
+set (GGML_METAL_STD "" CACHE STRING "ggml: metal standard version (-std flag)")
+option(GGML_OPENMP "ggml: use OpenMP" ON)
+option(GGML_RPC "ggml: use RPC" OFF)
+option(GGML_SYCL "ggml: use SYCL" OFF)
+option(GGML_SYCL_F16 "ggml: use 16 bit floats for sycl calculations" OFF)
+set (GGML_SYCL_TARGET "INTEL" CACHE STRING
+ "ggml: sycl target device")
+
+# extra artifacts
+option(GGML_BUILD_TESTS "ggml: build tests" ${GGML_STANDALONE})
+option(GGML_BUILD_EXAMPLES "ggml: build examples" ${GGML_STANDALONE})
 
 #
-# POSIX conformance
+# dependencies
 #
 
-# clock_gettime came in POSIX.1b (1993)
-# CLOCK_MONOTONIC came in POSIX.1-2001 / SUSv3 as optional
-# posix_memalign came in POSIX.1-2001 / SUSv3
-# M_PI is an XSI extension since POSIX.1-2001 / SUSv3, came in XPG1 (1985)
-add_compile_definitions(_XOPEN_SOURCE=600)
-
-# Somehow in OpenBSD whenever POSIX conformance is specified
-# some string functions rely on locale_t availability,
-# which was introduced in POSIX.1-2008, forcing us to go higher
-if (CMAKE_SYSTEM_NAME MATCHES "OpenBSD")
- remove_definitions(-D_XOPEN_SOURCE=600)
- add_compile_definitions(_XOPEN_SOURCE=700)
-endif()
-
-# Data types, macros and functions related to controlling CPU affinity
-# are available on Linux through GNU extensions in libc
-if (CMAKE_SYSTEM_NAME MATCHES "Linux")
- add_compile_definitions(_GNU_SOURCE)
-endif()
-
-# RLIMIT_MEMLOCK came in BSD, is not specified in POSIX.1,
-# and on macOS its availability depends on enabling Darwin extensions
-# similarly on DragonFly, enabling BSD extensions is necessary
-if (CMAKE_SYSTEM_NAME MATCHES "Darwin")
- add_compile_definitions(_DARWIN_C_SOURCE)
-endif()
-if (CMAKE_SYSTEM_NAME MATCHES "DragonFly")
- add_compile_definitions(_DARWIN_C_SOURCE)
-endif()
-
-# alloca is a non-standard interface that is not visible on BSDs when
-# POSIX conformance is specified, but not all of them provide a clean way
-# to enable it in such cases
-if (CMAKE_SYSTEM_NAME MATCHES "FreeBSD")
- add_compile_definitions(__BSD_VISIBLE)
-endif()
-if (CMAKE_SYSTEM_NAME MATCHES "NetBSD")
- add_compile_definitions(_NETBSD_SOURCE)
-endif()
-if (CMAKE_SYSTEM_NAME MATCHES "OpenBSD")
- add_compile_definitions(_BSD_SOURCE)
-endif()
+set(CMAKE_C_STANDARD 11)
+set(CMAKE_C_STANDARD_REQUIRED true)
 
-if (WHISPER_PERF)
- set(WHISPER_EXTRA_FLAGS ${WHISPER_EXTRA_FLAGS} -DGGML_PERF)
+if (GGML_SYCL)
+ set(CMAKE_CXX_STANDARD 17)
+else()
+ set(CMAKE_CXX_STANDARD 11)
 endif()
+set(CMAKE_CXX_STANDARD_REQUIRED true)
 
-# dependencies
-
-set(CMAKE_C_STANDARD 11)
-set(CMAKE_CXX_STANDARD 11)
+set(THREADS_PREFER_PTHREAD_FLAG ON)
 
 find_package(Threads REQUIRED)
 
-# main
-
-if (NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
- set(CMAKE_BUILD_TYPE Release CACHE STRING "Build type" FORCE)
- set_property(CACHE CMAKE_BUILD_TYPE PROPERTY STRINGS "Debug" "Release" "RelWithDebInfo")
-endif ()
-
-if (GGML_BUILD_TESTS)
- if (GGML_TEST_COVERAGE)
- if (CMAKE_C_COMPILER_ID MATCHES "Clang")
- set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fprofile-instr-generate -fcoverage-mapping")
- set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fprofile-instr-generate -fcoverage-mapping")
- else()
- message(WARNING "Test coverage is only supported for Clang")
- endif()
- endif()
-endif()
+#
+# build the library
+#
 
 add_subdirectory(src)
 
-if (GGML_BUILD_TESTS)
- enable_testing()
- add_subdirectory(tests)
-endif ()
+#
+# install
+#
 
-if (GGML_BUILD_EXAMPLES)
- add_subdirectory(examples)
-endif ()
+include(GNUInstallDirs)
+include(CMakePackageConfigHelpers)
+
+set(GGML_PUBLIC_HEADERS
+ include/ggml.h
+ include/ggml-alloc.h
+ include/ggml-backend.h
+ "${GGML_HEADERS_CUDA}"
+ "${GGML_HEADERS_METAL}"
+ "${GGML_HEADERS_EXTRA}")
+
+set_target_properties(ggml PROPERTIES PUBLIC_HEADER "${GGML_PUBLIC_HEADERS}")
+install(TARGETS ggml PUBLIC_HEADER)
+
+if (BUILD_SHARED_LIBS)
+ install(TARGETS ggml_shared LIBRARY)
+endif()
+
+if (GGML_METAL)
+ install(
+ FILES src/ggml-metal.metal
+ PERMISSIONS
+ OWNER_READ
+ OWNER_WRITE
+ GROUP_READ
+ WORLD_READ
+ DESTINATION ${CMAKE_INSTALL_BINDIR})
+ if (NOT GGML_METAL_EMBED_LIBRARY)
+ install(
+ FILES ${CMAKE_RUNTIME_OUTPUT_DIRECTORY}/default.metallib
+ DESTINATION ${CMAKE_INSTALL_BINDIR}
+ )
+ endif()
+endif()
 
-configure_file(${CMAKE_CURRENT_SOURCE_DIR}/ggml.pc.in
- ${CMAKE_CURRENT_BINARY_DIR}/ggml.pc
- @ONLY)
-install(FILES ${CMAKE_CURRENT_BINARY_DIR}/ggml.pc
- DESTINATION share/pkgconfig)