[llama-cpp] Create a new port

.circleci/port-setup.txt

@@ -18,4 +18,4 @@ opencl:x64-windows
 egl-registry:x64-windows
 opengl-registry:x64-windows
 vulkan-headers:x64-windows
-nlohmann-json:x64-windows
+nlohmann-json:x64-windows

.circleci/port-windows.txt

@@ -1,2 +1,2 @@
 grpc[codegen]:x64-windows
-onnxruntime[directml,xnnpack,test]:x64-windows
+onnxruntime[directml,xnnpack,test]:x64-windows

ports/llama-cpp/fix-cmake.patch

@@ -0,0 +1,55 @@
+diff --git a/scripts/build-info.cmake b/scripts/build-info.cmake
+index c86ab43..0571c20 100644
+--- a/scripts/build-info.cmake
++++ b/scripts/build-info.cmake
+@@ -1,9 +1,5 @@
+ set(TEMPLATE_FILE "${CMAKE_CURRENT_SOURCE_DIR}/scripts/build-info.h.in")
+ set(HEADER_FILE "${CMAKE_CURRENT_SOURCE_DIR}/build-info.h")
+-set(BUILD_NUMBER 0)
+-set(BUILD_COMMIT "unknown")
+-set(BUILD_COMPILER "unknown")
+-set(BUILD_TARGET "unknown")
+
+ # Look for git
+ find_package(Git)
+@@ -18,22 +14,7 @@ if(NOT Git_FOUND)
+ endif()
+
+ # Get the commit count and hash
+-if(Git_FOUND)
+-    execute_process(
+-        COMMAND ${GIT_EXECUTABLE} rev-parse --short HEAD
+-        WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
+-        OUTPUT_VARIABLE HEAD
+-        OUTPUT_STRIP_TRAILING_WHITESPACE
+-    )
+-    execute_process(
+-        COMMAND ${GIT_EXECUTABLE} rev-list --count HEAD
+-        WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
+-        OUTPUT_VARIABLE COUNT
+-        OUTPUT_STRIP_TRAILING_WHITESPACE
+-    )
+-    set(BUILD_COMMIT ${HEAD})
+-    set(BUILD_NUMBER ${COUNT})
+-endif()
++# see portfile.cmake
+
+ if(MSVC)
+     set(BUILD_COMPILER "${CMAKE_C_COMPILER_ID} ${CMAKE_C_COMPILER_VERSION}")
+@@ -72,3 +53,16 @@ if(EXISTS ${HEADER_FILE})
+ else()
+     configure_file(${TEMPLATE_FILE} ${HEADER_FILE})
+ endif()
++
++if(NOT DEFINED BUILD_NUMBER)
++    set(BUILD_NUMBER 0)
++endif()
++if(NOT DEFINED BUILD_COMMIT)
++    set(BUILD_COMMIT "unknown")
++endif()
++if(NOT DEFINED BUILD_COMPILER)
++    set(BUILD_COMPILER "unknown")
++endif()
++if(NOT DEFINED BUILD_TARGET)
++    set(BUILD_TARGET "unknown")
++endif()

ports/llama-cpp/fix-openblas.patch

@@ -0,0 +1,67 @@
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index 11ebe9e..8516e59 100644
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -137,8 +137,8 @@ if (LLAMA_OPENBLAS)
+             $ENV{OpenBLAS_HOME}
+             $ENV{OpenBLAS_HOME}/include
+             )
+-        find_path(OPENBLAS_INC NAMES cblas.h PATHS ${OPENBLAS_INCLUDE_SEARCH_PATHS})
+-        add_compile_options(-I${OPENBLAS_INC})
++        find_path(OPENBLAS_INC NAMES cblas.h openblas/cblas.h PATHS ${OPENBLAS_INCLUDE_SEARCH_PATHS})
++        include_directories(${OPENBLAS_INC})
+     else()
+         message(WARNING "OpenBLAS not found")
+     endif()
+diff --git a/ggml.c b/ggml.c
+index 8505889..6b5344c 100644
+--- a/ggml.c
++++ b/ggml.c
+@@ -146,7 +146,7 @@ inline static void* ggml_aligned_malloc(size_t size) {
+ #if defined(GGML_USE_ACCELERATE)
+ #include <Accelerate/Accelerate.h>
+ #elif defined(GGML_USE_OPENBLAS)
+-#include <cblas.h>
++#include <openblas/cblas.h>
+ #elif defined(GGML_USE_CUBLAS)
+ #include "ggml-cuda.h"
+ #endif
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index 8516e59..cb607b2 100644
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -139,6 +139,8 @@ if (LLAMA_OPENBLAS)
+             )
+         find_path(OPENBLAS_INC NAMES cblas.h openblas/cblas.h PATHS ${OPENBLAS_INCLUDE_SEARCH_PATHS})
+         include_directories(${OPENBLAS_INC})
++        find_library(OPENBLAS_LIB NAMES openblas REQUIRED)
++        link_libraries(${OPENBLAS_LIB})
+     else()
+         message(WARNING "OpenBLAS not found")
+     endif()
+diff --git a/CMakeLists.txt b/CMakeLists.txt
+index cb607b2..f4fd773 100644
+--- a/CMakeLists.txt
++++ b/CMakeLists.txt
+@@ -123,7 +123,7 @@ if (LLAMA_OPENBLAS)
+
+         add_compile_definitions(GGML_USE_OPENBLAS)
+         add_link_options(${BLAS_LIBRARIES})
+-        set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} openblas)
++        set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} BLAS::BLAS)
+
+         # find header file
+         set(OPENBLAS_INCLUDE_SEARCH_PATHS
+@@ -340,7 +340,11 @@ if (GGML_CUDA_SOURCES)
+     set_property(TARGET llama PROPERTY CUDA_ARCHITECTURES OFF)
+ endif()
+
+-
++install(TARGETS ggml llama
++    RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
++    ARCHIVE DESTINATION ${CMAKE_INSTALL_LIBDIR}
++    LIBRARY DESTINATION ${CMAKE_INSTALL_LIBDIR}
++)
+ #
+ # programs, examples and tests
+ #

ports/llama-cpp/portfile.cmake

@@ -0,0 +1,105 @@
+if(VCPKG_TARGET_IS_WINDOWS)
+    vcpkg_check_linkage(ONLY_DYNAMIC_LIBRARY)
+endif()
+
+vcpkg_from_github(
+    OUT_SOURCE_PATH SOURCE_PATH
+    REPO ggerganov/llama.cpp
+    REF b1273 # commit 99115f3fa654b593099c6719ad30e3f54ce231e1
+    SHA512 2b3e8fd9673647f59a4fa96621afe2f77ab10a2bee88a96b662b493beb2b66f17c854c1077f01f8ea8998d0296f92225d3033aae0adc756810f80caf45b9a456
+    HEAD_REF master
+    PATCHES
+        # fix-openblas.patch
+        fix-cmake.patch
+)
+
+vcpkg_find_acquire_program(PKGCONFIG)
+message(STATUS "Using pkgconfig: ${PKGCONFIG}")
+
+# check https://cmake.org/cmake/help/latest/module/FindBLAS.html#blas-lapack-vendors
+if(VCPKG_TARGET_IS_WINDOWS)
+    list(APPEND BLAS_OPTIONS -DLLAMA_BLAS_VENDOR=OpenBLAS)
+elseif(VCPKG_TARGET_IS_OSX OR VCPKG_TARGET_IS_IOS)
+    list(APPEND BLAS_OPTIONS -DLLAMA_BLAS_VENDOR=Apple)
+else()
+    # todo: Intel MKL, ARM, ACML, etc...
+    list(APPEND BLAS_OPTIONS -DLLAMA_BLAS_VENDOR=Generic)
+endif()
+
+if(VCPKG_TARGET_ARCHITECTURE STREQUAL "x64")
+    list(APPEND ARCH_OPTIONS
+        -DLLAMA_AVX512=ON -DLLAMA_AVX512_VBMI=ON -DLLAMA_AVX512_VNNI=ON
+        # -DLLAMA_AVX2=ON
+        # -DLLAMA_AVX=ON
+    )
+endif()
+
+vcpkg_check_features(OUT_FEATURE_OPTIONS FEATURE_OPTIONS
+    FEATURES
+        cublas  LLAMA_CUBLAS
+        cublas  LLAMA_CUDA_F16
+        clblast LLAMA_CLBLAST
+        mpi     LLAMA_MPI
+        test    LLAMA_BUILD_TESTS
+)
+
+string(COMPARE EQUAL "${VCPKG_CRT_LINKAGE}" "static" USE_STATIC)
+
+if(VCPKG_TARGET_IS_OSX OR VCPKG_TARGET_IS_IOS)
+    set(TARGET_IS_APPLE ON)
+else()
+    set(TARGET_IS_APPLE OFF)
+endif()
+
+vcpkg_cmake_configure(
+    SOURCE_PATH "${SOURCE_PATH}"
+    OPTIONS
+        ${ARCH_OPTIONS}
+        ${FEATURE_OPTIONS}
+        -DLLAMA_ACCELERATE=${TARGET_IS_APPLE}
+        -DLLAMA_METAL=${TARGET_IS_APPLE}
+        -DLLAMA_STATIC=${USE_STATIC}
+        -DLLAMA_BLAS=ON
+        ${BLAS_OPTIONS}
+        -DPKG_CONFIG_EXECUTABLE:FILEPATH="${PKGCONFIG}"
+        -DBUILD_COMMIT:STRING="99115f3fa654b593099c6719ad30e3f54ce231e1"
+        -DBUILD_NUMBER:STRING="1273"
+    OPTIONS_RELEASE
+        -DLLAMA_METAL_NDEBUG=ON
+)
+vcpkg_cmake_build(TARGET "llama" LOGFILE_BASE build-llama)
+vcpkg_cmake_install()
+vcpkg_cmake_config_fixup(CONFIG_PATH "lib/cmake/Llama" PACKAGE_NAME "Llama")
+vcpkg_copy_pdbs()
+
+vcpkg_copy_tools(TOOL_NAMES
+    baby-llama beam-search benchmark convert-llama2c-to-ggml embd-input-test embedding llama-bench
+    main perplexity quantize-stats quantize save-load-state server simple speculative train-text-from-scratch    
+    AUTO_CLEAN
+)
+if("test" IN_LIST FEATURES)
+    vcpkg_copy_tools(TOOL_NAMES
+        test-grad0 test-grammar-parser test-llama-grammar test-quantize-fns test-quantize-perf
+        test-sampling test-tokenizer-0-falcon test-tokenizer-0-llama test-tokenizer-1-llama
+        AUTO_CLEAN
+    )
+endif()
+
+file(INSTALL "${SOURCE_PATH}/llama.h" DESTINATION "${CURRENT_PACKAGES_DIR}/include")
+
+file(INSTALL    "${CURRENT_PACKAGES_DIR}/bin/convert.py"
+                "${CURRENT_PACKAGES_DIR}/bin/convert-lora-to-ggml.py"
+    DESTINATION "${CURRENT_PACKAGES_DIR}/tools/${PORT}"
+)
+
+file(REMOVE_RECURSE
+    "${CURRENT_PACKAGES_DIR}/debug/include"
+)
+file(REMOVE
+    "${CURRENT_PACKAGES_DIR}/bin/convert.py"
+    "${CURRENT_PACKAGES_DIR}/debug/bin/convert.py"
+    "${CURRENT_PACKAGES_DIR}/bin/convert-lora-to-ggml.py"
+    "${CURRENT_PACKAGES_DIR}/debug/bin/convert-lora-to-ggml.py"
+)
+
+vcpkg_install_copyright(FILE_LIST "${SOURCE_PATH}/LICENSE")

ports/llama-cpp/vcpkg.json

@@ -0,0 +1,51 @@
+{
+  "name": "llama-cpp",
+  "version-string": "b1273",
+  "description": "Port of Facebook's LLaMA model in C/C++",
+  "homepage": "https://github.com/ggerganov/llama.cpp",
+  "supports": "windows | linux | osx",
+  "dependencies": [
+    {
+      "name": "openblas",
+      "platform": "windows"
+    },
+    {
+      "name": "vcpkg-cmake",
+      "host": true
+    },
+    {
+      "name": "vcpkg-cmake-config",
+      "host": true
+    }
+  ],
+  "features": {
+    "clblast": {
+      "description": "Use CLBlast",
+      "dependencies": [
+        "clblast"
+      ]
+    },
+    "cublas": {
+      "description": "Use CUDA",
+      "dependencies": [
+        "cudnn"
+      ]
+    },
+    "mpi": {
+      "description": "Use MPI",
+      "dependencies": [
+        {
+          "name": "msmpi",
+          "platform": "windows"
+        },
+        {
+          "name": "openmpi",
+          "platform": "!windows"
+        }
+      ]
+    },
+    "test": {
+      "description": "Build tests"
+    }
+  }
+}

ports/zlib-ng/portfile.cmake

@@ -5,6 +5,13 @@ vcpkg_from_github(
     SHA512 1c19a62bb00727ac49049c299fb70060da95b5fafa448144ae4133372ec8c3da15cef6c1303485290f269b23c580696554ca0383dba3e1f9609f65c332981988
     HEAD_REF master
 )
+if(VCPKG_TARGET_ARCHITECTURE STREQUAL "x64")
+    # list(APPEND ARCH_OPTIONS)
+elseif(VCPKG_TARGET_ARCHITECTURE STREQUAL "x86")
+    # list(APPEND ARCH_OPTIONS)
+elseif(VCPKG_TARGET_ARCHITECTURE STREQUAL "arm64")
+    # list(APPEND ARCH_OPTIONS)
+endif()
 
 vcpkg_check_features(OUT_FEATURE_OPTIONS FEATURE_OPTIONS
     FEATURES
@@ -14,7 +21,7 @@ vcpkg_check_features(OUT_FEATURE_OPTIONS FEATURE_OPTIONS
 vcpkg_cmake_configure(
     SOURCE_PATH "${SOURCE_PATH}"
     OPTIONS
-        ${FEATURE_OPTIONS}
+        ${ARCH_OPTIONS} ${FEATURE_OPTIONS}
         -DZLIB_ENABLE_TESTS=OFF
         -DWITH_NEW_STRATEGIES=ON
         -DWITH_NATIVE_INSTRUCTIONS=OFF # `-march=native` breaks `check_c_source_compiles`

test/vcpkg.json

@@ -13,6 +13,13 @@
       "name": "apple-nio-ssl",
       "platform": "osx | ios"
     },
+    {
+      "name": "llama-cpp",
+      "features": [
+        "test"
+      ],
+      "platform": "windows | osx"
+    },
     "openssl3",
     {
       "name": "vcpkg-cmake",