ci : build with CLBlast + ggml-opencl use GGML_API (#1576)

* Build with CLBlast

* Declare GGML_API

After rebasing, examples/talk-llama failed:

"D:\a\whisper.cpp\whisper.cpp\build\ALL_BUILD.vcxproj" (build target) (1) ->
"D:\a\whisper.cpp\whisper.cpp\build\examples\talk-llama\talk-llama.vcxproj" (default target) (14) ->
(Link target) ->
  llama.obj : error LNK2019: unresolved external symbol ggml_cl_free_data referenced in function "public: __cdecl llama_model::~llama_model(void)" (??1llama_model@@QEAA@XZ) [D:\a\whisper.cpp\whisper.cpp\build\examples\talk-llama\talk-llama.vcxproj]
  llama.obj : error LNK2019: unresolved external symbol ggml_cl_transform_tensor referenced in function "public: void __cdecl llama_model_loader::load_all_data(struct ggml_context *,void (__cdecl*)(float,void *),void *,struct llama_mlock *)" (?load_all_data@llama_model_loader@@QEAAXPEAUggml_context@@P6AXMPEAX@Z1PEAUllama_mlock@@@Z) [D:\a\whisper.cpp\whisper.cpp\build\examples\talk-llama\talk-llama.vcxproj]
  D:\a\whisper.cpp\whisper.cpp\build\bin\Release\talk-llama.exe : fatal error LNK1120: 2 unresolved externals [D:\a\whisper.cpp\whisper.cpp\build\examples\talk-llama\talk-llama.vcxproj]

diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index 1d388d9e4599310c0f9a50d4e14b115f457e4b06..ae6c4ce9b450998d2b0eb7896402558c6e53174e 100644 (file)
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -223,9 +223,12 @@ jobs:
           - arch: Win32
             obzip: https://github.com/OpenMathLib/OpenBLAS/releases/download/v0.3.25/OpenBLAS-0.3.25-x86.zip
             s2arc: x86
+            clblast: OFF
           - arch: x64
             obzip: https://github.com/OpenMathLib/OpenBLAS/releases/download/v0.3.25/OpenBLAS-0.3.25-x64.zip
             s2arc: x64
+            clblast: ON
+            clver: 1.6.1
           - sdl2: ON
             s2ver: 2.28.5
 
@@ -252,6 +255,18 @@ jobs:
           7z x sdl2.zip
           echo "SDL2_DIR=$env:GITHUB_WORKSPACE/SDL2-${{ matrix.s2ver }}/cmake" >> $env:GITHUB_ENV
 
+      - name: Install OpenCL
+        if: matrix.clblast == 'ON'
+        run: vcpkg.exe --triplet=${{ matrix.arch }}-windows install opencl
+
+      - name: Fetch CLBlast and set CLBlast_DIR
+        if: matrix.clblast == 'ON'
+        run: |
+          C:/msys64/usr/bin/wget.exe -qO clblast.zip https://github.com/CNugteren/CLBlast/releases/download/${{ matrix.clver }}/CLBlast-${{ matrix.clver }}-windows-x64.zip
+          7z x clblast.zip
+          7z x CLBlast-${{ matrix.clver }}-windows-x64.7z
+          echo "CLBlast_DIR=$env:GITHUB_WORKSPACE/CLBlast-${{ matrix.clver }}-windows-x64/lib/cmake/CLBlast" >> $env:GITHUB_ENV
+
       - name: Configure
         run: >
           cmake -S . -B ./build -A ${{ matrix.arch }}
@@ -259,6 +274,7 @@ jobs:
           -DWHISPER_OPENBLAS=${{ matrix.blas }}
           -DCMAKE_LIBRARY_PATH="$env:OPENBLAS_PATH/lib"
           -DWHISPER_SDL2=${{ matrix.sdl2 }}
+          -DWHISPER_CLBLAST=${{ matrix.clblast }}
 
       - name: Build
         run: |
@@ -273,11 +289,15 @@ jobs:
         if: matrix.sdl2 == 'ON'
         run: copy "$env:SDL2_DIR/../lib/${{ matrix.s2arc }}/SDL2.dll" build/bin/${{ matrix.build }}
 
+      - name: Copy clblast.dll
+        if: matrix.clblast == 'ON'
+        run: copy "$env:CLBlast_DIR/../../clblast.dll" build/bin/${{ matrix.build }}
+
       - name: Upload binaries
         if: matrix.blas == 'ON' && matrix.sdl2 == 'ON'
         uses: actions/upload-artifact@v1
         with:
-          name: whisper-blas-bin-${{ matrix.arch }}
+          name: whisper-blas${{ matrix.clblast == 'ON' && '-clblast' || ''}}-bin-${{ matrix.arch }}
           path: build/bin/${{ matrix.build }}
 
   windows-cublas:

diff --git a/examples/CMakeLists.txt b/examples/CMakeLists.txt
index e6f837f3a774083c8d6745d0d4e68122aa3ca798..62b5b6fe495dcc40e68a0cde3cf052fe96a37a1d 100644 (file)
--- a/examples/CMakeLists.txt
+++ b/examples/CMakeLists.txt
@@ -14,6 +14,10 @@ if (WHISPER_SDL2)
     message(STATUS "SDL2_LIBRARIES = ${SDL2_LIBRARIES}")
 endif()
 
+if (WHISPER_CLBLAST)
+    find_package(CLBlast REQUIRED)
+endif()
+
 # common
 
 set(TARGET common)

diff --git a/examples/talk-llama/CMakeLists.txt b/examples/talk-llama/CMakeLists.txt
index e2130a4028a1f5aebbb0dd877f8448ad23450a7f..567db34499d444ce91e9a157e4877a5b7b8e4ed9 100644 (file)
--- a/examples/talk-llama/CMakeLists.txt
+++ b/examples/talk-llama/CMakeLists.txt
@@ -3,7 +3,11 @@ if (WHISPER_SDL2)
     set(TARGET talk-llama)
     add_executable(${TARGET} talk-llama.cpp llama.cpp)
     target_include_directories(${TARGET} PRIVATE ${SDL2_INCLUDE_DIRS})
-    target_link_libraries(${TARGET} PRIVATE common common-sdl whisper ${SDL2_LIBRARIES} ${CMAKE_THREAD_LIBS_INIT})
+
+    if (WHISPER_CLBLAST)
+        set(CLBLAST_LIBNAME clblast)
+    endif ()
+    target_link_libraries(${TARGET} PRIVATE common common-sdl whisper ${SDL2_LIBRARIES} ${CLBLAST_LIBNAME} ${CMAKE_THREAD_LIBS_INIT})
 
     if(WIN32)
         # It requires Windows 8.1 or later for PrefetchVirtualMemory

diff --git a/ggml-opencl.h b/ggml-opencl.h
index a92b445c9d7660da6ed5dfcbc4cce9ae7a5b9827..44d05bd64a3ad116c5ae416d1f106ce02998efd1 100644 (file)
--- a/ggml-opencl.h
+++ b/ggml-opencl.h
@@ -6,19 +6,19 @@
 extern "C" {
 #endif
 
-void ggml_cl_init(void);
+GGML_API void ggml_cl_init(void);
 
-void   ggml_cl_mul(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
-bool   ggml_cl_can_mul_mat(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
-size_t ggml_cl_mul_mat_get_wsize(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
-void   ggml_cl_mul_mat(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst, void * wdata, size_t wsize);
+GGML_API void   ggml_cl_mul(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
+GGML_API bool   ggml_cl_can_mul_mat(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
+GGML_API size_t ggml_cl_mul_mat_get_wsize(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst);
+GGML_API void   ggml_cl_mul_mat(const struct ggml_tensor * src0, const struct ggml_tensor * src1, struct ggml_tensor * dst, void * wdata, size_t wsize);
 
-void * ggml_cl_host_malloc(size_t size);
-void   ggml_cl_host_free(void * ptr);
+GGML_API void * ggml_cl_host_malloc(size_t size);
+GGML_API void   ggml_cl_host_free(void * ptr);
 
-void ggml_cl_free_data(const struct ggml_tensor* tensor);
+GGML_API void ggml_cl_free_data(const struct ggml_tensor* tensor);
 
-void ggml_cl_transform_tensor(void * data, struct ggml_tensor * tensor);
+GGML_API void ggml_cl_transform_tensor(void * data, struct ggml_tensor * tensor);
 
 #ifdef  __cplusplus
 }