wip4

slaren · Dec 1, 2024 · cc5a750 · cc5a750
1 parent edddcaf
commit cc5a750
Showing 1 changed file with 34 additions and 47 deletions.
diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
@@ -33,40 +33,31 @@ env:
   LLAMA_LOG_TIMESTAMPS: 1
 
 jobs:
-  windows-msys2-cpu-x86_64:
+  windows-mingw-cpu-x86_64:
     runs-on: windows-latest
 
     strategy:
       fail-fast: false
       matrix:
-        arch: [avx, avx2, avx512, amx]
         include:
-          - { sys: UCRT64, env: ucrt-x86_64, build: Release }
-          - { arch: avx,    defines: '-DGGML_AVX=ON -DGGML_AVX2=OFF -DGGML_AVX512=OFF -DGGML_AMX_TILE=OFF -DGGML_AMX_INT8=OFF' }
-          - { arch: avx2,   defines: '-DGGML_AVX=ON -DGGML_AVX2=ON  -DGGML_AVX512=OFF -DGGML_AMX_TILE=OFF -DGGML_AMX_INT8=OFF' }
-          - { arch: avx512, defines: '-DGGML_AVX=ON -DGGML_AVX2=ON  -DGGML_AVX512=ON  -DGGML_AMX_TILE=OFF -DGGML_AMX_INT8=OFF' }
-          - { arch: amx,    defines: '-DGGML_AVX=ON -DGGML_AVX2=ON  -DGGML_AVX512=ON  -DGGML_AMX_TILE=ON  -DGGML_AMX_INT8=ON' }
+          - { arch: sse,    defines: '-DGGML_FMA=OFF -DGGML_AVX=OFF -DGGML_AVX2=OFF -DGGML_AVX_VNNI=OFF -DGGML_F16C=OFF -DGGML_AVX512=OFF -DGGML_AVX512_VNNI=OFF -DGGML_AMX_TILE=OFF -DGGML_AMX_INT8=OFF' }
+          - { arch: avx,    defines: '-DGGML_FMA=ON  -DGGML_AVX=ON  -DGGML_AVX2=OFF -DGGML_AVX_VNNI=OFF -DGGML_F16C=ON  -DGGML_AVX512=OFF -DGGML_AVX512_VNNI=OFF -DGGML_AMX_TILE=OFF -DGGML_AMX_INT8=OFF' }
+          - { arch: avx2,   defines: '-DGGML_FMA=ON  -DGGML_AVX=ON  -DGGML_AVX2=ON  -DGGML_AVX_VNNI=ON  -DGGML_F16C=ON  -DGGML_AVX512=OFF -DGGML_AVX512_VNNI=OFF -DGGML_AMX_TILE=OFF -DGGML_AMX_INT8=OFF' }
+          - { arch: avx512, defines: '-DGGML_FMA=ON  -DGGML_AVX=ON  -DGGML_AVX2=ON  -DGGML_AVX_VNNI=ON  -DGGML_F16C=ON  -DGGML_AVX512=ON  -DGGML_AVX512_VNNI=ON  -DGGML_AMX_TILE=OFF -DGGML_AMX_INT8=OFF' }
+          - { arch: amx,    defines: '-DGGML_FMA=ON  -DGGML_AVX=ON  -DGGML_AVX2=ON  -DGGML_AVX_VNNI=ON  -DGGML_F16C=ON  -DGGML_AVX512=ON  -DGGML_AVX512_VNNI=ON  -DGGML_AMX_TILE=ON  -DGGML_AMX_INT8=ON'  }
 
     steps:
       - name: Clone
         uses: actions/checkout@v4
 
-      - name: Setup ${{ matrix.sys }}
-        uses: msys2/setup-msys2@v2
-        with:
-          update: true
-          msystem: ${{matrix.sys}}
-          install: >-
-            base-devel
-            mingw-w64-${{matrix.env}}-toolchain
-            mingw-w64-${{matrix.env}}-cmake
-            mingw-w64-${{matrix.env}}-ninja
+      - name: Install Ninja
+        run: |
+          choco install ninja
 
       - name: Build using CMake
-        shell: msys2 {0}
         run: |
             cmake -B build -G "Ninja Multi-Config" -DGGML_NATIVE=OFF -DBUILD_SHARED_LIBS=ON -DGGML_BACKEND_DL=ON ${{ matrix.defines }}
-            cmake --build build --config ${{ matrix.build }} -t ggml-cpu
+            cmake --build build --config Release -t ggml-cpu
 
       - name: Rename artifacts
         run: mv build/bin/Release/ggml-cpu.dll ggml-cpu-${{ matrix.arch }}.dll
@@ -78,43 +69,39 @@ jobs:
           path: ggml-cpu-${{ matrix.arch }}.dll
           name: ggml-cpu-${{ matrix.arch }}.dll
 
-  windows-msys2-x86_64:
+  windows-mingw-x86_64:
     runs-on: windows-latest
 
     needs:
-      - windows-msys2-cpu-x86_64
-
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - { sys: UCRT64, env: ucrt-x86_64, build: Release }
+      - windows-mingw-cpu-x86_64
 
     steps:
       - name: Clone
         uses: actions/checkout@v4
         with:
           fetch-depth: 0
 
-      - name: Setup ${{ matrix.sys }}
-        uses: msys2/setup-msys2@v2
-        with:
-          update: true
-          msystem: ${{matrix.sys}}
-          install: >-
-            base-devel
-            git
-            mingw-w64-${{matrix.env}}-toolchain
-            mingw-w64-${{matrix.env}}-cmake
-            mingw-w64-${{matrix.env}}-ninja
-            mingw-w64-${{matrix.env}}-curl
+      - name: Install Ninja
+        run: |
+          dir c:/mingw64
+          dir c:/mingw64/bin
+          choco install ninja
+
+      - name: Download curl for Windows
+        run: |
+          curl.exe -O https://curl.se/windows/dl-8.11.0_4/curl-8.11.0_4-win64-mingw.zip
+          7z x curl-8.11.0_4-win64-mingw.zip -ocurl
 
       - name: Build using CMake
-        shell: msys2 {0}
         run: |
-            cmake -B build -G "Ninja Multi-Config" -DLLAMA_CURL=ON -DLLAMA_BUILD_TESTS=OFF -DBUILD_SHARED_LIBS=ON -DGGML_BACKEND_DL=ON
-            cmake --build build --config ${{ matrix.build }}
-            rm build/bin/Release/ggml-cpu.dll
+          cmake -B build -G "Ninja Multi-Config" -DLLAMA_CURL=ON -DLLAMA_BUILD_TESTS=OFF -DBUILD_SHARED_LIBS=ON -DGGML_BACKEND_DL=ON -DGGML_NATIVE=OFF -DCURL_INCLUDE_DIR="curl/curl-8.11.0_4-win64-mingw/include/" -DCURL_LIBRARY="curl/curl-8.11.0_4-win64-mingw/lib/libcurl.dll.a"
+          cmake --build build --config Release -t llama-bench -t llama-cli -t llama-gguf-hash -t llama-gguf-split -t llama-imatrix -t llama-perplexity -t llama-quantize -t llama-simple -t llama-simple-chat -t llama-tokenize
+          rm build/bin/Release/ggml-cpu.dll
+          cp curl/curl-8.11.0_4-win64-mingw/bin/libcurl-x64.dll build/bin/Release
+          cp c:/mingw64/bin/libwinpthread-1.dll build/bin/Release
+          cp c:/mingw64/bin/libstdc++-6.dll build/bin/Release
+          cp c:/mingw64/bin/libgcc_s_seh-1.dll build/bin/Release
+          cp c:/mingw64/bin/libgomp-1.dll build/bin/Release
 
       - name: Download CPU backend artifacts
         uses: actions/download-artifact@v4
@@ -141,23 +128,23 @@ jobs:
         if: ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
         run: |
           Copy-Item LICENSE .\build\bin\Release\llama.cpp.txt
-          7z a llama-${{ steps.tag.outputs.name }}-bin-msys2-cpu-${{ matrix.build }}.zip .\build\bin\Release\*
+          7z a llama-${{ steps.tag.outputs.name }}-bin-mingw-cpu.zip .\build\bin\Release\*
 
       - name: Upload artifacts
         id: upload_artifacts
         if: ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
         uses: actions/upload-artifact@v4
         with:
-          path: llama-${{ steps.tag.outputs.name }}-bin-msys2-cpu-${{ matrix.build }}.zip
-          name: llama-${{ steps.tag.outputs.name }}-bin-msys2-cpu-${{ matrix.build }}.zip
+          path: llama-${{ steps.tag.outputs.name }}-bin-mingw-cpu.zip
+          name: llama-${{ steps.tag.outputs.name }}-bin-mingw-cpu.zip
 
   release:
     if: ${{ ( github.event_name == 'push' && github.ref == 'refs/heads/master' ) || github.event.inputs.create_release == 'true' }}
 
     runs-on: ubuntu-latest
 
     needs:
-      - windows-msys2-x86_64
+      - windows-mingw-x86_64
 
     steps:
       - name: Clone