fix(build): update LLAMA build flags for windows-gpu and disable unnecessary build steps

Author: Gioee

.github/workflows/main.yml

@@ -104,8 +104,8 @@ jobs:
           - os: windows-latest
             arch: x86_64
             name: windows-gpu
-            make: LLAMA="-DGGML_BACKEND_DL=ON -DGGML_NATIVE=OFF -DGGML_CPU=OFF -DGGML_CUDA=ON"
-            make-cuda: LLAMA_ARGS="--target ggml-cuda" LLAMA="-G \"Ninja Multi-Config\" -DGGML_BACKEND_DL=ON -DGGML_NATIVE=OFF -DGGML_CPU=OFF -DGGML_CUDA=ON -DGGML_HIP=ON -DGGML_SYCL=ON"
+            make: LLAMA="-G \"Ninja Multi-Config\" -DGGML_BACKEND_DL=ON -DGGML_NATIVE=OFF -DGGML_CPU=OFF -DGGML_CUDA=ON -DGGML_HIP=ON -DGGML_SYCL=ON"
+            make-cuda: LLAMA_ARGS="--target ggml-cuda" LLAMA="-G \"Ninja Multi-Config\" -DGGML_NATIVE=OFF -DGGML_CPU=OFF -DGGML_CUDA=ON"
             make-hip: LLAMA_ARGS="--target ggml-hip" LLAMA="-G \"Ninja\" -DGGML_BACKEND_DL=ON -DGGML_NATIVE=OFF -DGGML_CPU=OFF -DGGML_HIP=ON -DGGML_HIP_ROCWMMA_FATTN=ON -DAMDGPU_TARGETS=\"gfx1100;gfx1101;gfx1102;gfx1030;gfx1031;gfx1032\" -DCMAKE_C_COMPILER=\"%HIP_PATH%/bin/clang.exe\" -DCMAKE_CXX_COMPILER=\"%HIP_PATH%/bin/clang++.exe\" -DCMAKE_CXX_FLAGS=\"-I../../rocwmma/library/include/ -Wno-ignored-attributes -Wno-nested-anon-types\""
             make-sycl: LLAMA_ARGS="--target ggml-sycl" LLAMA="-G \"Ninja\" -DGGML_BACKEND_DL=ON -DCMAKE_C_COMPILER=cl -DCMAKE_CXX_COMPILER=icx -DGGML_NATIVE=OFF -DGGML_CPU=OFF -DGGML_SYCL=ON"
 
@@ -338,14 +338,14 @@ jobs:
           key: miniaudio-${{ matrix.name }}-${{ matrix.os }}-${{ matrix.arch }}-${{ steps.submodule-hashes.outputs.miniaudio }}-${{ hashFiles('modules/miniaudio/**') }}-${{ steps.submodule-hashes.outputs.make }}
 
       - name: build ggml-sycl
-        if: matrix.name == 'windows-gpu' && steps.cache-llama.outputs.cache-hit != 'true'
+        if: false #matrix.name == 'windows-gpu' && steps.cache-llama.outputs.cache-hit != 'true'
         shell: cmd
         run: |
           call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
           make build/llama.cpp.stamp ${{ matrix.make-sycl }}
 
       - name: build ggml-hip (msvc)
-        if: matrix.name == 'windows-gpu' && steps.cache-llama.outputs.cache-hit != 'true'
+        if: false #matrix.name == 'windows-gpu' && steps.cache-llama.outputs.cache-hit != 'true'
         shell: cmd
         run: |
           call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" x64
@@ -355,7 +355,7 @@ jobs:
           make build/llama.cpp.stamp ${{ matrix.make-hip }}
 
       - name: build ggml-cuda (msvc)
-        if: matrix.name == 'windows-gpu' && steps.cache-llama.outputs.cache-hit != 'true'
+        if: false #matrix.name == 'windows-gpu' && steps.cache-llama.outputs.cache-hit != 'true'
         shell: cmd
         run: |
           call "C:\Program Files\Microsoft Visual Studio\2022\Enterprise\VC\Auxiliary\Build\vcvarsall.bat" x64