[Inductor] Update pytorch inductor workflow

.github/actions/install-dpcpp/action.yml

@@ -97,10 +97,12 @@ runs:
         echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" | \
           sudo tee /etc/apt/sources.list.d/oneAPI.list
         sudo apt update
-        sudo apt install -y intel-oneapi-runtime-libs intel-oneapi-compiler-dpcpp-cpp
+        sudo apt install -y intel-oneapi-runtime-libs intel-oneapi-compiler-dpcpp-cpp intel-oneapi-mkl-devel
         . /opt/intel/oneapi/setvars.sh
+        . /opt/intel/oneapi/mkl/latest/env/vars.sh
         cat <<EOF | tee setvars.sh
         . /opt/intel/oneapi/setvars.sh
+        . /opt/intel/oneapi/mkl/latest/env/vars.sh
         export CC=icx
         export CXX=icpx
         export C_INCLUDE_PATH=$CMPLR_ROOT/include:$C_INCLUDE_PATH

.github/workflows/cutlass_inductor_test.yml

@@ -1,6 +1,10 @@
 name: "TorchInductor CUTLASS UTs"
 
 on:
+  push:
+    branches: [ "main" ]
+  pull_request:
+    branches: [ "main" ]
   schedule:
     # Runs at 00:00 AM UTC alternate days
     - cron: '0 0 * * 2,4,6'
@@ -30,22 +34,33 @@ jobs:
           GPU: 'BMG'
           IGC: 'ROLLING'
 
-      - name: Install MKL
+      - name: Install DPC++
+        id: install_dpcpp
+        uses: ./.github/actions/install-dpcpp
+        with:
+          DPCPP_RELEASE: RELEASE
+          DPCPP_PATH: ~/dpcpp
+
+      - name: Setup environment
         shell: bash
         run: |
-          export PT_DLE_DIR=pt_ut_dle
-          export ONEAPI_INSTALL_DIR_FOR_PT=$HOME/$PT_DLE_DIR
-          if [ -d "$ONEAPI_INSTALL_DIR_FOR_PT" ]; then
-            echo "Directory exists. Listing contents:"
-            ls "$ONEAPI_INSTALL_DIR_FOR_PT"
-          fi
-          rm -rf $HOME/intel/installercache/instances/pt_oneapi
-          wget -q https://registrationcenter-download.intel.com/akdlm/IRC_NAS/b3e6c1bf-a6d5-4580-8b1d-80cbfd38c8af/intel-deep-learning-essentials-2025.3.2.36_offline.sh
-          bash intel-deep-learning-essentials-2025.3.2.36_offline.sh -a --silent --eula accept --instance=pt_oneapi --install-dir=$ONEAPI_INSTALL_DIR_FOR_PT
-          source $ONEAPI_INSTALL_DIR_FOR_PT/mkl/latest/env/vars.sh
-          source $ONEAPI_INSTALL_DIR_FOR_PT/setvars.sh
-          export TORCH_WHL=$HOME/pytorch_ut/torch-2.13.0a0+git0002a87-cp312-cp312-linux_x86_64.whl
+          . setvars.sh
+          export CC=icx
+          export CXX=icpx
+          export CUTLASS_SYCL_PROFILING_ENABLED=ON
+          export ONEAPI_DEVICE_SELECTOR=level_zero:gpu
+          export CMAKE_BUILD_TYPE=Release
+          export IGC_VISAOptions="-perfmodel"
+          export IGC_VectorAliasBBThreshold=100000000000
+          export IGC_ExtraOCLOptions="-cl-intel-256-GRF-per-thread"
+          export SYCL_PROGRAM_COMPILE_OPTIONS="-ze-opt-large-register-file -gline-tables-only"
+          # Persist environment variables to following steps
           env >> $GITHUB_ENV
+          echo $PATH
+          echo $MKLROOT
+          which $CXX
+          $CXX --version
+          sycl-ls
 
       - name: Setup virtual environment
         shell: bash
@@ -64,10 +79,7 @@ jobs:
           source ~/.pytorch-venv/bin/activate
           python3 --version
           echo $TORCH_WHL
-          #python3 -m pip install --pre torch==2.13.0.dev20260521+xpu --index-url https://download.pytorch.org/whl/nightly/xpu --force-reinstall
-          cp $TORCH_WHL .
-          ls ./
-          python3 -m pip install torch-2.13.0a0+git0002a87-cp312-cp312-linux_x86_64.whl
+          python3 -m pip install torch==2.14.0.dev20260811+xpu --index-url https://download.pytorch.org/whl/nightly/xpu --force-reinstall
 
       - name: Clone PyTorch fork and install requirements
         shell: bash
@@ -76,7 +88,7 @@ jobs:
           cd ${{ runner.temp }}
           git clone https://github.com/pytorch/pytorch.git
           cd pytorch
-          git checkout 0002a8783abe6623eacc71df6c05d72869cd5ba5
+          git log -1 
           python3 -m pip install -r requirements.txt
           python3 -m pip install unittest-xml-reporting
           python3 -m pip install pytest
@@ -140,6 +152,7 @@ jobs:
         shell: bash
         run: |
           rm -rf /tmp/torchinductor_*
+          echo $PATH
           source ~/.pytorch-venv/bin/activate
           export USE_XPU=1
           export CUTLASS_USE_SYCL=1
@@ -179,8 +192,6 @@ jobs:
         run: |
           rm -rf ${{ runner.temp }}/pytorch
           rm -rf $HOME/.pytorch-venv
-          echo $ONEAPI_INSTALL_DIR_FOR_PT
-          rm -rf $ONEAPI_INSTALL_DIR_FOR_PT