Merge branch 'main' into addchanges

Author: rishi-yadav

.github/actions/install-intel-graphics/action.yml

@@ -57,9 +57,7 @@ runs:
         which sudo || alias sudo=""
         if [[ "${{ inputs.GPU }}" == "PVC" ]]; then
           # Purge existing intel graphics packages
-          sudo apt-get purge -y 'libigc*' 'intel-igc*' 'libigdgmm*' 'libigdfcl*' \
-          'libze-intel-gpu*' 'libze1' 'libze-dev' 'intel-opencl-icd' \
-          'intel-metrics-discovery' 'intel-gsc' 'intel-ocloc' || true
+          dpkg -l | grep -E 'libigc|intel-igc|libigdgmm|libigdfcl|libze-intel-gpu|libze1|libze-dev|intel-opencl-icd|intel-metrics-discovery|intel-gsc|intel-ocloc' | awk '{print $2}' | xargs -r sudo apt-get purge -y || true
           sudo rm -f /etc/apt/sources.list.d/kobuk-team-ubuntu-intel-graphics-*
           sudo rm -f /etc/apt/sources.list.d/intel-gpu-*.list
           
@@ -74,14 +72,9 @@ runs:
           echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-graphics.gpg] https://repositories.intel.com/gpu/ubuntu ${VERSION_CODENAME}/lts/2523 unified" | \
             sudo tee /etc/apt/sources.list.d/intel-gpu-${VERSION_CODENAME}.list
           sudo apt update
-          sudo apt install -y \
-              intel-opencl-icd libze-intel-gpu1 libze1 \
-              intel-media-va-driver-non-free libmfx-gen1 libvpl2 \
-              libegl-mesa0 libegl1-mesa-dev libgbm1 libgl1-mesa-dev libgl1-mesa-dri \
-              libglapi-mesa libgles2-mesa-dev libglx-mesa0 libigdgmm12 libxatracker2 mesa-va-drivers \
-              mesa-vdpau-drivers mesa-vulkan-drivers va-driver-all vainfo hwinfo clinfo
-          
-          
+          sudo apt-get install -y libze-intel-gpu1 libze-dev intel-metrics-discovery \
+            intel-opencl-icd ocl-icd-opencl-dev clinfo intel-gsc intel-ocloc g++
+          dpkg -l | grep libhwloc15 > /dev/null || sudo apt install -y libhwloc15
           rm -rf igc
           mkdir igc
           cd igc

.github/workflows/pvc_nightly.yml

@@ -26,7 +26,7 @@ jobs:
 
   gpp-host-test:
     needs: sycl-test
-    uses: ./.github/workflows/intel_test_gpp.yml
+    uses: ./.github/workflows/intel_test_gpp_host_reusable.yml
     with:
       compiler: '["RELEASE"]'
       intel_graphics: '["LTS"]'

.github/workflows/xe_benchmarks_ondemand.yml

@@ -92,6 +92,7 @@ jobs:
       - name: Run Benchmarks
         shell: bash
         run: |
+          export no_proxy=$no_proxy,${{ secrets.INFLUX_SERVER_IP }}
           export INFLUX_URL=${{ secrets.INFLUX_URL }}
           export INFLUX_ORG="cutlass"
           export INFLUX_BUCKET="benchmarks"
@@ -102,7 +103,6 @@ jobs:
         with:
           name: benchmark-artifacts
           path: logs/
-          retention-days: 3
       - name: Cleanup DPC++
         if: always()
         shell: bash

examples/06_bmg_flash_attention/06_xe_fmha_fwd.cpp

@@ -94,10 +94,10 @@ int main(int argc, const char **argv) {
   using SubgroupLayoutQK = Layout<Shape<_8, _1, _1>>;
 
 #elif HEAD_DIM == 128
-  using ShapeQK = Shape<_128, _32, _32>;
-  using ShapePV = Shape<_128, _32, _32>;
-  using ShapeOut = Shape<_128, _128>;
-  using SubgroupLayoutQK = Layout<Shape<_8, _1, _1>>;
+  using ShapeQK = Shape<_256, _32, _32>;
+  using ShapePV = Shape<_256, _32, _32>;
+  using ShapeOut = Shape<_256, _128>;
+  using SubgroupLayoutQK = Layout<Shape<_16, _1, _1>>;
 
 #elif HEAD_DIM == 192
   using ShapeQK = Shape<_256, _64, _32>;

examples/cute/tutorial/CMakeLists.txt

@@ -66,6 +66,16 @@ if (CUTLASS_ENABLE_SYCL)
       cute_tutorial_xe_gemm_slm
       xe_gemm_slm.cpp
     )
+
+    cutlass_example_add_executable(
+      cute_tutorial_xe_gemm_subgroup_specialization_slm
+      xe_gemm_subgroup_specialization_slm.cpp
+    )
+
+    cutlass_example_add_executable(
+      cute_tutorial_xe_two_gemm_fusion
+      xe_two_gemm_fusion.cpp
+    )
   endif()
 
   if (SYCL_NVIDIA_TARGET)