CI test

Author: mieshkiwrk

benchmarks/triton_kernels_benchmark/gemm_benchmark.py

@@ -472,7 +472,7 @@ def get_benchmark(
     The benchmark can then be executed by calling the :code:`.run` method on the return value.
     """
     supported_providers = {
-        'gluon': 'Gluon',
+        #'gluon': 'Gluon',
         'triton': 'Triton',
         'onednn': 'OneDNN',
     }

third_party/intel/backend/compiler.py

@@ -29,7 +29,7 @@ class XPUOptions:
     num_ctas: int = 1
     num_stages: int = 2
     cluster_dims: tuple = (1, 1, 1)
-    warp_size: int = 16  #32 # TODO:[mdziado]
+    warp_size: int = 32  # TODO:[mdziado]
     optimize_epilogue: bool = False
     enable_fp_fusion: bool = True
     launch_cooperative_grid: bool = False