Add new gemm tuned cfg

Author: yudongsi

benchmarks/triton_kernels_benchmark/gemm_benchmark.py

@@ -98,10 +98,18 @@ def matmul_kernel_with_block_pointers(
         triton.Config(
             {'BLOCK_SIZE_M': 256, 'BLOCK_SIZE_N': 128, 'BLOCK_SIZE_K': 32, 'GROUP_SIZE_M': 4, 'grf_mode': 'large'},
             num_stages=s, num_warps=32) for s in [2]
+    ] + [
+        triton.Config(
+            {'BLOCK_SIZE_M': 128, 'BLOCK_SIZE_N': 1024, 'BLOCK_SIZE_K': 16, 'GROUP_SIZE_M': 1, 'grf_mode': 'large'},
+            num_stages=s, num_warps=32) for s in [2]
     ] + [
         triton.Config(
             {'BLOCK_SIZE_M': 64, 'BLOCK_SIZE_N': 128, 'BLOCK_SIZE_K': 32, 'GROUP_SIZE_M': 4, 'grf_mode': 'large'},
             num_stages=s, num_warps=32) for s in [2]
+    ] + [
+        triton.Config(
+            {'BLOCK_SIZE_M': 16, 'BLOCK_SIZE_N': 1024, 'BLOCK_SIZE_K': 64, 'GROUP_SIZE_M': 32, 'grf_mode': 'large'},
+            num_stages=s, num_warps=32) for s in [2, 3]
     ] + [
         triton.Config(
             {'BLOCK_SIZE_M': 8, 'BLOCK_SIZE_N': 512, 'BLOCK_SIZE_K': 64, 'GROUP_SIZE_M': 1, 'grf_mode': 'large'},