Set one_matrix_per_load_for_bt to False

Signed-off-by: Whitney Tsang <[email protected]>

Author: whitneywhtsang

benchmarks/triton_kernels_benchmark/flash_attention_fwd_benchmark.py

@@ -157,7 +157,7 @@ def _attn_fwd(Q, K, V, sm_scale, M, Out,  #
 
 
 configs = [
-    triton.Config({'BLOCK_M': BM, 'BLOCK_N': BN, 'grf_mode': 'large', 'one_matrix_per_load_for_bt': True}, num_stages=s, num_warps=w) \
+    triton.Config({'BLOCK_M': BM, 'BLOCK_N': BN, 'grf_mode': 'large', 'one_matrix_per_load_for_bt': False}, num_stages=s, num_warps=w) \
     for BM in [128, 256] \
     for BN in [32, 64] \
     for s in [3, 4] \