diff --git a/tests/fmha_v3_stage_c_example9.py b/tests/fmha_v3_stage_c_example9.py index 19ac1128..27f4f261 100644 --- a/tests/fmha_v3_stage_c_example9.py +++ b/tests/fmha_v3_stage_c_example9.py @@ -217,10 +217,12 @@ class FmhaV3StageCMulti: cute.copy(tma_q, tAgQ[(None, Int32(0))], tAsQ[(None, qh.index)], tma_bar_ptr=qh.barrier) qp.tail() kvp.reset() + kv_coord = Int32(0) for kt in cutlass.range(0, n_kv_tiles, 1, unroll=1): kvh = kvp.acquire_and_advance() - cute.copy(tma_k, tBgK[(None, kvh.count)], tBsK[(None, kvh.index)], tma_bar_ptr=kvh.barrier) - cute.copy(tma_v, tVgV[(None, kvh.count)], tVsV[(None, kvh.index)], tma_bar_ptr=kvh.barrier) + cute.copy(tma_k, tBgK[(None, kv_coord)], tBsK[(None, kvh.index)], tma_bar_ptr=kvh.barrier) + cute.copy(tma_v, tVgV[(None, kv_coord)], tVsV[(None, kvh.index)], tma_bar_ptr=kvh.barrier) + kv_coord += 1 kvp.tail() # ===== MMA warp =====