From fdb41efbd3cc1e409c7f514315b0199121df86eb Mon Sep 17 00:00:00 2001 From: Chenggang Zhao Date: Mon, 23 Jun 2025 17:58:44 +0800 Subject: [PATCH] Fix bugs --- csrc/kernels/internode.cu | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/csrc/kernels/internode.cu b/csrc/kernels/internode.cu index c3f14bb..1e0a28e 100644 --- a/csrc/kernels/internode.cu +++ b/csrc/kernels/internode.cu @@ -470,7 +470,7 @@ dispatch(int4* recv_x, float* recv_x_scales, int64_t* recv_topk_idx, float* recv // Read RDMA rank existence uint64_t is_token_in_rank_uint64 = 0; if (lane_id < kNumRDMARanks) { - is_token_in_rank_uint64 = *reinterpret_cast(is_token_in_rank + token_idx * num_ranks + lane_id * NUM_MAX_NVL_PEERS); + is_token_in_rank_uint64 = __ldg(reinterpret_cast(is_token_in_rank + token_idx * num_ranks + lane_id * NUM_MAX_NVL_PEERS)); global_rdma_tail_idx += (is_token_in_rank_uint64 != 0); } __syncwarp();