DeepSeek/DeepEP
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepEP synced 2025-06-26 18:28:11 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Remove useless assertion

Browse Source
This commit is contained in:
Chenggang Zhao 2025-06-24 09:21:35 +08:00
parent bc118b248a
commit a15faa9ff0
1 changed files with 5 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
csrc/kernels/internode.cu
View File

@ -562,7 +562,7 @@ dispatch(int4* recv_x, float* recv_x_scales, int64_t* recv_topk_idx, float* recv
acquire_lock(rdma_send_channel_lock + lane_id); acquire_lock(rdma_send_channel_lock + lane_id);
// Release the transaction slot // Release the transaction slot
auto rdy_window = rdma_send_channel_window[lane_id]; auto window = rdma_send_channel_window[lane_id];
auto latest_tail = rdma_send_channel_tail[lane_id]; auto latest_tail = rdma_send_channel_tail[lane_id];
auto offset = rdma_tail_idx - latest_tail; auto offset = rdma_tail_idx - latest_tail;
@ -570,14 +570,13 @@ dispatch(int4* recv_x, float* recv_x_scales, int64_t* recv_topk_idx, float* recv
EP_STATIC_ASSERT(kNumDispatchRDMASenderWarps < 32, "Invalid warps"); EP_STATIC_ASSERT(kNumDispatchRDMASenderWarps < 32, "Invalid warps");
// Erase bit and move the ones if possible // Erase bit and move the ones if possible
rdy_window ^= 1u << offset; window ^= 1u << offset;
if (offset == 0) { if (offset == 0) {
EP_DEVICE_ASSERT(rdy_window & 1); auto num_empty_slots = __ffs(~window) - 1;
auto num_empty_slots = __ffs(~rdy_window) - 1;
st_release_cta(rdma_send_channel_tail + lane_id, latest_tail + num_empty_slots); st_release_cta(rdma_send_channel_tail + lane_id, latest_tail + num_empty_slots);
rdy_window >>= num_empty_slots; window >>= num_empty_slots;
} }
rdma_send_channel_window[lane_id] = rdy_window; rdma_send_channel_window[lane_id] = window;
// Release lock // Release lock
release_lock(rdma_send_channel_lock + lane_id); release_lock(rdma_send_channel_lock + lane_id);