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

Support zero-copy for low-latency combine

Browse Source
This commit is contained in:
Chenggang Zhao 2025-03-18 15:44:26 +08:00
parent dcaf73e5ff
commit 66465476ae
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
csrc/config.hpp
View File

@ -103,7 +103,7 @@ struct LowLatencyBuffer {
int* combine_rdma_recv_flag_buffer = nullptr; int* combine_rdma_recv_flag_buffer = nullptr;
void* combine_rdma_send_buffer_data_start = nullptr; void* combine_rdma_send_buffer_data_start = nullptr;
int num_bytes_per_combine_msg = 0; size_t num_bytes_per_combine_msg = 0;
std::pair<int*, int> clean_meta() { std::pair<int*, int> clean_meta() {
EP_HOST_ASSERT(dispatch_rdma_recv_count_buffer == combine_rdma_recv_flag_buffer); EP_HOST_ASSERT(dispatch_rdma_recv_count_buffer == combine_rdma_recv_flag_buffer);
@ -168,7 +168,7 @@ struct LowLatencyLayout {
advance(rdma_buffer, send_buffer_bytes * 2 + recv_buffer_bytes * i), advance(rdma_buffer, send_buffer_bytes * 2 + recv_buffer_bytes * i),
advance<int*>(rdma_buffer, send_buffer_bytes * 2 + recv_buffer_bytes * 2 + signaling_buffer_bytes * i), advance<int*>(rdma_buffer, send_buffer_bytes * 2 + recv_buffer_bytes * 2 + signaling_buffer_bytes * i),
advance(rdma_buffer, send_buffer_bytes * i + sizeof(int4)), advance(rdma_buffer, send_buffer_bytes * i + sizeof(int4)),
static_cast<int>(num_bytes_per_combine_msg) num_bytes_per_combine_msg
}; };
} }
} }