Skip to content

Allow using few SMs for low-latency mode #277

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Allow using few SMs for low-latency mode #277

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 6 additions & 2 deletions csrc/kernels/internode_ll.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -339,7 +339,9 @@ void dispatch(void* packed_recv_x, void* packed_recv_x_scales,
void* workspace, int num_device_sms,
cudaStream_t stream, int phases) {
constexpr int kNumMaxTopK = 9;
const int num_warp_groups = ceil_div(num_experts, num_device_sms);
const int num_warp_groups = ((phases & LOW_LATENCY_RECV_PHASE) == 0)
? 9
: ceil_div(num_experts, num_device_sms);
const int num_warps_per_group = 32 / num_warp_groups;
EP_HOST_ASSERT(num_warp_groups > 0 and num_warps_per_group > 0);
EP_HOST_ASSERT(kNumMaxTopK + 1 <= num_warp_groups * num_warps_per_group);
Expand Down Expand Up @@ -548,7 +550,9 @@ void combine(void* combined_x,
void* workspace, int num_device_sms,
cudaStream_t stream, int phases, bool zero_copy) {
constexpr int kNumMaxTopk = 9;
const int num_warp_groups = ceil_div(num_experts, num_device_sms);
const int num_warp_groups = ((phases & LOW_LATENCY_RECV_PHASE) == 0)
? 9
: ceil_div(num_experts, num_device_sms);
const int num_warps_per_group = 32 / num_warp_groups;
EP_HOST_ASSERT(num_warp_groups > 0 and num_warps_per_group > 0);

Expand Down