Skip to content

fix the hanging bug #355

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

Merged
Yael-X merged 1 commit into from
Jan 31, 2026
Merged

fix the hanging bug #355

Show file tree
Hide file tree
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
2 changes: 1 addition & 1 deletion csrc/deepep/ops/op_host/cam_moe_combine_normal_tiling.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,7 @@ constexpr size_t MAX_GROUP_NAME_LENGTH = 128UL;
constexpr int64_t MAX_EP_WORLD_SIZE = 384;
constexpr int64_t MIN_EP_WORLD_SIZE = 2;
constexpr int64_t MAX_TP_WORLD_SIZE = 2;
constexpr int64_t BS_UPPER_BOUND = 32768;
constexpr int64_t BS_UPPER_BOUND = 65536;
Comment thread
luanyundu marked this conversation as resolved.

constexpr uint32_t SYSTEM_NEED_WORKSPACE = 16 * 1024 * 1024;
constexpr int32_t HCCL_BUFFER_SIZE_DEFAULT = 200 * 1024 * 1024; // Bytes
Expand Down
2 changes: 1 addition & 1 deletion csrc/deepep/ops/op_host/cam_moe_dispatch_normal_tiling.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -64,7 +64,7 @@ constexpr size_t MAX_GROUP_NAME_LENGTH = 128UL;
constexpr int64_t MAX_EP_WORLD_SIZE = 384;
constexpr int64_t MIN_EP_WORLD_SIZE = 2;
constexpr int64_t MAX_TP_WORLD_SIZE = 2;
constexpr int64_t BS_UPPER_BOUND = 32768; // 最大bs
constexpr int64_t BS_UPPER_BOUND = 65536; // 最大bs
Comment thread
luanyundu marked this conversation as resolved.

constexpr uint32_t TILINGKEY_TP_WORLD_SIZE = 100;
constexpr uint32_t TP_WORLD_SIZE_TWO = 2;
Expand Down
10 changes: 5 additions & 5 deletions csrc/deepep/ops/op_kernel/cam_moe_combine_normal_multi_round.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -635,15 +635,15 @@ __aicore__ inline void CamMoeCombineNormalMultiRound<TemplateMC2TypeFunc>::Proce
if ASCEND_IS_AIV { // 全aiv处理
uint32_t realRound = (realMaxBs_ + perRoundTokens_ - 1) / perRoundTokens_;
while (roundIndex_ < realRound) {
if (roundIndex_ >= WAIT_ROUND_INDEX) {
CopyBufferToShareAndSetStatus();
ReadBufferFromRemote();
if (realRound > 1) {
SetRoundStatus();
WaitRoundStatus();
roundMagic_ = roundMagic_ == 0 ? 1 : 0;
SyncAll<true>();
}
CopyBufferToShareAndSetStatus();
ReadBufferFromRemote();
SetRoundStatus();
roundIndex_ += 1;
roundMagic_ = roundMagic_ == 0 ? 1 : 0;
}
Comment thread
Yael-X marked this conversation as resolved.
}
}
Expand Down