EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

fix: resolve sync bug in DispathFFNCombine when expert num per card is 32 (#6416)

Browse Source 
### What this PR does / why we need it?

Fix the synchronization deadlock issue in DispathFFNCombine module that
occurs on NPU cards when the number of experts per card exceeds 16 (the
bug manifests prominently when set to 32/128).

### Does this PR introduce _any_ user-facing change?
No, this is a bug fix for internal synchronization logic specific to NPU
expert dispatch, with no impact on external APIs, interfaces, or
end-user behaviors.


- vLLM version: v0.14.1
- vLLM main:
dc917cceb8

Signed-off-by: xulei_ict <xulei292@huawei.com>
Co-authored-by: xulei_ict <xulei292@huawei.com>
This commit is contained in:
xulei
2026-01-30 21:21:20 +08:00
committed by GitHub
parent 56f5d3bd49
commit 77ea873224
4 changed files with 69 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

47
csrc/dispatch_ffn_combine/op_kernel/dispatch_ffn_combine_kernel.hpp
View File

@@ -42,6 +42,10 @@ using namespace AscendC;
namespace Catlass::Gemm::Kernel { namespace Catlass::Gemm::Kernel {
constexpr uint16_t SYNCFLAGC2V = 9;
constexpr uint16_t SYNCFLAGV2C = 10;
constexpr uint16_t CROSS_CORE_FLAG_MAX_SET_COUNT = 15;
template < template <
class BlockMmad_, class BlockMmad_,
class BlockScheduler_, class BlockScheduler_,
@@ -189,7 +193,7 @@ public:
{ {
GMM1(params); GMM1(params);
AscendC::CrossCoreWaitFlag<0x2>(2); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGV2C);
GMM2(params); GMM2(params);
} }
@@ -201,7 +205,7 @@ public:
{ {
Dispatch(params); Dispatch(params);
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(2); AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(SYNCFLAGV2C);
Combine(params); Combine(params);
} }
@@ -324,9 +328,12 @@ private:
int64_t gmGroupOffsetC = 0; int64_t gmGroupOffsetC = 0;
uint32_t startCoreIdx = 0; uint32_t startCoreIdx = 0;
uint32_t syncGroupIdx = 0; uint32_t syncGroupIdx = 0;
AscendC::CrossCoreWaitFlag<0x2>(0); // Wait for AIV to finish cumsum for matmul
int64_t preCurrentmSum = 0; int64_t preCurrentmSum = 0;
int32_t syncLoopIdx = -1; int32_t syncLoopIdx = -1;
uint16_t syncgmmIdx = 0;
AscendC::CrossCoreWaitFlag<0x2>(syncgmmIdx / CROSS_CORE_FLAG_MAX_SET_COUNT); // Wait for AIV to finish cumsum for matmul
syncgmmIdx++;
constexpr uint32_t MAX_EXPERTS_PER_RANK = 32; constexpr uint32_t MAX_EXPERTS_PER_RANK = 32;
__gm__ ElementB* weight1Array[MAX_EXPERTS_PER_RANK]; __gm__ ElementB* weight1Array[MAX_EXPERTS_PER_RANK];
@@ -370,7 +377,8 @@ private:
for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) { for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) {
for(;syncGroupIdx <= groupIdx; syncGroupIdx++) { for(;syncGroupIdx <= groupIdx; syncGroupIdx++) {
AscendC::CrossCoreWaitFlag<0x2>(0); AscendC::CrossCoreWaitFlag<0x2>(syncgmmIdx / CROSS_CORE_FLAG_MAX_SET_COUNT);
syncgmmIdx ++;
} }
// Compute block location // Compute block location
GemmCoord blockCoord = blockScheduler.GetBlockCoord(loopIdx); GemmCoord blockCoord = blockScheduler.GetBlockCoord(loopIdx);
@@ -399,7 +407,7 @@ private:
if constexpr (BlockMmad::DispatchPolicy::ASYNC) { if constexpr (BlockMmad::DispatchPolicy::ASYNC) {
blockMmad.SynchronizeBlock(); blockMmad.SynchronizeBlock();
} }
blockMmad.Finalize(syncLoopIdx, 1); blockMmad.Finalize(syncLoopIdx, SYNCFLAGC2V);
} }
preCurrentmSum += currentM; preCurrentmSum += currentM;
@@ -410,10 +418,16 @@ private:
gmGroupOffsetC += inGroupProblemShape.m() * inGroupProblemShape.n(); gmGroupOffsetC += inGroupProblemShape.m() * inGroupProblemShape.n();
startCoreIdx = (startCoreIdx + coreLoops) % coreNum; startCoreIdx = (startCoreIdx + coreLoops) % coreNum;
} }
for(;syncGroupIdx < params.expertPerRank; syncGroupIdx++) {
AscendC::CrossCoreWaitFlag<0x2>(syncgmmIdx / CROSS_CORE_FLAG_MAX_SET_COUNT);
syncgmmIdx ++;
}
if constexpr (BlockMmad::DispatchPolicy::ASYNC) { if constexpr (BlockMmad::DispatchPolicy::ASYNC) {
blockMmad.SynchronizeBlock(); blockMmad.SynchronizeBlock();
} }
blockMmad.Finalize(syncLoopIdx + 1, 1); blockMmad.Finalize(syncLoopIdx + 1, SYNCFLAGC2V);
} }
CATLASS_DEVICE CATLASS_DEVICE
@@ -482,7 +496,7 @@ private:
uint32_t startLoopIdx = ((coreIdx < startCoreIdx) ? (coreIdx + coreNum) : coreIdx) - startCoreIdx; uint32_t startLoopIdx = ((coreIdx < startCoreIdx) ? (coreIdx + coreNum) : coreIdx) - startCoreIdx;
// Loop through the matmul of each groupIdx // Loop through the matmul of each groupIdx
if (params.expertPerRank > lastDequantExpertNum && groupIdx + 1 == params.expertPerRank - lastDequantExpertNum) { if (params.expertPerRank > lastDequantExpertNum && groupIdx + 1 == params.expertPerRank - lastDequantExpertNum) {
AscendC::CrossCoreWaitFlag<0x2>(2); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGV2C);
} }
for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) { for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) {
if (loopIdx + coreNum >= coreLoops) { if (loopIdx + coreNum >= coreLoops) {
@@ -508,7 +522,7 @@ private:
gmB2[gmGroupOffsetB + gmOffsetB], layoutB2, gmB2[gmGroupOffsetB + gmOffsetB], layoutB2,
gmC2[gmGroupOffsetC + gmOffsetC], layoutC, gmC2[gmGroupOffsetC + gmOffsetC], layoutC,
gmS2[gmOffsetS], layoutScale, gmS2[gmOffsetS], layoutScale,
actualBlockShape, syncLoopIdx, 3 actualBlockShape, syncLoopIdx, 0
); );
} }
} }
@@ -526,7 +540,7 @@ private:
if constexpr (BlockMmad::DispatchPolicy::ASYNC) { if constexpr (BlockMmad::DispatchPolicy::ASYNC) {
blockMmad.SynchronizeBlock(); blockMmad.SynchronizeBlock();
} }
blockMmad.Finalize(params.expertPerRank - 1, 3); blockMmad.Finalize(params.expertPerRank - 1, 0);
} }
CATLASS_DEVICE CATLASS_DEVICE
@@ -625,7 +639,9 @@ private:
GetCumsumForMMAIV(tokenPerExpert, cumsumMM, params.expertPerRank, params.rank, params.EP); GetCumsumForMMAIV(tokenPerExpert, cumsumMM, params.expertPerRank, params.rank, params.EP);
} }
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(0); uint16_t syncgmm1Idx = 0;
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(syncgmm1Idx / CROSS_CORE_FLAG_MAX_SET_COUNT);
syncgmm1Idx++;
uint32_t curGroupOffset = 0; uint32_t curGroupOffset = 0;
int32_t prevSumBeforeRank = 0; int32_t prevSumBeforeRank = 0;
@@ -673,10 +689,11 @@ private:
if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1) { if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1) {
syncLoopIdx++; syncLoopIdx++;
AscendC::CrossCoreWaitFlag<0x2>(syncLoopIdx / 8 + 1); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGC2V);
} }
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(0); // V notifies C that the current communication round is complete AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(syncgmm1Idx / CROSS_CORE_FLAG_MAX_SET_COUNT); // V notifies C that the current communication round is complete
syncgmm1Idx++;
if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1 && prevGroupSum1 > 0) { if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1 && prevGroupSum1 > 0) {
uint32_t rowStartThisCore = 0; uint32_t rowStartThisCore = 0;
@@ -699,7 +716,7 @@ private:
} }
} }
syncLoopIdx ++; syncLoopIdx ++;
AscendC::CrossCoreWaitFlag<0x2>(syncLoopIdx /8 + 1); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGC2V);
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
uint32_t lastDequantExpertNum = params.expertPerRank; uint32_t lastDequantExpertNum = params.expertPerRank;
@@ -707,7 +724,7 @@ private:
lastDequantExpertNum = params.expertPerRank - params.epilogueGranularity; lastDequantExpertNum = params.expertPerRank - params.epilogueGranularity;
} }
if (lastDequantExpertNum < params.expertPerRank) { if (lastDequantExpertNum < params.expertPerRank) {
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(2); AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(SYNCFLAGV2C);
} }
if (prevGroupSum1 - dequantSum < params.maxOutputSize) { if (prevGroupSum1 - dequantSum < params.maxOutputSize) {
uint32_t rowStartThisCore = prevGroupSum1 - dequantSum;; uint32_t rowStartThisCore = prevGroupSum1 - dequantSum;;
@@ -746,7 +763,7 @@ private:
BlockEpilogue2 blockEpilogue(resource, epilogueParams); BlockEpilogue2 blockEpilogue(resource, epilogueParams);
int32_t prevGroupSum2 = 0; int32_t prevGroupSum2 = 0;
for (uint32_t groupIdx = 0; groupIdx < params.expertPerRank; ++groupIdx) { for (uint32_t groupIdx = 0; groupIdx < params.expertPerRank; ++groupIdx) {
AscendC::CrossCoreWaitFlag<0x2>(groupIdx / 8 + 3); AscendC::CrossCoreWaitFlag<0x2>(groupIdx / CROSS_CORE_FLAG_MAX_SET_COUNT);
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
for(int32_t dstEpIdx = coreIdx; dstEpIdx < params.EP; dstEpIdx += coreNum) { for(int32_t dstEpIdx = coreIdx; dstEpIdx < params.EP; dstEpIdx += coreNum) {

4
csrc/dispatch_ffn_combine/op_kernel/utils/block_mmad_preload_async_fixpipe_quant.hpp
View File

@@ -22,6 +22,8 @@
namespace Catlass::Gemm::Block { namespace Catlass::Gemm::Block {
constexpr uint16_t CROSS_CORE_FLAG_MAX_SET_COUNT = 15;
template<AscendC::HardEvent event> template<AscendC::HardEvent event>
__aicore__ inline void SyncFlagFunc(int32_t eventID) __aicore__ inline void SyncFlagFunc(int32_t eventID)
{ {
@@ -271,7 +273,7 @@ public:
void Finalize(int32_t target, int32_t flag = 0) void Finalize(int32_t target, int32_t flag = 0)
{ {
for(;syncGroupIdx <= target; syncGroupIdx++) { for(;syncGroupIdx <= target; syncGroupIdx++) {
int32_t flagId = syncGroupIdx / 8 + flag; int32_t flagId = syncGroupIdx / CROSS_CORE_FLAG_MAX_SET_COUNT + flag;
AscendC::CrossCoreSetFlag<0x2, PIPE_FIX>(flagId); AscendC::CrossCoreSetFlag<0x2, PIPE_FIX>(flagId);
} }
} }

46
csrc/dispatch_ffn_combine_bf16/op_kernel/dispatch_ffn_combine_bf16_kernel.hpp
View File

@@ -42,6 +42,10 @@ using namespace AscendC;
namespace Catlass::Gemm::Kernel { namespace Catlass::Gemm::Kernel {
constexpr uint16_t SYNCFLAGC2V = 9;
constexpr uint16_t SYNCFLAGV2C = 10;
constexpr uint16_t CROSS_CORE_FLAG_MAX_SET_COUNT = 15;
template < template <
class BlockMmad_, class BlockMmad_,
class BlockScheduler_, class BlockScheduler_,
@@ -198,7 +202,7 @@ public:
{ {
GMM1(params); GMM1(params);
AscendC::CrossCoreWaitFlag<0x2>(2); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGV2C);
GMM2(params); GMM2(params);
} }
@@ -210,7 +214,7 @@ public:
{ {
Dispatch(params); Dispatch(params);
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(2); AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(SYNCFLAGV2C);
CombineV2(params); CombineV2(params);
} }
@@ -507,7 +511,9 @@ CATLASS_DEVICE
int64_t gmGroupOffsetC = 0; int64_t gmGroupOffsetC = 0;
uint32_t startCoreIdx = 0; uint32_t startCoreIdx = 0;
uint32_t syncGroupIdx = 0; uint32_t syncGroupIdx = 0;
AscendC::CrossCoreWaitFlag<0x2>(0); uint16_t syncgmmIdx = 0;
AscendC::CrossCoreWaitFlag<0x2>(syncgmmIdx / CROSS_CORE_FLAG_MAX_SET_COUNT); // Wait for AIV to finish cumsum for matmul
syncgmmIdx++;
AicSyncAll(); AicSyncAll();
int64_t preCurrentmSum = 0; int64_t preCurrentmSum = 0;
int32_t syncLoopIdx = -1; int32_t syncLoopIdx = -1;
@@ -553,7 +559,8 @@ CATLASS_DEVICE
for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) { for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) {
for(;syncGroupIdx <= groupIdx; syncGroupIdx++) { for(;syncGroupIdx <= groupIdx; syncGroupIdx++) {
AscendC::CrossCoreWaitFlag<0x2>(0); AscendC::CrossCoreWaitFlag<0x2>(syncgmmIdx / CROSS_CORE_FLAG_MAX_SET_COUNT);
syncgmmIdx ++;
} }
// Compute block location // Compute block location
GemmCoord blockCoord = blockScheduler.GetBlockCoord(loopIdx); GemmCoord blockCoord = blockScheduler.GetBlockCoord(loopIdx);
@@ -592,7 +599,7 @@ CATLASS_DEVICE
if constexpr (BlockMmad::DispatchPolicy::ASYNC) { if constexpr (BlockMmad::DispatchPolicy::ASYNC) {
blockMmad.SynchronizeBlock(); blockMmad.SynchronizeBlock();
} }
blockMmad.Finalize(syncLoopIdx, 1); blockMmad.Finalize(syncLoopIdx, SYNCFLAGC2V);
} }
preCurrentmSum += currentM; preCurrentmSum += currentM;
@@ -603,10 +610,16 @@ CATLASS_DEVICE
gmGroupOffsetC += inGroupProblemShape.m() * inGroupProblemShape.n(); gmGroupOffsetC += inGroupProblemShape.m() * inGroupProblemShape.n();
startCoreIdx = (startCoreIdx + coreLoops) % coreNum; startCoreIdx = (startCoreIdx + coreLoops) % coreNum;
} }
for(;syncGroupIdx < params.expertPerRank; syncGroupIdx++) {
AscendC::CrossCoreWaitFlag<0x2>(syncgmmIdx / CROSS_CORE_FLAG_MAX_SET_COUNT);
syncgmmIdx ++;
}
if constexpr (BlockMmad::DispatchPolicy::ASYNC) { if constexpr (BlockMmad::DispatchPolicy::ASYNC) {
blockMmad.SynchronizeBlock(); blockMmad.SynchronizeBlock();
} }
blockMmad.Finalize(syncLoopIdx + 1, 1); blockMmad.Finalize(syncLoopIdx + 1, SYNCFLAGC2V);
} }
CATLASS_DEVICE CATLASS_DEVICE
@@ -674,7 +687,7 @@ CATLASS_DEVICE
uint32_t startLoopIdx = ((coreIdx < startCoreIdx) ? (coreIdx + coreNum) : coreIdx) - startCoreIdx; uint32_t startLoopIdx = ((coreIdx < startCoreIdx) ? (coreIdx + coreNum) : coreIdx) - startCoreIdx;
// Loop through the matmul of each groupIdx // Loop through the matmul of each groupIdx
if (params.expertPerRank > lastDequantExpertNum && groupIdx + 1 == params.expertPerRank - lastDequantExpertNum) { if (params.expertPerRank > lastDequantExpertNum && groupIdx + 1 == params.expertPerRank - lastDequantExpertNum) {
AscendC::CrossCoreWaitFlag<0x2>(2); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGV2C);
} }
for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) { for (uint32_t loopIdx = startLoopIdx; loopIdx < coreLoops; loopIdx += coreNum) {
if (loopIdx + coreNum >= coreLoops) { if (loopIdx + coreNum >= coreLoops) {
@@ -701,7 +714,7 @@ CATLASS_DEVICE
gmB2[gmGroupOffsetB + gmOffsetB], layoutB2, gmB2[gmGroupOffsetB + gmOffsetB], layoutB2,
gmC2[gmGroupOffsetC + gmOffsetC], layoutC, gmC2[gmGroupOffsetC + gmOffsetC], layoutC,
gmS2[gmOffsetS], layoutScale, gmS2[gmOffsetS], layoutScale,
actualBlockShape, syncLoopIdx, 3 actualBlockShape, syncLoopIdx, 0
); );
} else { } else {
blockMmad( blockMmad(
@@ -709,7 +722,7 @@ CATLASS_DEVICE
gmB2[gmGroupOffsetB + gmOffsetB], layoutB2, gmB2[gmGroupOffsetB + gmOffsetB], layoutB2,
gmC2[gmGroupOffsetC + gmOffsetC], layoutC, gmC2[gmGroupOffsetC + gmOffsetC], layoutC,
gmS2, layoutScale, gmS2, layoutScale,
actualBlockShape, syncLoopIdx, 3 actualBlockShape, syncLoopIdx, 0
); );
} }
} }
@@ -749,7 +762,9 @@ CATLASS_DEVICE
GetCumsumForMMAIV(tokenPerExpert, cumsumMM, params.expertPerRank, params.rank, params.EP); GetCumsumForMMAIV(tokenPerExpert, cumsumMM, params.expertPerRank, params.rank, params.EP);
} }
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(0); uint16_t syncgmm1Idx = 0;
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(syncgmm1Idx / CROSS_CORE_FLAG_MAX_SET_COUNT);
syncgmm1Idx++;
uint32_t curGroupOffset = 0; uint32_t curGroupOffset = 0;
int32_t prevSumBeforeRank = 0; int32_t prevSumBeforeRank = 0;
@@ -791,10 +806,11 @@ CATLASS_DEVICE
if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1) { if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1) {
syncLoopIdx++; syncLoopIdx++;
AscendC::CrossCoreWaitFlag<0x2>(syncLoopIdx / 8 + 1); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGC2V);
} }
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(0); AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(syncgmm1Idx / CROSS_CORE_FLAG_MAX_SET_COUNT); // V notifies C that the current communication round is complete
syncgmm1Idx++;
if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1 && prevGroupSum1 > 0) { if ((params.epilogueGranularity < params.expertPerRank && params.epilogueGranularity > 0) && groupIdx == params.expertPerRank - 1 && prevGroupSum1 > 0) {
uint32_t rowStartThisCore = 0; uint32_t rowStartThisCore = 0;
@@ -829,9 +845,9 @@ CATLASS_DEVICE
lastDequantExpertNum = params.expertPerRank - params.epilogueGranularity; lastDequantExpertNum = params.expertPerRank - params.epilogueGranularity;
} }
if (lastDequantExpertNum < params.expertPerRank) { if (lastDequantExpertNum < params.expertPerRank) {
AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(2); AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(SYNCFLAGV2C);
} }
AscendC::CrossCoreWaitFlag<0x2>(syncLoopIdx /8 + 1); AscendC::CrossCoreWaitFlag<0x2>(SYNCFLAGC2V);
AscendC::SyncAll<true>(); AscendC::SyncAll<true>();
if (prevGroupSum1 - dequantSum < params.maxOutputSize) { if (prevGroupSum1 - dequantSum < params.maxOutputSize) {
uint32_t rowStartThisCore = prevGroupSum1 - dequantSum;; uint32_t rowStartThisCore = prevGroupSum1 - dequantSum;;
@@ -907,7 +923,7 @@ CATLASS_DEVICE
} }
if (loopIdx == startLoopIdx) { if (loopIdx == startLoopIdx) {
for (;syncLoopIdx <= groupIdx; syncLoopIdx++) { for (;syncLoopIdx <= groupIdx; syncLoopIdx++) {
int32_t flag_id = 3 + syncLoopIdx / 8; int32_t flag_id = syncLoopIdx / CROSS_CORE_FLAG_MAX_SET_COUNT;
AscendC::CrossCoreWaitFlag<0x2>(flag_id); AscendC::CrossCoreWaitFlag<0x2>(flag_id);
} }
} }

4
csrc/dispatch_ffn_combine_bf16/op_kernel/utils/block_mmad_preload_async_fixpipe_quant.hpp
View File

@@ -22,6 +22,8 @@
namespace Catlass::Gemm::Block { namespace Catlass::Gemm::Block {
constexpr uint16_t CROSS_CORE_FLAG_MAX_SET_COUNT = 15;
template<AscendC::HardEvent event> template<AscendC::HardEvent event>
__aicore__ inline void SyncFlagFunc(int32_t eventID) __aicore__ inline void SyncFlagFunc(int32_t eventID)
{ {
@@ -271,7 +273,7 @@ public:
void Finalize(int32_t target, int32_t flag = 0) void Finalize(int32_t target, int32_t flag = 0)
{ {
for(;syncGroupIdx <= target; syncGroupIdx++) { for(;syncGroupIdx <= target; syncGroupIdx++) {
int32_t flagId = syncGroupIdx / 8 + flag; int32_t flagId = syncGroupIdx / CROSS_CORE_FLAG_MAX_SET_COUNT + flag;
AscendC::CrossCoreSetFlag<0x2, PIPE_FIX>(flagId); AscendC::CrossCoreSetFlag<0x2, PIPE_FIX>(flagId);
} }
} }