Update dispatch_ffn_combine_kernel.hpp

LQLlulu · web-flow · commit a47adaa1a58c · 2026-02-01T00:15:22.000+08:00
Signed-off-by: LQLlulu &lt;39671654+LQLlulu@users.noreply.github.com&gt;
diff --git a/csrc/dispatch_ffn_combine/op_kernel/dispatch_ffn_combine_kernel.hpp b/csrc/dispatch_ffn_combine/op_kernel/dispatch_ffn_combine_kernel.hpp
@@ -688,9 +688,10 @@ class DispatchFFNCombineKernel {
                 }
             }
 
+
             AscendC::SyncAll<true>();
             AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(syncgmm1Idx / CROSS_CORE_FLAG_MAX_SET_COUNT);   // V notifies C that the current communication round is complete            prevGroupSum1 += currentM;
-
+            prevGroupSum1 += currentM;
             //第一次swglu的token数以及截断逻辑
             if (groupIdx + 1 <= params.epilogueGranularity) {
                 if (dequantSum1 + currentM <= params.maxOutputSize) {

Original file line number	Diff line number	Diff line change
`@@ -688,9 +688,10 @@ class DispatchFFNCombineKernel {`
`688`	`688`	`}`
`689`	`689`	`}`
`690`	`690`
	`691`	`+`
`691`	`692`	`AscendC::SyncAll<true>();`
`692`	`693`	`AscendC::CrossCoreSetFlag<0x2, PIPE_MTE3>(syncgmm1Idx / CROSS_CORE_FLAG_MAX_SET_COUNT); // V notifies C that the current communication round is complete prevGroupSum1 += currentM;`
`693`		`-`
	`694`	`+ prevGroupSum1 += currentM;`
`694`	`695`	`//第一次swglu的token数以及截断逻辑`
`695`	`696`	`if (groupIdx + 1 <= params.epilogueGranularity) {`
`696`	`697`	`if (dequantSum1 + currentM <= params.maxOutputSize) {`