From 42c1e4815a7ef0b3d6f3cd1b892969fbda3d720b Mon Sep 17 00:00:00 2001
From: hujiawen <hujiawen5@hisilicon.com>
Date: Sat, 2 Aug 2025 11:02:51 +0800
Subject: [PATCH] fix mm leakyrelu block24 aicore error

---
 .../op_host/matmul_leakyrelu_custom.cpp                |  4 ++--
 .../op_kernel/matmul_leakyrelu_custom.cpp              | 10 ++++++----
 2 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_host/matmul_leakyrelu_custom.cpp b/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_host/matmul_leakyrelu_custom.cpp
index b811c71ca..5e5da7928 100644
--- a/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_host/matmul_leakyrelu_custom.cpp
+++ b/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_host/matmul_leakyrelu_custom.cpp
@@ -60,8 +60,8 @@ static ge::graphStatus TilingFunc(gert::TilingContext *context)
         context->SetTilingKey(2);
     } else {
         /* SetBlockDim here refers to the number of cube cores, so for separated arch(AIC:AIV=1:2), 
-            vector cores number is set 48 by SetDim, cube core number need to be set 24 here.*/ 
-        context->SetBlockDim(24);
+           when vector cores number is set like 48 by SetDim, cube core number need to be set 24 here.*/ 
+        context->SetBlockDim(ascendcPlatform.GetCoreNumAic());
         context->SetTilingKey(1);
     }
     tiling.SaveToBuffer(context->GetRawTilingData()->GetData(), context->GetRawTilingData()->GetCapacity());
diff --git a/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_kernel/matmul_leakyrelu_custom.cpp b/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_kernel/matmul_leakyrelu_custom.cpp
index 4f56aca86..e1894f29f 100644
--- a/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_kernel/matmul_leakyrelu_custom.cpp
+++ b/operator/ascendc/0_introduction/12_matmulleakyrelu_frameworklaunch/MatmulLeakyReluCustom/op_kernel/matmul_leakyrelu_custom.cpp
@@ -147,11 +147,13 @@ template <typename aType, typename bType, typename cType, typename biasType>
 __aicore__ inline void MatmulLeakyKernel<aType, bType, cType, biasType>::CopyOut(uint32_t count)
 {
     reluOutQueue_.DeQue<cType>();
-    const uint32_t roundM = tiling.singleCoreM / tiling.baseM;
-    const uint32_t roundN = tiling.singleCoreN / tiling.baseN;
+    const uint32_t roundM = Ceiling(tiling.singleCoreM, tiling.baseM);
+    const uint32_t curCopyM = tiling.singleCoreM < tiling.baseM ? tiling.singleCoreM : tiling.baseM;
+    const uint32_t curCopyN = tiling.singleCoreN < tiling.baseN ? tiling.singleCoreN : tiling.baseN;
+    const uint32_t curCopyNStride = tiling.N < tiling.baseN ? tiling.N : tiling.N - tiling.baseN;
     uint32_t startOffset = (count % roundM * tiling.baseM * tiling.N + count / roundM * tiling.baseN);
-    DataCopyParams copyParam = {(uint16_t)tiling.baseM, (uint16_t)(tiling.baseN * sizeof(cType) / DEFAULT_C0_SIZE), 0,
-                                (uint16_t)((tiling.N - tiling.baseN) * sizeof(cType) / DEFAULT_C0_SIZE)};
+    DataCopyParams copyParam = {(uint16_t)curCopyM, (uint16_t)(curCopyN * sizeof(cType) / 32), 0,
+                                (uint16_t)(curCopyNStride * sizeof(cType) / 32)};
     DataCopy(cGlobal[startOffset], reluOutLocal, copyParam);
     reluOutQueue_.FreeTensor(reluOutLocal);
 }
-- 
Gitee