diff --git a/impl/matmul/scheduler/bias/bias_scheduler_base.h b/impl/matmul/scheduler/bias/bias_scheduler_base.h
index 295aaee8407a8f89416faf745ab1c78b5ae75da1..0005d79bf20962173d9cfa0c9d864915b28aa9ac 100644
--- a/impl/matmul/scheduler/bias/bias_scheduler_base.h
+++ b/impl/matmul/scheduler/bias/bias_scheduler_base.h
@@ -32,7 +32,7 @@ class BiasSchedulerBase
 {
     MATMUL_USE_MODULE(MatmulShapeTiling);
     using BiasT = typename BIAS_TYPE::T;
-    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (!MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
+    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (PhyPosIsUB(BIAS_TYPE::pos) && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
         && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1Singleshape())), GlobalTensor<BiasT>, LocalTensor<BiasT>>::type;
 
 public:
diff --git a/impl/matmul/scheduler/bias/bias_scheduler_intf.h b/impl/matmul/scheduler/bias/bias_scheduler_intf.h
index 40ff3fa2b939e1f5aa1157d2592304c2d55a6afe..7fea10bf4503a1b95aa8a7ea8d0793caf12f565c 100644
--- a/impl/matmul/scheduler/bias/bias_scheduler_intf.h
+++ b/impl/matmul/scheduler/bias/bias_scheduler_intf.h
@@ -28,7 +28,7 @@ namespace Detail {
 template <typename IMPL, class A_TYPE, class B_TYPE, class BIAS_TYPE, const auto &MM_CFG, typename = void>
 class BiasScheduler {
     using BiasT = typename BIAS_TYPE::T;
-    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (!MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
+    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (PhyPosIsUB(BIAS_TYPE::pos) && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
         && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1Singleshape())), GlobalTensor<BiasT>, LocalTensor<BiasT>>::type;
 
 public:
diff --git a/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_batch.h b/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_batch.h
index 7146a59eecbcfd33bc0d539fdf3e5f9f060cdaa2..6c5c0995be11a0772e405846339ca265bd695ae0 100644
--- a/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_batch.h
+++ b/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_batch.h
@@ -35,7 +35,7 @@ class CopyBiasIn<IMPL, A_TYPE, BIAS_TYPE, MM_CFG, enable_if_t<
     ToMatmulConfig(MM_CFG).batchMode != BatchMode::SINGLE_LARGE_THAN_L1 &&
     (PhyPosIsUB(BIAS_TYPE::pos) || PhyPosIsGM(BIAS_TYPE::pos))>> {
     using BiasT = typename BIAS_TYPE::T;
-    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()),
+    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (PhyPosIsUB(BIAS_TYPE::pos) && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1())),
                                          GlobalTensor<BiasT>, LocalTensor<BiasT>>::type;
 
 public:
diff --git a/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_intf.h b/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_intf.h
index 38cbb34d4682eb32810ae4c9aa7fcf353b967d57..3b7179908b758bc1ed328c86325a91debf570660 100644
--- a/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_intf.h
+++ b/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_intf.h
@@ -28,7 +28,7 @@ namespace Detail {
 template <typename IMPL, class A_TYPE, class BIAS_TYPE, const auto &MM_CFG, typename = void>
 class CopyBiasIn {
     using BiasT = typename BIAS_TYPE::T;
-    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (!MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
+    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (PhyPosIsUB(BIAS_TYPE::pos) && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
         && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1Singleshape())), GlobalTensor<BiasT>, LocalTensor<BiasT>>::type;
 public:
     __aicore__ inline CopyBiasIn() = default;
diff --git a/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_v220.h b/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_v220.h
index 492203476dee61b391566624ec8670fe3cf30bc8..c701517336e0166c16d0e05f97eeb802d7351078 100644
--- a/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_v220.h
+++ b/impl/matmul/stage/copy_cube_in/bias/copy_bias_in_v220.h
@@ -41,7 +41,7 @@ class CopyBiasIn<IMPL, A_TYPE, BIAS_TYPE, MM_CFG, enable_if_t<
     MATMUL_USE_MODULE(NLoop);
 
     using BiasT = typename BIAS_TYPE::T;
-    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (!MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
+    using TensorT = typename Conditional<(PhyPosIsGM(BIAS_TYPE::pos) || (PhyPosIsUB(BIAS_TYPE::pos) && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1()
         && !MatmulFeatureTrait<MM_CFG>::IsSupportUBToL1Singleshape())), GlobalTensor<BiasT>, LocalTensor<BiasT>>::type;
 
 public:
diff --git a/impl/matmul/stage/copy_cube_out/copy_cube_out_fixpipe.h b/impl/matmul/stage/copy_cube_out/copy_cube_out_fixpipe.h
index 9835a4856e6dd5ccafb4840ba07dc175c1488a36..6ccd5947fdd4707a59493aa3f2edab1c91661ae5 100644
--- a/impl/matmul/stage/copy_cube_out/copy_cube_out_fixpipe.h
+++ b/impl/matmul/stage/copy_cube_out/copy_cube_out_fixpipe.h
@@ -178,18 +178,21 @@ private:
         if constexpr (MatmulFeatureTrait<MM_CFG>::IsSupportL0CToUB() && PhyPosIsUB(C_TYPE::pos) &&
             (A_TYPE::ibShare && B_TYPE::ibShare)) {
             if constexpr (FIXPIPE_MODE != McgShfMode::DUAL_DST_SPLIT_N) {
-                dstOffset = (static_cast<int64_t> (static_cast<int64_t>(
-                    curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM() * stride)) >> 1) +
+                dstOffset = (static_cast<int64_t> (static_cast<int64_t>(static_cast<int64_t>(
+                    curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM()) * stride)) >> 1) +
                 static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN());
             } else {
                 dstOffset =
-                    static_cast<int64_t>(curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM() * stride) +
-                    static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN() * baseHeight);
+                    static_cast<int64_t>(static_cast<int64_t>(
+                        curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM()) * stride) +
+                    static_cast<int64_t>(static_cast<int64_t>(
+                        curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) * baseHeight);
                 dstOffset = dstOffset >> 1;
             }
         } else {
             dstOffset =
-                static_cast<int64_t>(curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM() * stride)+
+                static_cast<int64_t>(static_cast<int64_t>(
+                    curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM()) * stride)+
                 static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN());
         }
         return dstOffset;
@@ -221,7 +224,8 @@ private:
         if constexpr (!enSequentialWrite) {
             stride = static_cast<uint32_t>(GetOrgM<IS_INTRA_BLOCK>() * CHANNEL_MERGE_FACTOR * BLOCK_CUBE);
             if constexpr (!IsBasic(MM_CFG)) {
-                dstOffset = static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) * GetOrgM<IS_INTRA_BLOCK>() +
+                dstOffset = static_cast<int64_t>(static_cast<int64_t>(
+                    curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) * GetOrgM<IS_INTRA_BLOCK>()) +
                     static_cast<int64_t>(curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM()) * CHANNEL_MERGE_FACTOR * BLOCK_CUBE;
             }
             if constexpr (PhyPosIsUB(C_TYPE::pos) && ((A_TYPE::ibShare && B_TYPE::ibShare) ||
@@ -274,8 +278,9 @@ private:
         if constexpr (!enSequentialWrite) {
             stride = GetOrgHeight<IS_INTRA_BLOCK>();
             if constexpr (!IsBasic(MM_CFG)) {
-                dstOffset = static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN() *
-                    stride)+ static_cast<int64_t>(curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM());
+                dstOffset = static_cast<int64_t>(static_cast<int64_t>(
+                    curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) * stride) +
+                    static_cast<int64_t>(curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM());
             }
         }
         FixpipeAdaptor fixpipe(baseWidth, baseHeight, baseBlockWidth, baseBlockHeight,
@@ -293,11 +298,13 @@ private:
         uint32_t stride;
         if constexpr (!enSequentialWrite) {
             if constexpr (!ToMatmulConfig(MM_CFG).isEnableChannelSplit) {
-                dstOffset = static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) * GetOrgM<IS_INTRA_BLOCK>() +
+                dstOffset = static_cast<int64_t>(static_cast<int64_t>(
+                    curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) * GetOrgM<IS_INTRA_BLOCK>()) +
                     static_cast<int64_t>(curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM()) * BLOCK_CUBE;
                 stride = static_cast<uint32_t>(GetOrgM<IS_INTRA_BLOCK>() * BLOCK_CUBE * sizeof(DstT) / ONE_BLK_SIZE);
             } else {
-                dstOffset = static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) * Ceil(GetOrgM<IS_INTRA_BLOCK>(), BLOCK_CUBE) * BLOCK_CUBE +
+                dstOffset = static_cast<int64_t>(static_cast<int64_t>(curCol * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseN()) *
+                    Ceil(GetOrgM<IS_INTRA_BLOCK>(), BLOCK_CUBE)) * BLOCK_CUBE +
                     static_cast<int64_t>(curRow * MATMUL_MODULE(MatmulShapeTiling)->GetTiling().GetBaseM()) * B32_C0SIZE;
                 stride = static_cast<uint32_t>(Ceil(GetOrgM<IS_INTRA_BLOCK>() , BLOCK_CUBE) * BLOCK_CUBE *
                                                B32_C0SIZE * sizeof(DstT) / ONE_BLK_SIZE);
@@ -455,12 +462,14 @@ private:
             if constexpr (C_TYPE::format == CubeFormat::ND || C_TYPE::format == CubeFormat::ND_ALIGN) {
                 dstStrideIn = GetOrgWidth();
                 nSize = static_cast<uint16_t>(baseWidth);
-                dstOffset = static_cast<int64_t>(static_cast<int64_t>(curRow * baseM) * dstStrideIn) + static_cast<int64_t>(curCol * baseN);
+                dstOffset = static_cast<int64_t>(static_cast<int64_t>(curRow * baseM) * dstStrideIn) +
+                    static_cast<int64_t>(curCol * baseN);
             } else if constexpr (C_TYPE::format == CubeFormat::NZ) {
                 dstStrideIn = static_cast<uint32_t>((MATMUL_MODULE(MatmulShapeInfo)->GetOrgM()) *
                     BLOCK_CUBE * sizeof(DstT) / ONE_BLK_SIZE);
                 nSize = 0;
-                dstOffset = curCol * baseN * MATMUL_MODULE(MatmulShapeInfo)->GetOrgM() + curRow * baseM * BLOCK_CUBE;
+                dstOffset = static_cast<uint16_t>(static_cast<uint16_t>(curCol * baseN) * MATMUL_MODULE(MatmulShapeInfo)->GetOrgM()) +
+                    static_cast<uint16_t>(curRow * baseM * BLOCK_CUBE);
             }
         }
     }
diff --git a/impl/matmul/tiling/matmul_tiling_algorithm.cpp b/impl/matmul/tiling/matmul_tiling_algorithm.cpp
index fad57699c91237139ee336eeb72e97ac5dd8f55e..14145e00883bc2bf5b0cab57b2f4fdc532e75898 100644
--- a/impl/matmul/tiling/matmul_tiling_algorithm.cpp
+++ b/impl/matmul/tiling/matmul_tiling_algorithm.cpp
@@ -3045,8 +3045,8 @@ void MatmulTilingAlgorithm::CheckL0DB(SingleCoreStatus& singleCoreStatus, const
         tilingIns_->bType_.scalePos == TPosition::TSCM) {
         baseN = MathUtil::Align(singleCoreStatus.l0Status.nL0, L0_FACTOR_NUM_LIMIT) * C0_SIZE;
     }
-    const uint32_t aTypeSize = DTYPE_BIT_TAB.at(tilingIns_->aType_.dataType);
-    const uint32_t bTypeSize = DTYPE_BIT_TAB.at(tilingIns_->bType_.dataType);
+    const int32_t aTypeSize = DTYPE_BIT_TAB.at(tilingIns_->aType_.dataType);
+    const int32_t bTypeSize = DTYPE_BIT_TAB.at(tilingIns_->bType_.dataType);
     if ((baseM * baseK * aTypeSize / BITS_PER_BYTE) > (tilingIns_->bufferPool_.l0ASize / DB_ON)) {
         singleCoreStatus.l0Status.dbL0A = DB_OFF;
     }