diff --git a/src/04kernel/include/kernel/attributes/transpose_info.h b/src/04kernel/include/kernel/attributes/transpose_info.h
index 00c49315..8c8a30bb 100644
--- a/src/04kernel/include/kernel/attributes/transpose_info.h
+++ b/src/04kernel/include/kernel/attributes/transpose_info.h
@@ -23,6 +23,8 @@ namespace refactor::kernel {
 
         TransposeInfo(DataType, Shape const &, Permutation const &);
         dim_t locate(dim_t) const noexcept;
+        TransposeInfo reform(dim_t maxblockSize) const noexcept;
+        void reformAssign(dim_t maxblockSize) noexcept;
     };
 
 }// namespace refactor::kernel
diff --git a/src/04kernel/src/attributes/transpose_info.cc b/src/04kernel/src/attributes/transpose_info.cc
index 9ae385a9..44a32b23 100644
--- a/src/04kernel/src/attributes/transpose_info.cc
+++ b/src/04kernel/src/attributes/transpose_info.cc
@@ -118,4 +118,24 @@ namespace refactor::kernel {
         return ans;
     }
 
+    TransposeInfo TransposeInfo::reform(dim_t maxblockSize) const noexcept {
+        auto ans = *this;
+        ans.reformAssign(maxblockSize);
+        return ans;
+    }
+    void TransposeInfo::reformAssign(dim_t maxblockSize) noexcept {
+        if (dims.empty()) { return; }
+        auto blockSize_ = std::gcd(blockSize, maxblockSize);
+        if (blockSize_ == blockSize) { return; }
+        auto times = blockSize / blockSize_;
+        blockCount *= times;
+        blockSize = blockSize_;
+        for (auto &s : dims) {
+            s.strideI *= times;
+            s.strideO *= times;
+        }
+        dims.resize(dims.size() + 1);
+        dims.rbegin()[0] = {1, 1};
+    }
+
 }// namespace refactor::kernel
diff --git a/src/04kernel/src/kernels/transpose/cuda_kernel.cc b/src/04kernel/src/kernels/transpose/cuda_kernel.cc
index aed8d7b6..a557dc82 100644
--- a/src/04kernel/src/kernels/transpose/cuda_kernel.cc
+++ b/src/04kernel/src/kernels/transpose/cuda_kernel.cc
@@ -5,7 +5,7 @@ namespace refactor::kernel {
     using Info = TransposeInfo;
 
     K::TransposeCuda(Info info_) noexcept
-        : Kernel(), info(std::move(info_)) {}
+        : Kernel(), info(info_.reform(16)) {}
 
     auto K::build(Info info) noexcept -> KernelBox {
 #ifndef USE_CUDA