diff --git a/mace/ops/matmul.cc b/mace/ops/matmul.cc
index 9740a257ae112abc1c9ec93b82ac035e2e61e5a8..23db4e5208c20243b404a034dc3aba1ae58d903f 100644
--- a/mace/ops/matmul.cc
+++ b/mace/ops/matmul.cc
@@ -127,14 +127,6 @@ class MatMulOp<CPU, float> : public MatMulOpBase {
 
     auto scratch_buffer = context->device()->scratch_buffer();
     scratch_buffer->Rewind();
-    index_t scratch_size = C->raw_max_size();
-    if (!A->is_weight()) {
-      scratch_size += A->raw_max_size();
-    }
-    if (!B->is_weight()) {
-      scratch_size += B->raw_max_size();
-    }
-    scratch_buffer->GrowSize(scratch_size);
 
     sgemm_.Run(a_ptr_base,
                b_ptr_base,
diff --git a/mace/ops/sgemm.cc b/mace/ops/sgemm.cc
index 717cb05676a8629ae0a8815bd24d439db1d512a7..445b9cf6211f935034b2f4e4c74532f96e8708f1 100644
--- a/mace/ops/sgemm.cc
+++ b/mace/ops/sgemm.cc
@@ -56,7 +56,7 @@ void SGemm::operator()(const MatrixMap<const float> &lhs,
                        const MatrixMap<const float> &rhs,
                        MatrixMap<float> *result,
                        ScratchBuffer *scratch_buffer) {
-  if (rhs.col() < lhs.row()) {
+  if (lhs.is_const() && !rhs.is_const()) {
     MatrixMap<const float> lhs_transpose = lhs.transpose();
     MatrixMap<const float> rhs_transpose = rhs.transpose();
     MatrixMap<float> result_transpose = result->transpose();