fix(mgb): fix execution_policy set of matmul

GitOrigin-RevId: 90f539b0bed2e1bb5da0103eeb2b46e4e9690c6f

src/opr/impl/blas.cpp

@@ -98,6 +98,7 @@ size_t MatrixMul::get_workspace_size_bytes(
         param ^= 1;
     };
     MGB_TRY {
+        megdnn_opr()->execution_policy() = {};
         a = AlgoChooser<megdnn::MatrixMul>::setup_algo({i0, i1, out},
                                                        megdnn_opr(), this);
         //! Here we just want to save the execution policy got from setup_algo,
@@ -106,24 +107,28 @@ size_t MatrixMul::get_workspace_size_bytes(
         const_cast<MatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
         transpose(i0, tparam.transposeA);
         b = AlgoChooser<megdnn::MatrixMul>::setup_algo({i0, i1, out},
                                                        megdnn_opr(), this);
         const_cast<MatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
         transpose(i1, tparam.transposeB);
         c = AlgoChooser<megdnn::MatrixMul>::setup_algo({i0, i1, out},
                                                        megdnn_opr(), this);
         const_cast<MatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
         transpose(i0, tparam.transposeA);
         d = AlgoChooser<megdnn::MatrixMul>::setup_algo({i0, i1, out},
                                                        megdnn_opr(), this);
         const_cast<MatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
     }
     MGB_FINALLY({ tparam = this->param(); });
     return std::max(std::max(a, b), std::max(c, d));
@@ -252,29 +257,34 @@ size_t BatchedMatrixMul::get_workspace_size_bytes(
         param ^= 1;
     };
     MGB_TRY {
+        megdnn_opr()->execution_policy() = {};
         a = AlgoChooser<megdnn::BatchedMatrixMul>::setup_algo(
                 {i0, i1, out}, megdnn_opr(), this);
         const_cast<BatchedMatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
         transpose(i0, tparam.transposeA);
         b = AlgoChooser<megdnn::BatchedMatrixMul>::setup_algo(
                 {i0, i1, out}, megdnn_opr(), this);
         const_cast<BatchedMatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
         transpose(i1, tparam.transposeB);
         c = AlgoChooser<megdnn::BatchedMatrixMul>::setup_algo(
                 {i0, i1, out}, megdnn_opr(), this);
         const_cast<BatchedMatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
         transpose(i0, tparam.transposeA);
         d = AlgoChooser<megdnn::BatchedMatrixMul>::setup_algo(
                 {i0, i1, out}, megdnn_opr(), this);
         const_cast<BatchedMatrixMul*>(this)
                 ->m_cadidate_execution_policies[get_mask_from_matmul(tparam)] =
                 megdnn_opr()->execution_policy();
+        megdnn_opr()->execution_policy() = {};
     }
     MGB_FINALLY({ tparam = this->param(); });
     return std::max(std::max(a, b), std::max(c, d));

src/opr/impl/search_policy/algo_chooser.cpp

@@ -266,7 +266,6 @@ AlgoChooser<Opr>::ExeContext::choose_by_heuristic(bool reproducible) const {
     auto workspace_limit = WorkspaceLimitGetter::get_workspace_limit(
             opr->owner_graph(), opr->comp_node(),
             opr->execution_policy().workspace_limit);
-    m_megdnn_opr->execution_policy() = {};
     return APPLY(m_megdnn_opr->get_algorithm_info_heuristic(
                          args..., workspace_limit, reproducible),
                  m_layouts);