fix(mgb/gopt): fix cpu global layout transform fastrun error

GitOrigin-RevId: ea254297e5

dnn/include/megdnn/common.h

@@ -12,6 +12,7 @@
 #pragma once
 
 #include "megbrain_build_config.h"
+#include "megdnn/oprs/base.h"
 
 #if MGB_ENABLE_GETENV
 #define MGB_GETENV ::std::getenv
@@ -36,6 +37,11 @@ bool has_available_algo(Opr* opr, Args&&... args) {
     return !all_algos.empty();
 }
 
+template <class Opr, typename... Args>
+bool has_no_naive_heuristic_algo(Opr* opr, Args&&... args) {
+    auto&& algo = opr->get_algorithm_info_heuristic(std::forward<Args>(args)...);
+    return !static_cast<bool>(algo.attribute & detail::Algorithm::Attribute::NAIVE);
+}
 }  // namespace megdnn
 
 // vim: syntax=cpp.doxygen foldmethod=marker foldmarker=f{{{,f}}}

src/gopt/impl/global_layout_transform/opr_format_modifier.cpp

@@ -220,6 +220,28 @@ struct MultiAlgoOprTrait;
                     ::megdnn::has_available_algo(megdnn_opr, args...), array_layouts); \
             MIDOUT_E                                                                   \
         }                                                                              \
+        static bool has_no_naive_heuristic_algo(                                       \
+                const VarNodeArray& i, const cg::OperatorNodeBase* opr_) {             \
+            MIDOUT_B(                                                                  \
+                    midout_iv(MGB_HASH_STR(#_Opr)),                                    \
+                    midout_iv(MGB_HASH_STR("has_no_naive_heuristic_algo")))            \
+            auto&& opr = opr_->cast_final_safe<_Opr>();                                \
+            auto&& megdnn_opr = reinterpret_cast<MegDNNOpr*>(opr.megdnn_opr());        \
+            FixedTensorLayouts array_layouts;                                          \
+            size_t in = i.size() - 1;                                                  \
+            for (size_t idx = 0; idx < in; idx++) {                                    \
+                const auto& v = i[idx];                                                \
+                array_layouts[idx] =                                                   \
+                        TensorLayout{v->shape(), v->dtype(), v->format()};             \
+            }                                                                          \
+            const auto& v = i[in];                                                     \
+            array_layouts[arity - 1] =                                                 \
+                    TensorLayout{v->shape(), v->dtype(), v->format()};                 \
+            return APPLY(                                                              \
+                    ::megdnn::has_no_naive_heuristic_algo(megdnn_opr, args...),        \
+                    array_layouts);                                                    \
+            MIDOUT_E                                                                   \
+        }                                                                              \
     };
 INST(Convolution)
 INST(ConvBiasForward)
@@ -365,6 +387,23 @@ bool has_available_algo(const VarNodeArray& i, const cg::OperatorNodeBase* opr)
 #undef cb
 }
 
+bool has_no_naive_heuristic_algo(
+        const VarNodeArray& i, const cg::OperatorNodeBase* opr) {
+#define cb(_Opr)                                                             \
+    if (opr->dyn_typeinfo() == _Opr::typeinfo()) {                           \
+        MGB_MARK_USED_VAR(MultiAlgoOprTrait<_Opr>::has_algo);                \
+        VarNodeArray _ = i;                                                  \
+        _.emplace_back(opr->output(0));                                      \
+        return MultiAlgoOprTrait<_Opr>::has_no_naive_heuristic_algo(_, opr); \
+    } else
+    cb(Convolution) cb(ConvBiasForward) cb(ConvolutionBackwardData) cb(PoolingForward) {
+        mgb_throw(
+                InternalError, "invalid multi-algo operator(got:%s)",
+                opr->dyn_typeinfo()->name);
+    }
+#undef cb
+}
+
 bool has_opr_format(const cg::OperatorNodeBase* opr) {
     bool ret = false;
 #define cb(_Opr) ret |= opr->dyn_typeinfo() == _Opr::typeinfo();

src/gopt/impl/global_layout_transform/opr_format_modifier.h

@@ -27,6 +27,9 @@ namespace intl {
 
 bool has_available_algo(const VarNodeArray& i, const cg::OperatorNodeBase* opr);
 
+bool has_no_naive_heuristic_algo(
+        const VarNodeArray& i, const cg::OperatorNodeBase* opr);
+
 struct OprFormatInfo {
     opr::Convolution::Param::Format opr_format;
     struct TensorFormatsInfo {

src/gopt/impl/global_layout_transform/profiler_impl.cpp

@@ -331,7 +331,8 @@ float ProfilerImpl::profile_operator(
             opr::PoolingForward::typeinfo(),
     };
     if (multi_algo_oprs.count(opr->dyn_typeinfo()) &&
-        !mgb::gopt::intl::has_available_algo(new_inps, y->owner_opr()))
+        (!mgb::gopt::intl::has_available_algo(new_inps, y->owner_opr()) ||
+         !mgb::gopt::intl::has_no_naive_heuristic_algo(new_inps, y->owner_opr())))
         return PROFILE_TIME_OUT;
     if (!m_opr_filter(opr, y->owner_opr()))
         return PROFILE_TIME_OUT;