chore(format): fix compile bugs after code format

GitOrigin-RevId: 11a4b06f6fb0fb6353c85b9c76c0139ab30c158d

dnn/src/aarch64/matrix_mul/int8x8x16/kernel_mk4_4x4x8_a72.h

@@ -53,7 +53,9 @@ static inline void kern_4x4(const int8_t* packA, const int8_t* packB, int K,
     const int8_t* b_ptr = packB;
 
     LDC = LDC * sizeof(int8_t);
-// clang-format off
+
+    // clang-format off
+
     #define STORE_LINE(reg0)                 \
     "cmp w10, #0 \n"                         \
     "beq 101f\n"                             \

dnn/src/arm_common/elemwise/opr_impl.h

@@ -10,7 +10,6 @@
  * implied.
  */
 #pragma once
-
 #include "src/fallback/elemwise/opr_impl.h"
 
 #include "src/arm_common/elemwise_op.h"

dnn/src/arm_common/matrix_mul/int8/gemv.cpp

@@ -10,6 +10,7 @@
  */
 
 #include "src/arm_common/simd_macro/marm_neon.h"
+
 #include "src/arm_common/matrix_mul/int8/gemv.h"
 #include "src/common/utils.h"
 #include "megdnn/oprs.h"

dnn/src/common/cv/interp_helper.cpp

@@ -60,11 +60,8 @@
 #pragma GCC diagnostic ignored "-Wnon-virtual-dtor"
 // TableHolderBase has no problem; ignore the warning for old clang versions
 
-#include "./helper.h"
 #include "./interp_helper.h"
 
-#include "src/common/utils.h"
-
 using namespace megdnn;
 using namespace megdnn::megcv;
 

dnn/src/common/cv/interp_helper.h

@@ -62,7 +62,9 @@
 #pragma once
 
 #include "src/common/cv/aligned_allocator.h"
+#include "src/common/utils.h"
 
+#include "./helper.h"
 #include "megdnn/opr_param_defs.h"
 
 #include <cstdint>

dnn/src/cuda/batch_conv_bias/helper.cuh

@@ -10,6 +10,7 @@
  */
 #pragma once
 #include "src/cuda/convolution_helper/parameter.cuh"
+#include "src/cuda/utils.cuh"
 
 namespace megdnn {
 namespace cuda {

dnn/src/cuda/concat/concat.cuh

@@ -10,6 +10,7 @@
  */
 #pragma once
 #include <stdint.h>
+#include "src/cuda/utils.cuh"
 
 namespace megdnn {
 namespace cuda {

dnn/src/cuda/dot/dot.cuh

@@ -10,6 +10,7 @@
  */
 #pragma once
 #include "megdnn/dtype.h"
+#include "src/cuda/utils.cuh"
 
 namespace megdnn {
 namespace cuda {

dnn/src/cuda/repeat/repeat.cuh

@@ -9,6 +9,7 @@
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
 #pragma once
+#include "src/cuda/utils.cuh"
 
 namespace megdnn {
 namespace cuda {

dnn/src/cuda/tile/tile.cuh

@@ -9,6 +9,7 @@
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
 #pragma once
+#include "src/cuda/utils.cuh"
 
 namespace megdnn {
 namespace cuda {

dnn/src/cuda/warp_affine/common.cuh

@@ -9,6 +9,7 @@
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
 #pragma once
+#include "src/cuda/utils.cuh"
 
 namespace megdnn {
 namespace cuda {

dnn/src/cuda/warp_perspective/common.cuh

@@ -9,6 +9,7 @@
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
 #pragma once
+#include "src/cuda/utils.cuh"
 
 namespace megdnn {
 namespace cuda {

dnn/src/rocm/handle.h

@@ -9,6 +9,8 @@
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
 #pragma once
+#include "src/rocm/miopen_wrapper.h"
+
 #include "megcore_rocm.h"
 #include "megdnn/basic_types.h"
 #include "megdnn/handle.h"
@@ -16,7 +18,6 @@
 
 #include "src/common/handle_impl.h"
 #include "src/common/utils.h"
-#include "src/rocm/miopen_with_check.h"
 
 #include <rocblas.h>
 #include <atomic>

dnn/src/x86/avx_helper.h

@@ -13,9 +13,11 @@
 #include "megdnn/arch.h"
 
 #include <immintrin.h>
+#ifdef WIN32
 #include <avxintrin.h>
 #include <avx2intrin.h>
 #include <fmaintrin.h>
+#endif
 
 #if !defined (__clang__)
 #pragma GCC target ("avx")

dnn/src/x86/conv_bias/f32/do_conv_stride2.h

@@ -9,7 +9,8 @@
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
 #pragma once
-
+// clang-format off
 #include "src/x86/simd_macro/sse_helper.h"
 #include "src/fallback/convolution/do_conv_stride2_decl.inl"
 #include "src/x86/simd_macro/sse_helper_epilogue.h"
+// clang-format on

dnn/src/x86/convolution/avx/convolution_conv_fh1_avx.cpp

@@ -801,8 +801,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_conv_fh2_avx.cpp

@@ -896,8 +896,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_conv_fh3_avx.cpp

@@ -943,8 +943,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_conv_fh4_avx.cpp

@@ -948,8 +948,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_conv_fh5_avx.cpp

@@ -917,8 +917,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_conv_fh6_avx.cpp

@@ -856,8 +856,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_conv_fh7_avx.cpp

@@ -771,8 +771,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_xcorr_fh1_avx.cpp

@@ -788,8 +788,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_xcorr_fh2_avx.cpp

@@ -872,8 +872,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_xcorr_fh3_avx.cpp

@@ -910,8 +910,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_xcorr_fh4_avx.cpp

@@ -908,8 +908,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_xcorr_fh5_avx.cpp

@@ -872,8 +872,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_xcorr_fh6_avx.cpp

@@ -808,8 +808,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/avx/convolution_xcorr_fh7_avx.cpp

@@ -722,8 +722,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_conv_fh1_fma.cpp

@@ -785,9 +785,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_conv_fh2_fma.cpp

@@ -827,9 +827,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_conv_fh3_fma.cpp

@@ -842,9 +842,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_conv_fh4_fma.cpp

@@ -833,9 +833,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_conv_fh5_fma.cpp

@@ -803,9 +803,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_conv_fh6_fma.cpp

@@ -755,9 +755,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_conv_fh7_fma.cpp

@@ -692,9 +692,7 @@
         }                                                     \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_xcorr_fh1_fma.cpp

@@ -771,9 +771,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_xcorr_fh2_fma.cpp

@@ -801,9 +801,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_xcorr_fh3_fma.cpp

@@ -806,9 +806,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_xcorr_fh4_fma.cpp

@@ -789,9 +789,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_xcorr_fh5_fma.cpp

@@ -753,9 +753,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_xcorr_fh6_fma.cpp

@@ -701,9 +701,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/convolution/fma/convolution_xcorr_fh7_fma.cpp

@@ -636,9 +636,7 @@
         }                                                                  \
     } while (0)
 
-#include <immintrin.h>
-#include <avxintrin.h>
-#include <fmaintrin.h>
+#include "src/x86/avx_helper.h"
 #include <algorithm>
 
 #include "../convolution_direct_special_cases.h"

dnn/src/x86/local/local_avx.cpp

@@ -8,6 +8,8 @@
  * software distributed under the License is distributed on an
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
+// clang-format off
 #include "src/x86/simd_helper.h"
 #include "src/x86/simd_macro/avx_helper.h"
 #include "src/common/local/local_def.inl"
+// clang-format on

dnn/src/x86/local/local_fma.cpp

@@ -8,6 +8,8 @@
  * software distributed under the License is distributed on an
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
+// clang-format off
 #include "src/x86/simd_helper.h"
 #include "src/x86/simd_macro/fma_helper.h"
 #include "src/common/local/local_def.inl"
+// clang-format on

dnn/src/x86/local/local_simd.h

@@ -10,6 +10,7 @@
  */
 #pragma once
 
+// clang-format off
 #include "src/x86/simd_macro/sse_helper.h"
 #include "src/common/local/local_decl.inl"
 #include "src/x86/simd_macro/sse_helper_epilogue.h"
@@ -21,3 +22,4 @@
 #include "src/x86/simd_macro/fma_helper.h"
 #include "src/common/local/local_decl.inl"
 #include "src/x86/simd_macro/fma_helper_epilogue.h"
+// clang-format on

dnn/src/x86/local/local_sse.cpp

@@ -8,6 +8,8 @@
  * software distributed under the License is distributed on an
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
+// clang-format off
 #include "src/x86/simd_helper.h"
 #include "src/x86/simd_macro/sse_helper.h"
 #include "src/common/local/local_def.inl"
+// clang-form on

dnn/src/x86/matrix_mul/common/common.h

@@ -11,7 +11,6 @@
  */
 #pragma once
 #include <x86intrin.h>
-
 #ifdef WIN32
 #include <avx2intrin.h>
 #include <avxintrin.h>

dnn/src/x86/simd_helper.h

@@ -13,9 +13,11 @@
 #include "megdnn/arch.h"
 
 #include <immintrin.h>
+#ifdef WIN32
 #include <xmmintrin.h>
 #include <avxintrin.h>
 #include <fmaintrin.h>
+#endif
 #include <cmath>
 #include <algorithm>
 

imperative/tablegen/emitter.h

@@ -17,6 +17,7 @@
 #include "llvm/Support/raw_ostream.h"
 
 namespace mlir::tblgen {
+using llvm::raw_ostream;
 
 struct Environment {
     std::unordered_map<unsigned int, std::pair<llvm::StringRef, llvm::StringRef>> enumAlias;
@@ -37,4 +38,4 @@ protected:
     Environment* env_p = nullptr;
 };
 
-} // namespace mlir::tblgen
+} // namespace mlir::tblgen
\ No newline at end of file

imperative/tablegen/targets/macros.cpp

@@ -9,6 +9,7 @@
  * "AS IS" BASIS, WITHOUT ARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  */
 
+#include "./macros.h"
 #include "./cpp_class.h"
 #include "../emitter.h"
 

src/core/impl/graph/var_node_mem_mgr.cpp

@@ -125,7 +125,7 @@ StaticDeviceMemoryManager::make_default_impl() {
 #endif  // MGB_THREAD_SAFE
 
 /* ==================== AsyncVarReleaser ==================== */
-#if MGB_CUDA || MGB_ATLAS || MGB_CAMBRICON  || MGB_ROCM
+#if MGB_COMMON_ASYNC_COMPNODE
 class VarNodeMemManager::AsyncVarReleaser {
     struct WaiterParam {
         CompNode cn;
@@ -248,7 +248,7 @@ bool VarNodeMemManager::ImpureMemPlanManager::check_need_realloc() {
 VarNodeMemManager::VarNodeMemManager(ComputingGraphImpl* graph)
         : m_owner_graph(graph),
           m_seq_mem_opt(graph)
-#if MGB_CUDA || MGB_ATLAS || MGB_CAMBRICON  || MGB_ROCM
+#if MGB_COMMON_ASYNC_COMPNODE
           ,m_asyn_var_releaser(new AsyncVarReleaser)
 #endif
 {
@@ -256,7 +256,7 @@ VarNodeMemManager::VarNodeMemManager(ComputingGraphImpl* graph)
         MGB_MARK_USED_VAR(ev);
         // async release is only used for sync between multiple comp nodes, and
         // does not wait for device to finish
-#if MGB_CUDA || MGB_ATLAS || MGB_CAMBRICON  || MGB_ROCM
+#if MGB_COMMON_ASYNC_COMPNODE
         m_asyn_var_releaser->wait_release_finish();
 #endif
         m_cpu_async_release_barrier.wait_zero();
@@ -297,8 +297,7 @@ VarNodeMemManager::VarNodeMemManager(ComputingGraphImpl* graph)
     graph->event().register_receiver_permanent<event::CompSeqExecError>(
             on_comp_seq_error);
 
-#if MGB_ENABLE_VAR_DEV_MEM_DEFRAGMENTER &&                                   \
-        (MGB_CUDA || MGB_ATLAS || MGB_CAMBRICON  || MGB_ROCM)
+#if MGB_ENABLE_VAR_DEV_MEM_DEFRAGMENTER && MGB_COMMON_ASYNC_COMPNODE
     auto on_mem_defrag_start = [this](const event::BeforeMemDefrag&) {
         m_asyn_var_releaser->wait_release_finish();
     };

src/core/impl/graph/var_node_mem_mgr.h

@@ -445,7 +445,12 @@ class VarNodeMemManager {
 
         SyncableCounter m_cpu_async_release_barrier;
 
-#if MGB_CUDA || MGB_ATLAS || MGB_CAMBRICON  || MGB_ROCM
+// clang-format off
+#define MGB_COMMON_ASYNC_COMPNODE \
+    (MGB_CUDA || MGB_ATLAS || MGB_CAMBRICON  || MGB_ROCM)
+    // clang-format on
+
+#if MGB_COMMON_ASYNC_COMPNODE
         //! release dynamic var on after compnode event finishes
         class AsyncVarReleaser;
         std::unique_ptr<AsyncVarReleaser> m_asyn_var_releaser;

src/core/include/megbrain/utils/thread_impl_spinlock.h

@@ -14,6 +14,7 @@
 #include "megbrain/common.h"
 #include <thread>
 #include <atomic>
+#include "megbrain/utils/metahelper.h"
 
 namespace mgb {
 
@@ -24,7 +25,7 @@ class Spinlock final: public NonCopyableObj {
     public:
 
         void lock() {
-            while (m_state.test_and_set(std::memory_order_acquire));
+            while (m_state.test_and_set(std::memory_order_acquire)) {};
         }
 
         void unlock() {

src/opr/include/megbrain/opr/basic_arith.h

@@ -281,8 +281,8 @@ MGB_DEFINE_OPR_CLASS(AddUpdate,
  * Mode specifies the actual arithmetic; and exactly one of *axis* and
  * *target_shape* must be provided, to specify output shape.
  */
-MGB_DEFINE_OPR_CLASS(Reduce, intl::DynamicOutputIfInputDynamic<
-        intl::OutshapeBySymvarSCNOpr<mixin::MegDNNOprHolder>>) //  {
+MGB_DEFINE_OPR_CLASS(Reduce,
+        intl::DynamicOutputIfInputDynamic<intl::OutshapeBySymvarSCNOpr<mixin::MegDNNOprHolder>>) // {
 
     public:
         using Param = megdnn::param::Reduce;
@@ -350,16 +350,17 @@ MGB_DEFINE_OPR_CLASS(Reduce, intl::DynamicOutputIfInputDynamic<
  * the optimizer.
  */
 MGB_DEFINE_OPR_CLASS(PowC, intl::MegDNNOprWrapperFwd<megdnn::PowC>) // {
+public:
+    PowC(VarNode* inp, const Param& param, const OperatorNodeConfig& config);
+    static SymbolVar make(SymbolVar inp, const Param& param = {},
+                          const OperatorNodeConfig& config = {});
+
+private:
     void add_input_layout_constraint() override;
     void init_output_static_infer_desc() override;
     void mem_plan_fwd_in2out_writable() override;
     NodeProp* do_make_node_prop() const override;
     void scn_do_execute() override;
-
-public:
-    PowC(VarNode* inp, const Param& param, const OperatorNodeConfig& config);
-    static SymbolVar make(SymbolVar inp, const Param& param = {},
-                          const OperatorNodeConfig& config = {});
 };
 
 } // namespace opr

src/opr/test/atlas_models.h

-//generated by tools/atlas/embed.py
+// generated by tools/atlas/embed.py
+// clang-format off
 #pragma once
 #include <map>
 #include <string>