!3898 matmul bug

Merge pull request !3898 from ling/conv1x1

mindspore/lite/src/runtime/kernel/arm/base/arg_min_max_base.h

@@ -26,12 +26,10 @@ class ArgMinMaxBaseCPUKernel : public LiteKernel {
   ArgMinMaxBaseCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
                          const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx)
       : LiteKernel(parameter, inputs, outputs), context_(ctx), data_from_allocator_(false) {
-    opParameter->thread_num_ = ctx->threadNum;
+    opParameter->thread_num_ = ctx->thread_num_;
   }
 
-  virtual ~ArgMinMaxBaseCPUKernel() {
-    FreeTmpMemory();
-  }
+  virtual ~ArgMinMaxBaseCPUKernel() { FreeTmpMemory(); }
 
   int Init() override;
 
@@ -40,6 +38,7 @@ class ArgMinMaxBaseCPUKernel : public LiteKernel {
   int Run() override;
 
   void FreeTmpMemory();
+
  private:
   const lite::Context *context_;
   bool data_from_allocator_;

mindspore/lite/src/runtime/kernel/arm/base/batch_to_space_base.h

@@ -27,7 +27,7 @@ class BatchToSpaceBaseCPUKernel : public LiteKernel {
   BatchToSpaceBaseCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
                             const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx)
       : LiteKernel(parameter, inputs, outputs) {
-    opParameter->thread_num_ = ctx->threadNum;
+    opParameter->thread_num_ = ctx->thread_num_;
   }
 
   virtual ~BatchToSpaceBaseCPUKernel() = default;
@@ -38,9 +38,8 @@ class BatchToSpaceBaseCPUKernel : public LiteKernel {
 
   int Run() override { return 0; }
 
-  bool IsNoCrop() const {
-    return no_crop_;
-  }
+  bool IsNoCrop() const { return no_crop_; }
+
  private:
   bool no_crop_;
 };

mindspore/lite/src/runtime/kernel/arm/base/depth_to_space_base.h

@@ -27,7 +27,7 @@ class DepthToSpaceBaseCPUKernel : public LiteKernel {
   DepthToSpaceBaseCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
                             const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx)
       : LiteKernel(parameter, inputs, outputs) {
-    opParameter->thread_num_ = ctx->threadNum;
+    opParameter->thread_num_ = ctx->thread_num_;
   }
 
   virtual ~DepthToSpaceBaseCPUKernel() = default;
@@ -39,5 +39,4 @@ class DepthToSpaceBaseCPUKernel : public LiteKernel {
   int Run() override { return 0; }
 };
 }  // namespace mindspore::kernel
-
 #endif  // MINDSPORE_LITE_SRC_RUNTIME_KERNEL_ARM_BASE_DEPTH_TO_SPACE_BASE_H_

mindspore/lite/src/runtime/kernel/arm/base/matmul_base.h

@@ -29,7 +29,7 @@ class MatmulBaseCPUKernel : public LiteKernel {
  public:
   MatmulBaseCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
                       const std::vector<lite::tensor::Tensor *> &outputs, const Context *ctx)
-      : LiteKernel(parameter, inputs, outputs), ctx_(ctx), thread_count_(ctx->threadNum) {
+      : LiteKernel(parameter, inputs, outputs), ctx_(ctx), thread_count_(ctx->thread_num_) {
     params_ = reinterpret_cast<MatMulParameter *>(opParameter);
   }
   ~MatmulBaseCPUKernel() = default;

mindspore/lite/src/runtime/kernel/arm/opclib/fp32/matmul.cc

@@ -76,12 +76,6 @@ void MatMul8x8(const float *a, const float *b, float *c, const float *bias, ActT
 
 void MatMul(const float *a, const float *b, float *c, const float *bias, ActType act_type, int deep, int row_8_,
             int col_8_) {
-#ifdef __aarch64__
-  float minf = (act_type == ActType_No) ? FLT_MIN : 0.f;
-  float maxf = (act_type == ActType_Relu6) ? 6.0f : FLT_MAX;
-  MatMulFloatNeon64(a, b, c, bias, maxf, minf, deep, row_8_, col_8_);
-#else
   MatMul8x8(a, b, c, bias, act_type, deep, row_8_, col_8_);
-#endif
   return;
 }

mindspore/lite/test/ut/src/runtime/kernel/arm/fp32/matmul_fp32_tests.cc

@@ -67,7 +67,7 @@ TEST_F(TestMatMulFp32, simple) {
   std::vector<int> c_shape = {1, 2, 3};
   int total_size = MMTestInit(&inputs_, &outputs_, a, b, a_shape, b_shape, c_shape);
   auto ctx = new lite::Context;
-  ctx->threadNum = 2;
+  ctx->thread_num_ = 2;
   auto mm = new kernel::MatmulCPUKernel(reinterpret_cast<OpParameter *>(matmul_param), inputs_, outputs_, ctx);
   mm->Init();
   mm->Run();
@@ -98,7 +98,7 @@ TEST_F(TestMatMulFp32, simple_transb) {
   std::vector<int> c_shape = {1, 2, 3};
   int total_size = MMTestInit(&inputs_, &outputs_, a, b, a_shape, b_shape, c_shape);
   auto ctx = new lite::Context;
-  ctx->threadNum = 2;
+  ctx->thread_num_ = 2;
   auto mm = new kernel::MatmulCPUKernel(reinterpret_cast<OpParameter *>(matmul_param), inputs_, outputs_, ctx);
   mm->Init();
   mm->Run();
@@ -148,7 +148,7 @@ TEST_F(TestMatMulFp32, batch) {
   std::vector<int> c_shape = {3, 2, 3};
   int total_size = MMTestInit(&inputs_, &outputs_, a, b, a_shape, b_shape, c_shape);
   auto ctx = new lite::Context;
-  ctx->threadNum = 1;
+  ctx->thread_num_ = 1;
   auto mm = new kernel::MatmulCPUKernel(reinterpret_cast<OpParameter *>(matmul_param), inputs_, outputs_, ctx);
   mm->Init();
   mm->Run();

mindspore/lite/test/ut/src/runtime/kernel/arm/int8/matmul_int8_tests.cc

@@ -106,7 +106,7 @@ TEST_F(TestMatmulInt8, mmint8) {
   int output_zp;
   int total_size = MMInt8TestInit(&inputs_, &outputs_, matmul_param, &correct, &output_scale, &output_zp);
   auto ctx = new lite::Context;
-  ctx->threadNum = 2;
+  ctx->thread_num_ = 2;
   kernel::MatmulInt8CPUKernel *mm =
     new kernel::MatmulInt8CPUKernel(reinterpret_cast<OpParameter *>(matmul_param), inputs_, outputs_, ctx);