mindspore-Ecosystem
/
mindspore
21
0
Fork 27
Code Issues Pull Requests 6 Projects Releases Wiki Activity

!7384 opencl convolutoin kernel run segmentfault with hasbias=false 

Merge pull request !7384 from 王东旭/fix_conv_no_bias_bug
This commit is contained in:
mindspore-ci-bot 2020-10-16 13:51:09 +08:00 committed by Gitee
parent 554123ed49 2abd080334
commit fa37f8fde3
6 changed files with 144 additions and 177 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
mindspore/lite/src/runtime/kernel/opencl/kernel/concat.cc
View File

@ -76,7 +76,9 @@ int ConcatOpenCLKernel::Init() {
auto param = reinterpret_cast<ConcatParameter *>(this->op_parameter_);
MS_LOG(DEBUG) << " concat at axis=: " << param->axis_;
param->axis_ = (param->axis_ == -1) ? (in_tensors_[0]->shape().size() - 1) : param->axis_;
if (param->axis_ < 0) {
param->axis_ += in_tensors_.front()->shape().size();
}
if (param->axis_ < 0 || param->axis_ > 3) {
MS_LOG(ERROR) << " only support axis >= 0 and axis <= 3 ";
return RET_ERROR;

68
mindspore/lite/src/runtime/kernel/opencl/kernel/convolution.cc
View File

@ -65,6 +65,7 @@ int ConvolutionOpenCLKernel::Init() {
CO_SLICES_ = UP_DIV(CO_, C4NUM);
KH_ = param->kernel_h_;
KW_ = param->kernel_w_;
has_bias_ = in_tensors_.size() == 3;
// note: TILES_X TILES_Y TILES_XY is only used when use_winograd_=true
TILES_X_ = UP_DIV(OW_, 4);
@ -243,7 +244,9 @@ int ConvolutionOpenCLKernel::InitBias() {
int ConvolutionOpenCLKernel::InitBuffer() {
InitWeight();
InitBias();
if (has_bias_) {
InitBias();
}
return RET_OK;
}
@ -298,7 +301,9 @@ int ConvolutionOpenCLKernel::Run() {
cl_int4 _36to4x4_out_shape = {1, OH_, OW_, CO_SLICES_};
ocl_runtime_->SetKernelArg(kernel_36to4x4_, arg_cn++, winograd_mem1_, lite::opencl::MemType::IMG);
ocl_runtime_->SetKernelArg(kernel_36to4x4_, arg_cn++, out_tensors_[0]->data_c(), lite::opencl::MemType::IMG);
ocl_runtime_->SetKernelArg(kernel_36to4x4_, arg_cn++, packed_bias_, lite::opencl::MemType::BUF);
if (has_bias_) {
ocl_runtime_->SetKernelArg(kernel_36to4x4_, arg_cn++, packed_bias_, lite::opencl::MemType::BUF);
}
ocl_runtime_->SetKernelArg(kernel_36to4x4_, arg_cn++, _36to4x4_in_shape);
ocl_runtime_->SetKernelArg(kernel_36to4x4_, arg_cn++, _36to4x4_out_shape);
} else {
@ -306,7 +311,9 @@ int ConvolutionOpenCLKernel::Run() {
ocl_runtime_->SetKernelArg(kernel_conv_, arg_cn++, in_tensors_[0]->data_c(), lite::opencl::MemType::IMG);
ocl_runtime_->SetKernelArg(kernel_conv_, arg_cn++, out_tensors_[0]->data_c(), lite::opencl::MemType::IMG);
ocl_runtime_->SetKernelArg(kernel_conv_, arg_cn++, packed_weight_, lite::opencl::MemType::BUF);
ocl_runtime_->SetKernelArg(kernel_conv_, arg_cn++, packed_bias_, lite::opencl::MemType::BUF);
if (has_bias_) {
ocl_runtime_->SetKernelArg(kernel_conv_, arg_cn++, packed_bias_, lite::opencl::MemType::BUF);
}
if (op_format_ == Format_NC4HW4) {
cl_int4 input_shape = {1, IH_, IW_, CI_SLICES_};
cl_int4 output_shape = {1, OH_, OW_, CO_SLICES_};
@ -372,10 +379,14 @@ std::string ConvolutionOpenCLKernel::CodeGenConvolutionNHWC4() {
code +=
"__kernel void Convolution(__read_only image2d_t input,\n"
" __write_only image2d_t output,\n"
" __global FLT4 *weight,\n"
" __global FLT4 *bias)"
"{\n";
" __write_only image2d_t output,\n";
if (has_bias_) {
code +=
" __global FLT4 *weight,\n"
" __global FLT4 *bias) {\n";
} else {
code += " __global FLT4 *weight) {\n";
}
code += " int n_oh = get_global_id(0); // [0, N*OH)\n";
if (batch_size_ == 1) {
@ -426,17 +437,20 @@ std::string ConvolutionOpenCLKernel::CodeGenConvolutionNHWC4() {
" }\n"
" }\n"
" }\n\n";
code += " FLT4 out0_c4_bias = out0_c4 + bias[co_slice];\n";
if (has_bias_) {
code += " out0_c4 = out0_c4 + bias[co_slice];\n";
}
if (param->act_type_ == ActType_Relu) {
code += " out0_c4_bias = max(out0_c4_bias, (FLT4)(0.0f));\n";
code += " out0_c4 = max(out0_c4, (FLT4)(0.0f));\n";
} else if (param->act_type_ == ActType_Relu6) {
code += " out0_c4_bias = clamp(out0_c4_bias, (FLT4)(0.0f), (FLT4)(6.0f));\n";
code += " out0_c4 = clamp(out0_c4, (FLT4)(0.0f), (FLT4)(6.0f));\n";
}
if (OW_ * CO_SLICES_ <= MAX_IMAGE2D_SIZE) {
code += " WRITE_IMAGE(output, (int2)(ow * CO_SLICES + co_slice, n_oh), out0_c4_bias);// NHWC4: NH WC\n}";
code += " WRITE_IMAGE(output, (int2)(ow * CO_SLICES + co_slice, n_oh), out0_c4);// NHWC4: NH WC\n}";
} else {
code += " WRITE_IMAGE(output, (int2)(n_oh * CO_SLICES + co_slice, ow), out0_c4_bias);\n}";
code += " WRITE_IMAGE(output, (int2)(n_oh * CO_SLICES + co_slice, ow), out0_c4);\n}";
}
return code;
}
@ -460,8 +474,11 @@ std::string ConvolutionOpenCLKernel::CodeGenConvolutionNC4HW4() {
"\n"
"__kernel void Convolution(__read_only image2d_t input,\n"
" __write_only image2d_t output,\n"
" __global FLT4 *weight,\n"
" __global FLT4 *bias,\n"
" __global FLT4 *weight,\n";
if (has_bias_) {
code += " __global FLT4 *bias,\n";
}
code +=
" const int4 input_shape,\n"
" const int4 output_shape)\n"
"{\n";
@ -578,7 +595,10 @@ std::string ConvolutionOpenCLKernel::CodeGenConvolutionNC4HW4() {
" }\n"
" }\n\n";
code += " out0 = out0 + bias[co_slice];\n";
if (has_bias_) {
code += " out0 = out0 + bias[co_slice];\n";
}
if (param->act_type_ == ActType_Relu) {
code += " out0 = max(out0, (FLT4)(0.0f));\n";
} else if (param->act_type_ == ActType_Relu6) {
@ -591,7 +611,9 @@ std::string ConvolutionOpenCLKernel::CodeGenConvolutionNC4HW4() {
" if (last_is_double)"
" {\n";
}
code += " out1 = out1 + bias[co_slice];\n";
if (has_bias_) {
code += " out1 = out1 + bias[co_slice];\n";
}
if (param->act_type_ == ActType_Relu) {
code += " out1 = max(out1, (FLT4)(0.0f));\n";
} else if (param->act_type_ == ActType_Relu6) {
@ -788,8 +810,11 @@ std::string ConvolutionOpenCLKernel::CodeGenWinograd36To4x4() {
"};\n"
"\n"
"__kernel void Winograd36To4x4(__read_only image2d_t input,\n"
" __write_only image2d_t output,\n"
" __global FLT4 *bias,\n"
" __write_only image2d_t output,\n";
if (has_bias_) {
code += " __global FLT4 *bias,\n";
}
code +=
" int4 input_shape, // N 36 H/4*W/4 CO_SLICES\n"
" int4 output_shape) // N H W CO_SLICES\n"
"{\n"
@ -824,9 +849,10 @@ std::string ConvolutionOpenCLKernel::CodeGenWinograd36To4x4() {
" for (int y = 0; y < 6; y++)\n"
" {\n"
" acc += AtM_row[y] * At[x * 6 + y];\n"
" }\n"
" acc += bias[slice];\n"
"\n";
" }\n";
if (has_bias_) {
code += " acc += bias[slice];\n";
}
auto param = reinterpret_cast<ConvParameter *>(op_parameter_);
if (param->act_type_ == ActType_Relu) {

60
mindspore/lite/src/runtime/kernel/opencl/kernel/convolution.h
View File

@ -35,44 +35,15 @@ class ConvolutionOpenCLKernel : public OpenCLKernel {
int Init() override;
int Run() override;
int InitBuffer();
int GetImageSize(size_t idx, std::vector<size_t> *img_size) override;
private:
bool use_fp16_ = false;
int batch_size_{};
int CI_{};
int IH_{};
int IW_{};
int CO_{};
int OH_{};
int OW_{};
int CI_SLICES_{};
int CO_SLICES_{};
int KH_{};
int KW_{};
void *packed_weight_ = nullptr;
void *packed_bias_ = nullptr;
bool use_winograd_ = false;
int TILES_X_{};
int TILES_Y_{};
int TILES_XY_{};
void *winograd_mem0_ = nullptr;
void *winograd_mem1_ = nullptr;
cl::Kernel kernel_4x4to36_;
cl::Kernel kernel_conv_;
cl::Kernel kernel_36to4x4_;
int InitBuffer();
int InitWeight();
int InitBias();
int GenerateWinogradWeight();
std::string CodeGenConvolutionNHWC4();
std::string CodeGenConvolutionNC4HW4();
std::string CodeGenWinograd4x4To36();
std::string CodeGenWinogradConvolution();
std::string CodeGenWinograd36To4x4();
@ -110,6 +81,7 @@ class ConvolutionOpenCLKernel : public OpenCLKernel {
param->pad_r_,
param->dilation_h_,
param->dilation_w_,
has_bias_,
use_fp16_,
op_format_,
param->act_type_};
@ -119,6 +91,34 @@ class ConvolutionOpenCLKernel : public OpenCLKernel {
}
return code_id;
}
bool use_fp16_ = false;
int batch_size_{};
int CI_{};
int IH_{};
int IW_{};
int CO_{};
int OH_{};
int OW_{};
int CI_SLICES_{};
int CO_SLICES_{};
int KH_{};
int KW_{};
void *packed_weight_ = nullptr;
void *packed_bias_ = nullptr;
bool has_bias_ = false;
bool use_winograd_ = false;
int TILES_X_{};
int TILES_Y_{};
int TILES_XY_{};
void *winograd_mem0_ = nullptr;
void *winograd_mem1_ = nullptr;
cl::Kernel kernel_4x4to36_;
cl::Kernel kernel_conv_;
cl::Kernel kernel_36to4x4_;
};
} // namespace mindspore::kernel

143
mindspore/lite/src/runtime/kernel/opencl/kernel/to_format.cc
View File

@ -34,134 +34,71 @@ namespace mindspore::kernel {
int ToFormatOpenCLKernel::Init() {
auto parameter = reinterpret_cast<OpenCLToFormatParameter *>(op_parameter_);
out_mem_type_ = parameter->out_mem_type;
std::string program_name = "to_format";
std::map<schema::Format, std::string> format_str{
{schema::Format::Format_NCHW, "NCHW"}, {schema::Format::Format_NHWC, "NHWC"},
{schema::Format::Format_NC4HW4, "NC4HW4"}, {schema::Format::Format_NC4, "NHWC4"},
{schema::Format::Format_NC, "NHWC"}, {schema::Format::Format_NHWC4, "NHWC4"}};
std::string kernel_name =
"to_format_" + format_str[in_tensors_[0]->GetFormat()] + "_to_" + format_str[out_tensors_[0]->GetFormat()];
std::map<TypeId, std::string> dtype_str{
{kNumberTypeFloat32, "float"}, {kNumberTypeFloat16, "half"}, {kNumberTypeInt8, "Int8"}};
if (out_mem_type_ == OpenCLMemType::IMG) {
kernel_name += "_IMG_" + dtype_str[in_tensors_[0]->data_type()];
std::map<TypeId, std::string> dtype_str{{kNumberTypeFloat32, "float"}, {kNumberTypeFloat16, "half"}};
std::string kernel_name;
if (parameter->out_mem_type == OpenCLMemType::IMG) {
kernel_name = "to_format_NHWC_to_NHWC4_IMG_" + dtype_str[in_tensors_[0]->data_type()];
} else {
kernel_name += "_BUF_" + dtype_str[out_tensors_[0]->data_type()];
kernel_name = "to_format_NHWC4_to_NHWC_BUF_" + dtype_str[out_tensors_[0]->data_type()];
}
this->set_name(kernel_name);
#ifdef PROGRAM_WITH_IL
kernel_ = ocl_runtime_->GetKernelFromBinary(kernel_name);
#else
std::string program_name = "to_format";
std::set<std::string> build_options;
std::string source = to_format_source;
ocl_runtime_->LoadSource(program_name, source);
ocl_runtime_->BuildKernel(kernel_, program_name, kernel_name, build_options);
#endif
InitNHWCShape();
InitNHWC();
MS_LOG(DEBUG) << kernel_name << " Init Done!";
return RET_OK;
}
int ToFormatOpenCLKernel::InitNHWCShape() {
std::vector<int> shapex = out_tensors_[0]->shape();
size_t n, h, w, c;
if (shapex.size() == 2) {
n = shapex[0];
h = 1;
w = 1;
c = shapex[1];
nhwc_shape_ = {n, h, w, c};
return RET_OK;
int ToFormatOpenCLKernel::InitNHWC() {
std::vector<int> out_shape = out_tensors_[0]->shape();
if (out_shape.size() == 1) {
N_ = out_shape[0];
H_ = 1;
W_ = 1;
C_ = 1;
} else if (out_shape.size() == 2) {
N_ = out_shape[0];
H_ = 1;
W_ = 1;
C_ = out_shape[1];
} else if (out_shape.size() == 3) {
N_ = out_shape[0];
H_ = 1;
W_ = out_shape[1];
C_ = out_shape[2];
} else if (out_shape.size() == 4) {
N_ = out_shape[0];
H_ = out_shape[1];
W_ = out_shape[2];
C_ = out_shape[3];
}
if (shapex.size() == 3) {
n = 1;
h = 1;
w = 1;
c = 1;
nhwc_shape_ = {n, h, w, c};
return RET_OK;
}
if (out_tensors_[0]->GetFormat() == schema::Format::Format_NC4HW4 ||
out_tensors_[0]->GetFormat() == schema::Format::Format_NHWC4 ||
out_tensors_[0]->GetFormat() == schema::Format::Format_NHWC) {
n = shapex[0];
h = shapex[1];
w = shapex[2];
c = shapex[3];
} else if (out_tensors_[0]->GetFormat() == schema::Format::Format_NCHW) {
n = shapex[0];
h = shapex[2];
w = shapex[3];
c = shapex[1];
} else if (out_tensors_[0]->GetFormat() == schema::Format::Format_NC4 ||
out_tensors_[0]->GetFormat() == schema::Format::Format_NC) {
n = shapex[0];
h = 1;
w = 1;
c = shapex[1];
} else {
n = shapex[0];
h = shapex[1];
w = shapex[2];
c = shapex[3];
}
nhwc_shape_ = {n, h, w, c};
return RET_OK;
}
int ToFormatOpenCLKernel::ReSize() { return RET_OK; }
int ToFormatOpenCLKernel::GetGlobalSize(size_t idx, std::vector<size_t> *global_size) {
std::vector<size_t> vec = {nhwc_shape_[0] * nhwc_shape_[1], nhwc_shape_[2], UP_DIV(nhwc_shape_[3], C4NUM)};
*global_size = std::move(vec);
return RET_OK;
}
int ToFormatOpenCLKernel::GetLocalSize(size_t idx, const std::vector<size_t> &global_size,
std::vector<size_t> *local_size) {
return RET_OK;
}
int ToFormatOpenCLKernel::GetImageSize(size_t idx, std::vector<size_t> *img_size) {
size_t im_dst_x, im_dst_y;
if (out_tensors_[0]->GetFormat() == schema::Format::Format_NC4HW4) {
int c = nhwc_shape_[3];
int h = nhwc_shape_[1];
int w = nhwc_shape_[2];
im_dst_y = nhwc_shape_[0] * h * UP_DIV(c, C4NUM);
im_dst_x = w;
} else if (out_tensors_[0]->GetFormat() == schema::Format::Format_NHWC4) {
int h = nhwc_shape_[0] * nhwc_shape_[1];
int w = nhwc_shape_[2];
int c = nhwc_shape_[3];
im_dst_x = w * UP_DIV(c, C4NUM);
im_dst_y = h;
} else if (out_tensors_[0]->GetFormat() == schema::Format::Format_NC4) {
int c = nhwc_shape_[3];
im_dst_x = UP_DIV(c, C4NUM);
im_dst_y = 1;
} else {
MS_LOG(ERROR) << "Unsupported format. " << out_tensors_[0]->GetFormat();
return RET_ERROR;
}
img_size->clear();
auto enable_fp16_ = ocl_runtime_->GetFp16Enable();
size_t img_dtype = CL_FLOAT;
if (enable_fp16_) {
img_dtype = CL_HALF_FLOAT;
}
std::vector<size_t> vec{im_dst_x, im_dst_y, img_dtype};
*img_size = vec;
size_t img_height = N_ * H_;
size_t img_width = W_ * UP_DIV(C_, C4NUM);
size_t img_dtype = ocl_runtime_->GetFp16Enable() ? CL_HALF_FLOAT : CL_FLOAT;
*img_size = {img_width, img_height, img_dtype};
return RET_OK;
}
int ToFormatOpenCLKernel::Run() {
MS_LOG(DEBUG) << this->name() << " Running!";
std::vector<size_t> local = {};
std::vector<size_t> global;
GetGlobalSize(0, &global);
cl_int4 shape{(cl_int)nhwc_shape_[0], (cl_int)nhwc_shape_[1], (cl_int)nhwc_shape_[2], (cl_int)nhwc_shape_[3]};
std::vector<size_t> global = {N_ * H_, W_, UP_DIV(C_, C4NUM)};
std::vector<size_t> local = {16, 8, 1};
cl_int4 shape{(cl_int)N_, (cl_int)H_, (cl_int)W_, (cl_int)C_};
cl_int4 gsize{(cl_int)global[0], (cl_int)global[1], (cl_int)global[2], 1};
auto src_mem_type = (out_mem_type_ == OpenCLMemType::IMG) ? lite::opencl::MemType::BUF : lite::opencl::MemType::IMG;
auto dst_mem_type = (out_mem_type_ == OpenCLMemType::IMG) ? lite::opencl::MemType::IMG : lite::opencl::MemType::BUF;
ocl_runtime_->SetKernelArg(kernel_, 0, in_tensors_[0]->data_c(), src_mem_type);

12
mindspore/lite/src/runtime/kernel/opencl/kernel/to_format.h
View File

@ -31,16 +31,18 @@ class ToFormatOpenCLKernel : public OpenCLKernel {
~ToFormatOpenCLKernel() override{};
int Init() override;
int ReSize() override;
int ReSize() override { return RET_OK; };
int Run() override;
int GetImageSize(size_t idx, std::vector<size_t> *img_size) override;
int GetGlobalSize(size_t idx, std::vector<size_t> *global_size) override;
int GetLocalSize(size_t idx, const std::vector<size_t> &global_size, std::vector<size_t> *local_size) override;
int InitNHWCShape();
private:
int InitNHWC();
cl::Kernel kernel_;
std::vector<size_t> nhwc_shape_;
size_t N_{1};
size_t H_{1};
size_t W_{1};
size_t C_{1};
};
} // namespace mindspore::kernel

34
mindspore/lite/src/runtime/kernel/opencl/utils.cc
View File

@ -237,33 +237,33 @@ void PrintTensor(lite::Tensor *tensor, int num, const std::string &out_file) {
}
auto runtime_wrapper = lite::opencl::OpenCLRuntimeWrapper();
auto runtime = runtime_wrapper.GetInstance();
runtime->SyncCommandQueue();
auto allocator = runtime->GetAllocator();
auto origin_data = tensor->data_c();
allocator->MapBuffer(origin_data, CL_MAP_READ | CL_MAP_WRITE, nullptr, true);
tensor->SetData(origin_data);
runtime->SyncCommandQueue();
allocator->MapBuffer(origin_data, CL_MAP_READ, nullptr, true);
auto Batch = tensor->Batch();
auto Height = tensor->shape().size() == 4 ? tensor->Height() : 1;
auto Width = tensor->shape().size() == 4 ? tensor->Width() : 1;
auto SLICES = UP_DIV(tensor->Channel(), C4NUM);
auto shape = tensor->shape();
auto N = shape.size() > 0 ? shape[0] : 1;
auto H = shape.size() > 1 ? shape[1] : 1;
auto W = shape.size() > 2 ? shape[2] : 1;
auto C = shape.size() > 3 ? shape[3] : 1;
auto SLICES = UP_DIV(C, C4NUM);
auto ElementsC4Num = N * H * W * UP_ROUND(C, C4NUM);
auto alignment = runtime->GetImagePitchAlignment();
auto dtype_size = tensor->data_type() == kNumberTypeFloat16 ? sizeof(cl_half4) : sizeof(cl_float4);
auto row_pitch = (Width * SLICES + alignment - 1) / alignment * alignment * dtype_size;
auto row_size = Width * SLICES * dtype_size;
std::vector<char> data(tensor->Size());
for (int i = 0; i < Batch * Height; ++i) {
auto FLT4_size = tensor->data_type() == kNumberTypeFloat16 ? sizeof(cl_half4) : sizeof(cl_float4);
auto row_pitch = (W * SLICES + alignment - 1) / alignment * alignment * FLT4_size;
auto row_size = W * SLICES * FLT4_size;
std::vector<char> data(N * H * row_size);
for (int i = 0; i < N * H; ++i) {
memcpy(static_cast<char *>(data.data()) + i * row_size, static_cast<char *>(origin_data) + i * row_pitch, row_size);
}
std::cout << "shape=(";
for (auto x : tensor->shape()) {
for (auto x : shape) {
printf("%3d,", x);
}
printf("): ");
for (size_t i = 0; i < num && i < tensor->ElementsNum(); ++i) {
for (size_t i = 0; i < num && i < ElementsC4Num; ++i) {
if (tensor->data_type() == kNumberTypeFloat16)
printf("%zu %6.3f | ", i, (reinterpret_cast<float16_t *>(data.data()))[i]);
else
@ -272,7 +272,7 @@ void PrintTensor(lite::Tensor *tensor, int num, const std::string &out_file) {
printf("\n");
if (!out_file.empty()) {
Write2File(data.data(), out_file, tensor->Size());
Write2File(data.data(), out_file, data.size());
}
allocator->UnmapBuffer(origin_data);
}