mmdeploy/csrc/preprocess/cuda/pad_impl.cpp
lvhan028 36124f6205
Merge sdk (#251)
* check in cmake

* move backend_ops to csrc/backend_ops

* check in preprocess, model, some codebase and their c-apis

* check in CMakeLists.txt

* check in parts of test_csrc

* commit everything else

* add readme

* update core's BUILD_INTERFACE directory

* skip codespell on third_party

* update trt_net and ort_net's CMakeLists

* ignore clion's build directory

* check in pybind11

* add onnx.proto. Remove MMDeploy's dependency on ncnn's source code

* export MMDeployTargets only when MMDEPLOY_BUILD_SDK is ON

* remove useless message

* target include directory is wrong

* change target name from mmdeploy_ppl_net to mmdeploy_pplnn_net

* skip install directory

* update project's cmake

* remove useless code

* set CMAKE_BUILD_TYPE to Release by force if it isn't set by user

* update custom ops CMakeLists

* pass object target's source lists

* fix lint end-of-file

* fix lint: trailing whitespace

* fix codespell hook

* remove bicubic_interpolate to csrc/backend_ops/

* set MMDEPLOY_BUILD_SDK OFF

* change custom ops build command

* add spdlog installation command

* update docs on how to checkout pybind11

* move bicubic_interpolate to backend_ops/tensorrt directory

* remove useless code

* correct cmake

* fix typo

* fix typo

* fix install directory

* correct sdk's readme

* set cub dir when cuda version < 11.0

* change directory where clang-format will apply to

* fix build command

* add .clang-format

* change clang-format style from google to file

* reformat csrc/backend_ops

* format sdk's code

* turn off clang-format for some files

* add -Xcompiler=-fno-gnu-unique

* fix trt topk initialize

* check in config for sdk demo

* update cmake script and csrc's readme

* correct config's path

* add cuda include directory, otherwise compile failed in case of tensorrt8.2

* clang-format onnx2ncnn.cpp

Co-authored-by: zhangli <lzhang329@gmail.com>
Co-authored-by: grimoire <yaoqian@sensetime.com>
2021-12-07 10:57:55 +08:00

109 lines
4.0 KiB
C++

// Copyright (c) OpenMMLab. All rights reserved.
#include "core/utils/formatter.h"
#include "ppl/cv/cuda/copymakeborder.h"
#include "preprocess/transform/pad.h"
#include "preprocess/transform/transform_utils.h"
using namespace std;
using namespace ppl::cv::cuda;
namespace mmdeploy {
namespace cuda {
class PadImpl : public ::mmdeploy::PadImpl {
public:
explicit PadImpl(const Value& args) : ::mmdeploy::PadImpl(args) {
map<string, ppl::cv::BorderType> border_map{{"constant", ppl::cv::BORDER_TYPE_CONSTANT},
{"edge", ppl::cv::BORDER_TYPE_REPLICATE},
{"reflect", ppl::cv::BORDER_TYPE_REFLECT_101},
{"symmetric", ppl::cv::BORDER_TYPE_REFLECT}};
if (border_map.find(arg_.padding_mode) == border_map.end()) {
ERROR("unsupported padding_mode '{}'", arg_.padding_mode);
throw_exception(eNotSupported);
}
padding_mode_ = border_map[arg_.padding_mode];
}
protected:
Result<Tensor> PadImage(const Tensor& img, const array<int, 4>& padding) override {
OUTCOME_TRY(auto src_tensor, MakeAvailableOnDevice(img, device_, stream_));
auto desc = src_tensor.desc();
int height = desc.shape[1];
int width = desc.shape[2];
int c = desc.shape[3];
auto dst_height = height + padding[1] + padding[3];
auto dst_width = width + padding[0] + padding[2];
TensorShape dst_shape{1, dst_height, dst_width, c};
TensorDesc dst_desc{device_, desc.data_type, dst_shape, ""};
Tensor dst_tensor(dst_desc);
ppl::common::RetCode ret = 0;
cudaStream_t stream = ::mmdeploy::GetNative<cudaStream_t>(stream_);
if (desc.data_type == DataType::kFLOAT) {
auto src_buffer = src_tensor.data<float>();
auto dst_buffer = dst_tensor.data<float>();
if (3 == c) {
ret = CopyMakeBorder<float, 3>(stream, height, width, width * c, src_buffer, dst_width * c,
dst_buffer, padding[1], padding[3], padding[0], padding[2],
padding_mode_, arg_.pad_val);
} else if (1 == c) {
ret = CopyMakeBorder<float, 1>(stream, height, width, width * c, src_buffer, dst_width * c,
dst_buffer, padding[1], padding[3], padding[0], padding[2],
padding_mode_, arg_.pad_val);
} else {
ERROR("unsupported channels {}", c);
assert(0);
return Status(eNotSupported);
}
} else if (desc.data_type == DataType::kINT8) {
auto src_buffer = src_tensor.data<uint8_t>();
auto dst_buffer = dst_tensor.data<uint8_t>();
if (3 == c) {
ret = CopyMakeBorder<ppl::cv::uchar, 3>(
stream, height, width, width * c, src_buffer, dst_width * c, dst_buffer, padding[1],
padding[3], padding[0], padding[2], padding_mode_, (ppl::cv::uchar)arg_.pad_val);
} else if (1 == c) {
ret = CopyMakeBorder<ppl::cv::uchar, 1>(
stream, height, width, width * c, src_buffer, dst_width * c, dst_buffer, padding[1],
padding[3], padding[0], padding[2], padding_mode_, (ppl::cv::uchar)arg_.pad_val);
} else {
ERROR("unsupported channels {}", c);
assert(0);
return Status(eNotSupported);
}
} else {
ERROR("unsupported data type {}", desc.data_type);
assert(0);
return Status(eNotSupported);
}
if (ret != 0) {
ERROR("unexpected exception happened");
assert(0);
return Status(eNotSupported);
}
return dst_tensor;
}
private:
ppl::cv::BorderType padding_mode_;
};
class PadCreator : public Creator<::mmdeploy::PadImpl> {
public:
const char* GetName() const override { return "cuda"; }
int GetVersion() const override { return 1; }
ReturnType Create(const Value& args) override { return make_unique<PadImpl>(args); }
};
} // namespace cuda
} // namespace mmdeploy
using ::mmdeploy::PadImpl;
using mmdeploy::cuda::PadCreator;
REGISTER_MODULE(PadImpl, PadCreator);