Hu Chunming / FFNvDecoder_WIN

Commit 615017759733017e12ae70bada5092928ebcc311

Authored by Hu Chunming
1 parent a11c5683

添加showFrame

Inline Side-by-side
Showing 7 changed files with 61 additions and 59 deletions
FFNvDecoder/FFNvDecoder.vcxproj
  View file @6150177
... ... @@ -48,13 +48,13 @@
48 48 <Optimization>Disabled</Optimization>
49 49 <PreprocessorDefinitions>WIN32;WIN64;_DEBUG;_CONSOLE;%(PreprocessorDefinitions)</PreprocessorDefinitions>
50 50 <AdditionalOptions>/utf-8</AdditionalOptions>
51   - <AdditionalIncludeDirectories>..\3rdparty\ffmpeg-5.0.1-win64-dev\include;./;./common/inc;./common/UtilNPP;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
  51 + <AdditionalIncludeDirectories>..\3rdparty\ffmpeg-5.0.1-win64-dev\include;./;./common/inc;./common/UtilNPP;D:\win_dev\opencv\build\include;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
52 52 </ClCompile>
53 53 <Link>
54 54 <GenerateDebugInformation>true</GenerateDebugInformation>
55 55 <SubSystem>Console</SubSystem>
56   - <AdditionalDependencies>avcodec.lib;avdevice.lib;avfilter.lib;avformat.lib;avutil.lib;postproc.lib;swresample.lib;swscale.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;nvjpeg.lib;%(AdditionalDependencies)</AdditionalDependencies>
57   - <AdditionalLibraryDirectories>..\3rdparty\ffmpeg-5.0.1-win64-dev\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
  56 + <AdditionalDependencies>avcodec.lib;avdevice.lib;avfilter.lib;avformat.lib;avutil.lib;postproc.lib;swresample.lib;swscale.lib;cudart_static.lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;nvjpeg.lib;opencv_world455d.lib;freeglut.lib;glew64.lib;%(AdditionalDependencies)</AdditionalDependencies>
  57 + <AdditionalLibraryDirectories>..\3rdparty\ffmpeg-5.0.1-win64-dev\lib;D:\win_dev\opencv\build\x64\vc14\lib;../3rdparty/gl;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
58 58 </Link>
59 59 <CudaCompile>
60 60 <TargetMachinePlatform>64</TargetMachinePlatform>
... ... @@ -80,7 +80,6 @@
80 80 </CudaCompile>
81 81 </ItemDefinitionGroup>
82 82 <ItemGroup>
83   - <CudaCompile Include="cudaHeader.cu" />
84 83 <CudaCompile Include="NV12ToRGB.cu" />
85 84 </ItemGroup>
86 85 <ItemGroup>
... ... @@ -94,7 +93,6 @@
94 93 </ItemGroup>
95 94 <ItemGroup>
96 95 <ClInclude Include="check_tool.h" />
97   - <ClInclude Include="cudaHeader.h" />
98 96 <ClInclude Include="cuda_kernels.h" />
99 97 <ClInclude Include="FFCuContextManager.h" />
100 98 <ClInclude Include="FFNvDecoder.h" />
... ...
FFNvDecoder/FFNvDecoder.vcxproj.filters
  View file @6150177
1 1 <?xml version="1.0" encoding="utf-8"?>
2 2 <Project ToolsVersion="4.0" xmlns="http://schemas.microsoft.com/developer/msbuild/2003">
3 3 <ItemGroup>
4   - <CudaCompile Include="cudaHeader.cu">
5   - <Filter>cu_src</Filter>
6   - </CudaCompile>
7 4 <CudaCompile Include="NV12ToRGB.cu">
8 5 <Filter>cu_src</Filter>
9 6 </CudaCompile>
... ... @@ -58,9 +55,6 @@
58 55 <ClInclude Include="FrameQueue.h">
59 56 <Filter>include</Filter>
60 57 </ClInclude>
61   - <ClInclude Include="cudaHeader.h">
62   - <Filter>cu_src</Filter>
63   - </ClInclude>
64 58 <ClInclude Include="cuda_kernels.h">
65 59 <Filter>cu_src</Filter>
66 60 </ClInclude>
... ...
FFNvDecoder/NV12ToRGB.cu
  View file @6150177
... ... @@ -52,9 +52,9 @@ namespace cuda_common
52 52 return x;
53 53 }
54 54 }
55   -
56 55 // CUDA kernel for outputing the final RGB output from NV12;
57   - __global__ void NV12ToRGB_drvapi2(uint32 *srcImage, size_t nSourcePitch, unsigned char *dstImage, int width, int height)
  56 + extern "C"
  57 + __global__ void NV12ToRGB_drvapi2(uint32 *srcImage, size_t nSourcePitch, unsigned char *dstImage, int width, int height)
58 58 {
59 59  
60 60 int32 x, y;
... ... @@ -169,8 +169,9 @@ namespace cuda_common
169 169  
170 170 }
171 171  
172   - // CUDA kernel for outputing the final RGB output from NV12;
173   - __global__ void CUDAToBGR_drvapi(uint32 *dataY, uint32 *dataUV, size_t pitchY, size_t pitchUV, unsigned char *dstImage, int width, int height)
  172 + // CUDA kernel for outputing the final RGB output from NV12;
  173 + extern "C"
  174 + __global__ void CUDAToBGR_drvapi(uint32 *dataY, uint32 *dataUV, size_t pitchY, size_t pitchUV, unsigned char *dstImage, int width, int height)
174 175 {
175 176  
176 177 int32 x, y;
... ... @@ -307,7 +308,7 @@ namespace cuda_common
307 308 {
308 309 dim3 block(32, 16, 1);
309 310 dim3 grid((width + (2 * block.x - 1)) / (2 * block.x), (height + (block.y - 1)) / block.y, 1);
310   - NV12ToRGB_drvapi2 <<< grid, block >>>((uint32 *)d_srcNV12, nSourcePitch, d_dstRGB, width, height);
  311 + NV12ToRGB_drvapi2 << < grid, block >> >((uint32 *)d_srcNV12, nSourcePitch, d_dstRGB, width, height);
311 312 cudaError_t cudaStatus = cudaGetLastError();
312 313 if (cudaStatus != cudaSuccess) {
313 314 fprintf(stderr, "NV12ToRGB_drvapi launch failed: %s\n", cudaGetErrorString(cudaStatus));
... ... @@ -327,7 +328,7 @@ namespace cuda_common
327 328 {
328 329 dim3 block(32, 16, 1);
329 330 dim3 grid((width + (2 * block.x - 1)) / (2 * block.x), (height + (block.y - 1)) / block.y, 1);
330   - CUDAToBGR_drvapi <<< grid, block >>>((uint32 *)dataY, (uint32 *)dataUV, pitchY, pitchUV, d_dstRGB, width, height);
  331 + CUDAToBGR_drvapi << < grid, block >> >((uint32 *)dataY, (uint32 *)dataUV, pitchY, pitchUV, d_dstRGB, width, height);
331 332 cudaError_t cudaStatus = cudaGetLastError();
332 333 if (cudaStatus != cudaSuccess) {
333 334 fprintf(stderr, "NV12ToRGB_drvapi launch failed: %s\n", cudaGetErrorString(cudaStatus));
... ...
FFNvDecoder/cudaHeader.cu deleted
View file @a11c568
1   -#include"cudaHeader.h"
2   -#include<iostream>
3   -
4   -#include "cuda_runtime.h"
5   -#include <cuda.h>
6   -
7   -//核函数，计算a+b
8   -__global__ void add(int a,int b,int *c)
9   -{
10   - //保存a+b的计算结果
11   - *c=a+b;
12   -}
13   -
14   -
15   -//cuda测试函数的实现
16   -void cudaTest()
17   -{
18   - int c = 0;
19   - //在gpu上开辟一个相同的内存
20   - int *deviceC;
21   - cudaMalloc((void**)&deviceC,sizeof(int));
22   - //调用核函数
23   - add<<<1,1>>>(3,7,deviceC);
24   - //把计算结果复制到cpu上
25   - cudaMemcpy(&c,deviceC,sizeof(int),cudaMemcpyDeviceToHost);
26   - //展示计算结果
27   - std::cout<<c<<std::endl;
28   - //释放内存
29   - cudaFree(deviceC);
30   -}
31 0 \ No newline at end of file
FFNvDecoder/cudaHeader.h deleted
View file @a11c568
1   -#ifndef _CUDA_HEADER_H_
2   -#define _CUDA_HEADER_H_
3   -//用于测试cuda能否正常工作的函数
4   -void cudaTest();
5   -#endif
6 0 \ No newline at end of file
FFNvDecoder/dec.jpg deleted
View file @a11c568

638 KB
FFNvDecoder/main.cpp
  View file @6150177
... ... @@ -7,12 +7,13 @@
7 7  
8 8 #include "check_tool.h"
9 9  
10   -#include "cudaHeader.h"
11   -
12 10 #include "cuda_kernels.h"
13 11 #include "NvJpegEncoder.h"
14 12  
  13 +#include "opencv2\opencv.hpp"
  14 +
15 15 using namespace std;
  16 +using namespace cv;
16 17  
17 18 unsigned char *pHwRgb[2] = {nullptr, nullptr};
18 19  
... ... @@ -44,6 +45,50 @@ void saveFrame(AVFrame * gpuFrame, string file_name) {
44 45 pHwData = nullptr;
45 46 }
46 47  
  48 +mutex m_mutex_show;
  49 +void showFrame(AVFrame * gpuFrame) {
  50 + std::lock_guard<std::mutex> l(m_mutex_show);
  51 +
  52 + unsigned char *pHwData = nullptr;
  53 + cudaError_t cudaStatus = cudaMalloc((void **)&pHwData, 3 * gpuFrame->width * gpuFrame->height * sizeof(unsigned char));
  54 +
  55 + cuda_common::setColorSpace(ITU709, 0);
  56 + cudaStatus = cuda_common::CUDAToBGR((CUdeviceptr)gpuFrame->data[0], (CUdeviceptr)gpuFrame->data[1], gpuFrame->linesize[0], gpuFrame->linesize[1], pHwData, gpuFrame->width, gpuFrame->height);
  57 + cudaDeviceSynchronize();
  58 + if (cudaStatus != cudaSuccess) {
  59 + cout << "CUDAToBGR failed !!!" << endl;
  60 + return;
  61 + }
  62 +
  63 +
  64 +
  65 + unsigned char * pHwRgb = pHwData;
  66 + int channel = 3;
  67 + int width = gpuFrame->width;
  68 + int height = gpuFrame->height;
  69 +
  70 + if (pHwRgb != nullptr && channel > 0 && width > 0 && height > 0) {
  71 + int nSize = channel * height * width;
  72 + unsigned char* cpu_data = new unsigned char[nSize];
  73 +
  74 + cudaMemcpy(cpu_data, pHwRgb, nSize * sizeof(unsigned char), cudaMemcpyDeviceToHost);
  75 + cudaDeviceSynchronize();
  76 +
  77 + cv::Mat img_(height, width, CV_8UC3, cpu_data);
  78 + bool bWrite = cv::imwrite("dec0.jpg", img_);
  79 +
  80 + imshow("show", img_);
  81 + waitKey(0);
  82 +
  83 + delete[] cpu_data;
  84 + cpu_data = nullptr;
  85 +
  86 + }
  87 +
  88 + cudaFree(pHwData);
  89 + pHwData = nullptr;
  90 +}
  91 +
47 92 /**
48 93 * 注意： gpuFrame 在解码器设置的显卡上，后续操作要十分注意这一点，尤其是多线程情况
49 94 * */
... ... @@ -64,6 +109,7 @@ void postDecoded(const void * userPtr, AVFrame * gpuFrame){
64 109 cudaSetDevice(atoi(decoder->m_cfg.gpuid.c_str()));
65 110  
66 111 saveFrame(gpuFrame, decoder->getName());
  112 + showFrame(gpuFrame);
67 113 }
68 114 }
69 115 }
... ... @@ -82,7 +128,7 @@ static long long get_cur_time(){
82 128 return tpMicro.time_since_epoch().count();
83 129 }
84 130  
85   -static int sum = 0;
  131 +static int suming = 0;
86 132 unsigned char *pHwData = nullptr;
87 133  
88 134 void postDecoded0(const void * userPtr, AVFrame * gpuFrame){
... ... @@ -101,7 +147,7 @@ void postDecoded0(const void * userPtr, AVFrame * gpuFrame){
101 147 end_time = start_time = get_cur_time();
102 148 }
103 149 count_num++;
104   - sum ++ ;
  150 + suming ++ ;
105 151 if (count_num >= count_std)
106 152 {
107 153 // end_time = get_cur_time();
... ... @@ -112,7 +158,7 @@ void postDecoded0(const void * userPtr, AVFrame * gpuFrame){
112 158  
113 159 count_flag = false;
114 160 }
115   - cout << "帧数：" << sum << endl;
  161 + cout << "帧数：" << suming << endl;
116 162 }
117 163 }
118 164 }
... ... @@ -163,8 +209,6 @@ void logFF(void *, int level, const char *fmt, va_list ap)
163 209  
164 210 int main(int argc, char* argv[]) {
165 211  
166   - cudaTest();
167   -
168 212 printf("start \n");
169 213 if (argc != 3) {
170 214 fprintf(stderr, "./xxx uri gpu_id\n");
... ...