测试代码优化

ming
1 parent 2b980c5a
Showing 3 changed files with 121 additions and 67 deletions
.vscode/launch.json
src/Makefile
src/main.cpp
@@ -6,7 +6,25 @@
         "type": "cppdbg",
         "request": "launch",
         "program": "${workspaceFolder}/bin/lib/test",
-        "args": ["rtsp://176.10.0.4:8554/stream","0"],
+        "args": ["rtsp://122.97.218.170:8604/openUrl/V5nXRHa?params=eyJwcm90b2NhbCI6InJ0c3AiLCJjbGllbnRUeXBlIjoib3Blbl9hcGkiLCJleHByaWVUaW1lIjotMSwicHJvdG9jb2wiOiJydHNwIiwiZXhwaXJlVGltZSI6MzAwLCJlbmFibGVNR0MiOnRydWUsImV4cGFuZCI6InN0YW5kYXJkPXJ0c3Amc3RyZWFtZm9ybT1ydHAiLCJhIjoiMTBjZjM4N2JjY2Y5NDg3YzhjNWYzNjE2M2ViMWUyNTJ8MXwwfDEiLCJ0IjoxfQ==","3"],
+        "stopAtEntry": false,
+        "cwd": "${workspaceFolder}/bin/lib",
+        "environment": [],
+        "externalConsole": false,
+        "MIMode": "gdb",
+        "setupCommands": [
+          {
+            "description": "Enable pretty-printing for gdb",
+            "text": "-enable-pretty-printing",
+            "ignoreFailures": true
+          }
+        ]
+      },{
+        "name": "ffmpeg",
+        "type": "cppdbg",
+        "request": "launch",
+        "program": "${workspaceFolder}/bin/lib/ffmpeg_g",
+        "args": ["-vsync","0","-hwaccel","cuvid","-hwaccel_device","1","-c:v","h264_cuvid","-i","rtmp://192.168.10.56:1935/objecteye/116","-c:a","copy","-vf","scale_npp=800:480","-c:v","h264","/mnt/data/cmhu/FFNvDecoder/data/output1.mp4"],
         "stopAtEntry": false,
         "cwd": "${workspaceFolder}/bin/lib",
         "environment": [],
-
 XX = g++
-NVCC = nvcc
+
+CUDA_ROOT = /usr/local/cuda-11.1
+NVCC = $(CUDA_ROOT)/bin/nvcc
 PROJECT_ROOT= /mnt/data/cmhu/FFNvDecoder
 DEPEND_DIR = $(PROJECT_ROOT)/bin
 SRC_ROOT = $(PROJECT_ROOT)/src
-CUDA_ROOT = /usr/local/cuda-11.1
 TARGET= $(DEPEND_DIR)/lib/test
-
 INCLUDE= -I $(DEPEND_DIR)/include \
   -I $(CUDA_ROOT)/include \
   -I $(SRC_ROOT)/common/inc \
@@ -19,7 +18,8 @@ INCLUDE= -I $(DEPEND_DIR)/include \
   -I $(SRC_ROOT)\
 LIBSPATH= -L $(DEPEND_DIR)/lib -lavformat -lavcodec -lswscale -lavutil -lavfilter -lswresample -lavdevice \
-   -L $(CUDA_ROOT)/lib64 -lcuda -lcudart -lnvcuvid -lcurand -lcublas -lnvjpeg \
+   -L $(CUDA_ROOT)/lib64 -lcuda -lcudart -lnvcuvid \
+   -L $(CUDA_ROOT)/lib64/stubs -lcurand -lcublas -lnvjpeg \
 CFLAGS= -g -fPIC -O0 $(INCLUDE) -pthread -lrt -lz -std=c++11 -fvisibility=hidden -Wl,-Bsymbolic -ldl
 	# -DUNICODE -D_UNICODE
@@ -47,6 +47,4 @@ $(TARGET):$(OBJS) $(CU_OBJS)
 	$(NVCC) $(NFLAGS) -o $@ $< 
 clean:
-	rm -f *.o $(TARGET)
-
-
+	rm -f *.o $(TARGET)
 \ No newline at end of file
@@ -17,10 +17,79 @@ int sum2 = 0;
 cudaStream_t stream[2];
+string data_home = "/mnt/data/cmhu/FFNvDecoder/data/";
+
+
+#define checkCudaErrors(S) do {CUresult  status; \
+        status = S; \
+        if (status != CUDA_SUCCESS ) std::cout << __LINE__ <<" checkCudaErrors - status = " << status << std::endl; \
+        } while (false)
+
+
+static void gpu_helper(int gpuid)
+{
+    cudaSetDevice(gpuid);
+
+    // int *dn;
+    // cudaMalloc((void **)&dn, 1 * sizeof(int));
+
+	size_t free_byte;
+	size_t total_byte;
+
+	CUresult cuda_status = cuMemGetInfo(&free_byte, &total_byte);
+
+	const char *pStr = nullptr;
+	if (CUDA_SUCCESS != cuda_status) {
+		cuGetErrorString(cuda_status, &pStr);
+		printf("Error: cudaMemGetInfo fails, %s \n", pStr);
+		return;
+	}
+
+	double free_db = (double)free_byte;
+	double total_db = (double)total_byte;
+	double used_db_1 = (total_db - free_db) / 1024.0 / 1024.0;
+
+	std::cout <<"显存已使用 " << used_db_1 << " MB\n";
+
+    // cudaFree(dn);
+}
+
+int CheckCUDAProperty( int devId )
+{
+    cuInit(0);
+
+	CUdevice dev = devId;
+	size_t memSize = 0;
+	char devName[256] = {0};
+	int major = 0, minor = 0;
+	CUresult rlt = CUDA_SUCCESS;
+
+    rlt = cuDeviceGetAttribute(&major, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, dev);
+    checkCudaErrors( rlt );
+
+    rlt = cuDeviceGetAttribute(&minor, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR, dev);
+	checkCudaErrors( rlt );
+
+	rlt = cuDeviceGetName( devName, sizeof( devName ), dev );
+	checkCudaErrors( rlt );
+
+	printf( "Using GPU Device %d: %s has SM %d.%d compute capability\n",
+		    dev, devName, major, minor );
+
+	rlt = cuDeviceTotalMem( &memSize, dev );
+	checkCudaErrors( rlt );
+
+	printf( "Total amount of global memory:   %4.4f MB\n",
+		   (float)memSize / ( 1024 * 1024 ) );
+
+	return 0;
+}
+
 /**
  * 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
  * */
 void postDecoded(const void * userPtr, AVFrame * gpuFrame){
+
     FFNvDecoder* decoder = (FFNvDecoder*)userPtr;
     if (decoder!= nullptr)
     {
@@ -39,7 +108,7 @@ void postDecoded(const void * userPtr, AVFrame * gpuFrame){
                 if (gpuFrame->format == AV_PIX_FMT_CUDA)
                 {   
-                    cout << "gpuid = " << atoi(decoder->m_cfg.gpuid.c_str()) << endl;
+                    // cout << "gpuid = " << atoi(decoder->m_cfg.gpuid.c_str()) << endl;
                     cudaSetDevice(atoi(decoder->m_cfg.gpuid.c_str()));
                     cudaError_t cudaStatus;
                     if(pHwRgb[0] == nullptr){
@@ -54,7 +123,7 @@ void postDecoded(const void * userPtr, AVFrame * gpuFrame){
                         return;
                     }
-                    string path = "/mnt/f/fiss/data/" + decoder->getName() + "/" + to_string(sum) + ".jpg";
+                    string path = data_home + decoder->getName() + ".jpg";
                     saveJpeg(path.c_str(), pHwRgb[0], gpuFrame->width, gpuFrame->height, stream[0]);  // 验证 CUDAToRGB 
                 }
             } else if (decoder->getName() == "dec2") 
@@ -64,7 +133,7 @@ void postDecoded(const void * userPtr, AVFrame * gpuFrame){
                 if (gpuFrame->format == AV_PIX_FMT_CUDA)
                 {   
-                    cout << "gpuid = " << atoi(decoder->m_cfg.gpuid.c_str()) << endl;
+                    // cout << "gpuid = " << atoi(decoder->m_cfg.gpuid.c_str()) << endl;
                     cudaSetDevice(atoi(decoder->m_cfg.gpuid.c_str()));
                     cudaError_t cudaStatus;
                     if(pHwRgb[1] == nullptr){
@@ -79,7 +148,7 @@ void postDecoded(const void * userPtr, AVFrame * gpuFrame){
                         return;
                     }
-                    string path = "/mnt/f/fiss/data/" + decoder->getName() + "/" + to_string(sum) + ".jpg";
+                    string path = data_home + decoder->getName() + ".jpg";
                     saveJpeg(path.c_str(), pHwRgb[1], gpuFrame->width, gpuFrame->height, stream[1]);  // 验证 CUDAToRGB 
                 }
             }
@@ -126,12 +195,12 @@ void postDecoded0(const void * userPtr, AVFrame * gpuFrame){
                 // long time_using = end_time - start_time;
                 // double time_per_frame = double(time_using)/count_std ;
                 // cout << count_std << "帧用时:" << time_using << "ms 每帧用时：" << time_per_frame << "ms" << endl;
-                cout << "keyframe: " << gpuFrame->key_frame << " width: " << gpuFrame->width << " height: "<< gpuFrame->height << endl;
-                cout << gpuFrame->pts << endl;
+                cout << decoder->getName() << " keyframe: " << gpuFrame->key_frame << " width: " << gpuFrame->width << " height: "<< gpuFrame->height << endl;
+                // cout << gpuFrame->pts << endl;
                 count_flag = false;
             }
-            cout << "帧数：" << sum << endl;
+            // cout << "帧数：" << sum << endl;
             if (gpuFrame->format == AV_PIX_FMT_CUDA)
             {   
@@ -149,7 +218,7 @@ void postDecoded0(const void * userPtr, AVFrame * gpuFrame){
                     return;
                 }
-                string path = "/mnt/f/fiss/data/" + to_string(sum) + ".jpg";
+                string path = data_home + decoder->getName() + ".jpg";
                 saveJpeg(path.c_str(), pHwData, gpuFrame->width, gpuFrame->height, nullptr);  // 验证 CUDAToRGB 
             }
         }
@@ -166,8 +235,9 @@ void decode_finished_cbk(const void* userPtr){
 // string test_uri = "rtsp://176.10.0.2:8554/stream";
 // string test_uri = "/mnt/f/fiss/test_data/h265.mp4";
 string test_uri = "rtsp://176.10.0.4:8554/stream";
+char* gpuid = "0";
-void createDecode(int index){
+void createDecode(int index, const char* gpu_id){
     FFNvDecoderManager* pDecManager = FFNvDecoderManager::getInstance();
     MgrDecConfig config;
     config.name = "dec" + to_string(index);
@@ -176,14 +246,15 @@ void createDecode(int index){
     config.cfg.decode_finished_cbk = decode_finished_cbk;
     config.cfg.force_tcp = true;
-    if (index % 2 == 0)
-    {
-        config.cfg.gpuid = "0";
-    }
-    else
-    {
-        config.cfg.gpuid = "0";
-    }
+    config.cfg.gpuid = gpu_id;
+    // if (index % 2 == 0)
+    // {
+    //     config.cfg.gpuid = gpu_id;
+    // }
+    // else
+    // {
+    //     config.cfg.gpuid = "0";
+    // }
     FFNvDecoder* decoder = pDecManager->createDecoder(config);
     if (!decoder)
@@ -194,42 +265,6 @@ void createDecode(int index){
     pDecManager->startDecodeByName(config.name);
 }
-#define checkCudaErrors(S) do {CUresult  status; \
-        status = S; \
-        if (status != CUDA_SUCCESS ) std::cout << __LINE__ <<" checkCudaErrors - status = " << status << std::endl; \
-        } while (false)
-
-int CheckCUDAProperty( int devId )
-{
-    cuInit(0);
-
-	CUdevice dev = devId;
-	size_t memSize = 0;
-	char devName[256] = {0};
-	int major = 0, minor = 0;
-	CUresult rlt = CUDA_SUCCESS;
-
-    rlt = cuDeviceGetAttribute(&major, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MAJOR, dev);
-    checkCudaErrors( rlt );
-
-    rlt = cuDeviceGetAttribute(&minor, CU_DEVICE_ATTRIBUTE_COMPUTE_CAPABILITY_MINOR, dev);
-	checkCudaErrors( rlt );
-
-	rlt = cuDeviceGetName( devName, sizeof( devName ), dev );
-	checkCudaErrors( rlt );
-
-	printf( "Using GPU Device %d: %s has SM %d.%d compute capability\n",
-		    dev, devName, major, minor );
-
-	rlt = cuDeviceTotalMem( &memSize, dev );
-	checkCudaErrors( rlt );
-
-	printf( "Total amount of global memory:   %4.4f MB\n",
-		   (float)memSize / ( 1024 * 1024 ) );
-
-	return 0;
-}
-
 void logFF(void *, int level, const char *fmt, va_list ap)
 {
     vfprintf(stdout, fmt, ap);
@@ -239,18 +274,19 @@ void logFF(void *, int level, const char *fmt, va_list ap)
 int main(int argc, char* argv[]){
     test_uri = argv[1];
+    gpuid = argv[2];
     cout << test_uri << endl;
     // av_log_set_callback(&logFF);
-    CheckCUDAProperty(0);
+    CheckCUDAProperty(atoi(gpuid));
     FFNvDecoderManager* pDecManager = FFNvDecoderManager::getInstance();
-    // int count = 99;
+    // int count = 4;
     // for (size_t i = 0; i < count ; i++)
     // {
-    //     createDecode(i);
+    //     createDecode(i, gpuid);
     // }
     MgrDecConfig config;
@@ -259,7 +295,7 @@ int main(int argc, char* argv[]){
     config.cfg.post_decoded_cbk = postDecoded0;
     config.cfg.decode_finished_cbk = decode_finished_cbk;
     config.cfg.force_tcp = true;
-    config.cfg.gpuid = "0";
+    config.cfg.gpuid = gpuid;
     FFNvDecoder* dec2 = pDecManager->createDecoder(config);
     if (!dec2)
     {
@@ -277,12 +313,14 @@ int main(int argc, char* argv[]){
     pthread_create(&m_decode_thread,0,
         [](void* arg)
         {
+            // cudaSetDevice(atoi(gpuid));
             while (true)
             {
                 std::this_thread::sleep_for(std::chrono::milliseconds(5000));
                 FFNvDecoderManager* pDecManager = FFNvDecoderManager::getInstance();
                 int count = pDecManager->count();
-                cout << "当前运行路数： " << pDecManager->count() << endl;
+                cout << "当前时间戳：" << get_cur_time() <<  " 当前运行路数：" << pDecManager->count() << endl;
+                // gpu_helper(atoi(gpuid));
                 if (count <= 0)
                 {
                     break;