summaryrefslogtreecommitdiffstats
path: root/cuda/2d
diff options
context:
space:
mode:
Diffstat (limited to 'cuda/2d')
-rw-r--r--cuda/2d/arith.cu14
-rw-r--r--cuda/2d/fan_bp.cu15
-rw-r--r--cuda/2d/fan_fp.cu11
-rw-r--r--cuda/2d/par_bp.cu10
-rw-r--r--cuda/2d/par_fp.cu19
-rw-r--r--cuda/2d/sart.cu2
-rw-r--r--cuda/2d/util.cu69
7 files changed, 52 insertions, 88 deletions
diff --git a/cuda/2d/arith.cu b/cuda/2d/arith.cu
index aa0edae..45622d0 100644
--- a/cuda/2d/arith.cu
+++ b/cuda/2d/arith.cu
@@ -451,7 +451,7 @@ void processData(float* pfOut, unsigned int pitch, unsigned int width, unsigned
devtoD<op, 32><<<gridSize, blockSize>>>(pfOut, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), __FUNCTION__);
}
template<typename op>
@@ -462,7 +462,7 @@ void processData(float* pfOut, float fParam, unsigned int pitch, unsigned int wi
devFtoD<op, 32><<<gridSize, blockSize>>>(pfOut, fParam, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), __FUNCTION__);
}
template<typename op>
@@ -473,7 +473,7 @@ void processData(float* pfOut1, float* pfOut2, float fParam1, float fParam2, uns
devFFtoDD<op, 32><<<gridSize, blockSize>>>(pfOut1, pfOut2, fParam1, fParam2, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), __FUNCTION__);
}
@@ -485,7 +485,7 @@ void processData(float* pfOut, const float* pfIn, unsigned int pitch, unsigned i
devDtoD<op, 32><<<gridSize, blockSize>>>(pfOut, pfIn, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), __FUNCTION__);
}
template<typename op>
@@ -496,7 +496,7 @@ void processData(float* pfOut, const float* pfIn, float fParam, unsigned int pit
devDFtoD<op, 32><<<gridSize, blockSize>>>(pfOut, pfIn, fParam, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), __FUNCTION__);
}
template<typename op>
@@ -507,7 +507,7 @@ void processData(float* pfOut, const float* pfIn1, const float* pfIn2, float fPa
devDDFtoD<op, 32><<<gridSize, blockSize>>>(pfOut, pfIn1, pfIn2, fParam, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), __FUNCTION__);
}
template<typename op>
@@ -518,7 +518,7 @@ void processData(float* pfOut, const float* pfIn1, const float* pfIn2, unsigned
devDDtoD<op, 32><<<gridSize, blockSize>>>(pfOut, pfIn1, pfIn2, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), __FUNCTION__);
}
diff --git a/cuda/2d/fan_bp.cu b/cuda/2d/fan_bp.cu
index fc42456..2068d03 100644
--- a/cuda/2d/fan_bp.cu
+++ b/cuda/2d/fan_bp.cu
@@ -322,13 +322,12 @@ bool FanBP_internal(float* D_volumeData, unsigned int volumePitch,
else
devFanBP<false><<<dimGrid, dimBlock, 0, stream>>>(D_volumeData, volumePitch, i, dims, fOutputScale);
}
- cudaThreadSynchronize();
- cudaTextForceKernelsCompletion();
+ ok = checkCuda(cudaStreamSynchronize(stream), "FanBP");
cudaStreamDestroy(stream);
- return true;
+ return ok;
}
bool FanBP_FBPWeighted_internal(float* D_volumeData, unsigned int volumePitch,
@@ -354,13 +353,12 @@ bool FanBP_FBPWeighted_internal(float* D_volumeData, unsigned int volumePitch,
for (unsigned int i = 0; i < dims.iProjAngles; i += g_anglesPerBlock) {
devFanBP<true><<<dimGrid, dimBlock, 0, stream>>>(D_volumeData, volumePitch, i, dims, fOutputScale);
}
- cudaThreadSynchronize();
- cudaTextForceKernelsCompletion();
+ ok = checkCuda(cudaStreamSynchronize(stream), "FanBP_FBPWeighted");
cudaStreamDestroy(stream);
- return true;
+ return ok;
}
// D_projData is a pointer to one padded sinogram line
@@ -382,11 +380,8 @@ bool FanBP_SART(float* D_volumeData, unsigned int volumePitch,
(dims.iVolHeight+g_blockSliceSize-1)/g_blockSliceSize);
devFanBP_SART<<<dimGrid, dimBlock>>>(D_volumeData, volumePitch, dims, fOutputScale);
- cudaThreadSynchronize();
- cudaTextForceKernelsCompletion();
-
- return true;
+ return checkCuda(cudaThreadSynchronize(), "FanBP_SART");
}
bool FanBP(float* D_volumeData, unsigned int volumePitch,
diff --git a/cuda/2d/fan_fp.cu b/cuda/2d/fan_fp.cu
index 9bbb401..342ca4c 100644
--- a/cuda/2d/fan_fp.cu
+++ b/cuda/2d/fan_fp.cu
@@ -268,16 +268,17 @@ bool FanFP_internal(float* D_volumeData, unsigned int volumePitch,
for (unsigned int i = 0; i < dims.iVolHeight; i += g_blockSlices)
FanFPvertical<<<dimGrid, dimBlock, 0, stream2>>>(D_projData, projPitch, i, blockStart, blockEnd, dims, outputScale);
- cudaStreamDestroy(stream1);
- cudaStreamDestroy(stream2);
+ bool ok = true;
- cudaThreadSynchronize();
+ ok &= checkCuda(cudaStreamSynchronize(stream1), "fan_fp hor");
+ cudaStreamDestroy(stream1);
- cudaTextForceKernelsCompletion();
+ ok &= checkCuda(cudaStreamSynchronize(stream2), "fan_fp ver");
+ cudaStreamDestroy(stream2);
cudaFreeArray(D_dataArray);
- return true;
+ return ok;
}
bool FanFP(float* D_volumeData, unsigned int volumePitch,
diff --git a/cuda/2d/par_bp.cu b/cuda/2d/par_bp.cu
index b50b5a5..d7c3ab0 100644
--- a/cuda/2d/par_bp.cu
+++ b/cuda/2d/par_bp.cu
@@ -231,13 +231,12 @@ bool BP_internal(float* D_volumeData, unsigned int volumePitch,
else
devBP<<<dimGrid, dimBlock, 0, stream>>>(D_volumeData, volumePitch, i, dims, fOutputScale);
}
- cudaThreadSynchronize();
- cudaTextForceKernelsCompletion();
+ bool ok = checkCuda(cudaStreamSynchronize(stream), "par_bp");
cudaStreamDestroy(stream);
- return true;
+ return ok;
}
bool BP(float* D_volumeData, unsigned int volumePitch,
@@ -284,11 +283,8 @@ bool BP_SART(float* D_volumeData, unsigned int volumePitch,
(dims.iVolHeight+g_blockSliceSize-1)/g_blockSliceSize);
devBP_SART<<<dimGrid, dimBlock>>>(D_volumeData, volumePitch, angle_offset, angle_scaled_sin, angle_scaled_cos, dims, fOutputScale);
- cudaThreadSynchronize();
-
- cudaTextForceKernelsCompletion();
- return true;
+ return checkCuda(cudaThreadSynchronize(), "BP_SART");
}
diff --git a/cuda/2d/par_fp.cu b/cuda/2d/par_fp.cu
index 8c48280..e947428 100644
--- a/cuda/2d/par_fp.cu
+++ b/cuda/2d/par_fp.cu
@@ -305,8 +305,8 @@ bool FP_simple_internal(float* D_volumeData, unsigned int volumePitch,
dim3 dimGrid((blockEnd-blockStart+g_anglesPerBlock-1)/g_anglesPerBlock,
(dims.iProjDets+g_detBlockSize-1)/g_detBlockSize); // angle blocks, detector blocks
- // TODO: check if we can't immediately
- // destroy the stream after use
+ // TODO: consider limiting number of handle (chaotic) geoms
+ // with many alternating directions
cudaStream_t stream;
cudaStreamCreate(&stream);
streams.push_back(stream);
@@ -323,19 +323,16 @@ bool FP_simple_internal(float* D_volumeData, unsigned int volumePitch,
}
}
- for (std::list<cudaStream_t>::iterator iter = streams.begin(); iter != streams.end(); ++iter)
- cudaStreamDestroy(*iter);
-
- streams.clear();
-
- cudaThreadSynchronize();
+ bool ok = true;
- cudaTextForceKernelsCompletion();
+ for (std::list<cudaStream_t>::iterator iter = streams.begin(); iter != streams.end(); ++iter) {
+ ok &= checkCuda(cudaStreamSynchronize(*iter), "par_fp");
+ cudaStreamDestroy(*iter);
+ }
cudaFreeArray(D_dataArray);
-
- return true;
+ return ok;
}
bool FP_simple(float* D_volumeData, unsigned int volumePitch,
diff --git a/cuda/2d/sart.cu b/cuda/2d/sart.cu
index 29f5b43..89d58c2 100644
--- a/cuda/2d/sart.cu
+++ b/cuda/2d/sart.cu
@@ -54,7 +54,7 @@ void MUL_SART(float* pfOut, const float* pfIn, unsigned int pitch, unsigned int
devMUL_SART<<<gridSize, blockSize>>>(pfOut, pfIn, pitch, width);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), "MUL_SART");
}
diff --git a/cuda/2d/util.cu b/cuda/2d/util.cu
index 2a47472..ac360f0 100644
--- a/cuda/2d/util.cu
+++ b/cuda/2d/util.cu
@@ -40,12 +40,8 @@ bool copyVolumeToDevice(const float* in_data, unsigned int in_pitch,
{
size_t width = dims.iVolWidth;
size_t height = dims.iVolHeight;
- // TODO: memory order
- cudaError_t err;
- err = cudaMemcpy2D(outD_data, sizeof(float)*out_pitch, in_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyHostToDevice);
- ASTRA_CUDA_ASSERT(err);
- assert(err == cudaSuccess);
- return true;
+
+ return checkCuda(cudaMemcpy2D(outD_data, sizeof(float)*out_pitch, in_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyHostToDevice), "copyVolumeToDevice");
}
bool copyVolumeFromDevice(float* out_data, unsigned int out_pitch,
@@ -54,10 +50,8 @@ bool copyVolumeFromDevice(float* out_data, unsigned int out_pitch,
{
size_t width = dims.iVolWidth;
size_t height = dims.iVolHeight;
- // TODO: memory order
- cudaError_t err = cudaMemcpy2D(out_data, sizeof(float)*out_pitch, inD_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyDeviceToHost);
- ASTRA_CUDA_ASSERT(err);
- return true;
+
+ return checkCuda(cudaMemcpy2D(out_data, sizeof(float)*out_pitch, inD_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyDeviceToHost), "copyVolumeFromDevice");
}
@@ -67,10 +61,8 @@ bool copySinogramFromDevice(float* out_data, unsigned int out_pitch,
{
size_t width = dims.iProjDets;
size_t height = dims.iProjAngles;
- // TODO: memory order
- cudaError_t err = cudaMemcpy2D(out_data, sizeof(float)*out_pitch, inD_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyDeviceToHost);
- ASTRA_CUDA_ASSERT(err);
- return true;
+
+ return checkCuda(cudaMemcpy2D(out_data, sizeof(float)*out_pitch, inD_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyDeviceToHost), "copySinogramFromDevice");
}
bool copySinogramToDevice(const float* in_data, unsigned int in_pitch,
@@ -79,20 +71,15 @@ bool copySinogramToDevice(const float* in_data, unsigned int in_pitch,
{
size_t width = dims.iProjDets;
size_t height = dims.iProjAngles;
- // TODO: memory order
- cudaError_t err;
- err = cudaMemcpy2D(outD_data, sizeof(float)*out_pitch, in_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyHostToDevice);
- ASTRA_CUDA_ASSERT(err);
- return true;
+
+ return checkCuda(cudaMemcpy2D(outD_data, sizeof(float)*out_pitch, in_data, sizeof(float)*in_pitch, sizeof(float)*width, height, cudaMemcpyHostToDevice), "copySinogramToDevice");
}
bool allocateVolume(float*& ptr, unsigned int width, unsigned int height, unsigned int& pitch)
{
size_t p;
- cudaError_t ret = cudaMallocPitch((void**)&ptr, &p, sizeof(float)*width, height);
- if (ret != cudaSuccess) {
- reportCudaError(ret);
+ if (!checkCuda(cudaMallocPitch((void**)&ptr, &p, sizeof(float)*width, height), "allocateVolume")) {
ASTRA_ERROR("Failed to allocate %dx%d GPU buffer", width, height);
return false;
}
@@ -104,11 +91,9 @@ bool allocateVolume(float*& ptr, unsigned int width, unsigned int height, unsign
return true;
}
-void zeroVolume(float* data, unsigned int pitch, unsigned int width, unsigned int height)
+bool zeroVolume(float* data, unsigned int pitch, unsigned int width, unsigned int height)
{
- cudaError_t err;
- err = cudaMemset2D(data, sizeof(float)*pitch, 0, sizeof(float)*width, height);
- ASTRA_CUDA_ASSERT(err);
+ return checkCuda(cudaMemset2D(data, sizeof(float)*pitch, 0, sizeof(float)*width, height), "zeroVolume");
}
bool allocateVolumeData(float*& D_ptr, unsigned int& pitch, const SDimensions& dims)
@@ -121,14 +106,14 @@ bool allocateProjectionData(float*& D_ptr, unsigned int& pitch, const SDimension
return allocateVolume(D_ptr, dims.iProjDets, dims.iProjAngles, pitch);
}
-void zeroVolumeData(float* D_ptr, unsigned int pitch, const SDimensions& dims)
+bool zeroVolumeData(float* D_ptr, unsigned int pitch, const SDimensions& dims)
{
- zeroVolume(D_ptr, pitch, dims.iVolWidth, dims.iVolHeight);
+ return zeroVolume(D_ptr, pitch, dims.iVolWidth, dims.iVolHeight);
}
-void zeroProjectionData(float* D_ptr, unsigned int pitch, const SDimensions& dims)
+bool zeroProjectionData(float* D_ptr, unsigned int pitch, const SDimensions& dims)
{
- zeroVolume(D_ptr, pitch, dims.iProjDets, dims.iProjAngles);
+ return zeroVolume(D_ptr, pitch, dims.iProjDets, dims.iProjAngles);
}
void duplicateVolumeData(float* D_dst, float* D_src, unsigned int pitch, const SDimensions& dims)
@@ -231,7 +216,7 @@ float dotProduct2D(float* D_data, unsigned int pitch,
// Step 1: reduce 2D from image to a single vector, taking sum of squares
reduce2D<<< dimGrid2, dimBlock2, shared_mem2>>>(D_data, D_buf, pitch, width, height);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), "dotProduct2D reduce2D");
// Step 2: reduce 1D: add up elements in vector
if (bx * by > 512)
@@ -248,31 +233,21 @@ float dotProduct2D(float* D_data, unsigned int pitch,
float x;
cudaMemcpy(&x, D_res, 4, cudaMemcpyDeviceToHost);
- cudaTextForceKernelsCompletion();
+ checkCuda(cudaThreadSynchronize(), "dotProduct2D");
cudaFree(D_buf);
return x;
}
-
-bool cudaTextForceKernelsCompletion()
+bool checkCuda(cudaError_t err, const char *msg)
{
- cudaError_t returnedCudaError = cudaThreadSynchronize();
-
- if(returnedCudaError != cudaSuccess) {
- ASTRA_ERROR("Failed to force completion of cuda kernels: %d: %s.", returnedCudaError, cudaGetErrorString(returnedCudaError));
+ if (err != cudaSuccess) {
+ ASTRA_ERROR("%s: CUDA error %d: %s.", msg, err, cudaGetErrorString(err));
return false;
+ } else {
+ return true;
}
-
- return true;
}
-void reportCudaError(cudaError_t err)
-{
- if(err != cudaSuccess)
- ASTRA_ERROR("CUDA error %d: %s.", err, cudaGetErrorString(err));
-}
-
-
}
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-05 14:04:21 +0000