sunwen/URDepends
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: refactor precalculateAverageSpeed functions, remove redudent args

Browse Source
This commit is contained in:
kradchen
2024-11-29 15:28:09 +08:00
parent b599153837
commit c83c161bc6
6 changed files with 25 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
SAFT_TOFI/src/kernel/precalculateSpeedOfSoundKernel.cu
View File

@@ -45,21 +45,15 @@ texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToRecei
texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth1_preprocess; texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth1_preprocess;
texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth2_preprocess; texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth2_preprocess;
__global__ void precalculateAverageSpeedOfSoundKernel(cudaArray *deviceSosAttFieldCuArray, ///< CuArray fuer SosAttFieldTextur __global__ void precalculateAverageSpeedOfSoundKernel(int firstZLayer, ///< First z-layer in the speed of sound grid the
int firstZLayer, ///< First z-layer in the speed of sound grid the ///< pre-calculation is performed for.
///< pre-calculation is performed for. int sosZLayerCount, ///< Number of z-layers in the speed of sound grid the
int sosZLayerCount, ///< Number of z-layers in the speed of sound grid the ///< pre-calculation is performed for.
///< pre-calculation is performed for. int geometry, ///< emitters=0 or receivers=1.
int geometry, ///< emitters=0 or receivers=1. int geometryElementCount, ///< Number of elements in the geometry array.
int geometryElementCount, ///< Number of elements in the geometry array. int maxSoSReceiverArrayForTexture, ///< max amount of elements in the
int maxSoSReceiverArrayForTexture, ///< max amount of elements in the ///< receiver CUDA array.
///< receiver CUDA array. float3 sosOffset, float3 regionOfInterestOffset, float IMAGE_RESOLUTION, float SOS_RESOLUTION, float debugMode, float debugModeParameter)
float *deviceVoxelCountOutputFloat, ///< fuer Count im Floatformat gedacht
///< fuer Texturmemory.
float *speedOfSoundSumOutput, ///< fuer SoS im Floatformat gedacht fuer
///< Texturmemory.
int3 SOSGrid_XYZ, float3 sosOffset, float3 regionOfInterestOffset, float IMAGE_RESOLUTION, float SOS_RESOLUTION, float debugMode,
float debugModeParameter)
{ {
dim3 SosVoxel(threadIdx.x, // SoS-Voxel X ? Threads fangen bei 0 an dim3 SosVoxel(threadIdx.x, // SoS-Voxel X ? Threads fangen bei 0 an
blockIdx.x, // SoS-Voxel Y blockIdx.x, // SoS-Voxel Y
@@ -123,12 +117,10 @@ __global__ void precalculateAverageSpeedOfSoundKernel(cudaArray *deviceSosAttFie
// Bestimmen der SoS-Koordinaten fuer die Sender/Empfuenger-Koordinaten // Bestimmen der SoS-Koordinaten fuer die Sender/Empfuenger-Koordinaten
determineSpeedOfSoundFieldVoxelFloat(currentGeometry, SosGeometryVoxelFloat, sosOffset, determineSpeedOfSoundFieldVoxelFloat(currentGeometry, SosGeometryVoxelFloat, sosOffset,
SOS_RESOLUTION_FACTOR); // SoSVoxel von E/R bestimmen // SOS_RESOLUTION_FACTOR); // SoSVoxel von E/R bestimmen //
// currentGeometry --> SosGeometryVoxel Float // currentGeometry --> SosGeometryVoxel Float
performRayTracedSpeedAdditionTexture(voxelCount, averageSpeed, totalAttenuation, SosGeometryVoxelFloat, SosVoxel, deviceSosAttFieldCuArray, SOSGrid_XYZ, sosOffset, SOS_RESOLUTION, performRayTracedSpeedAdditionTexture(voxelCount, averageSpeed, totalAttenuation, SosGeometryVoxelFloat, SosVoxel, sosOffset); // SosGeometryVoxelFloat im Floatformat,
IMAGE_RESOLUTION, regionOfInterestOffset, // SoSVoxel als Integer
geometry); // SosGeometryVoxelFloat im Floatformat, SoSVoxel als
// Integer
if (geometry == 0) // Emitter if (geometry == 0) // Emitter
{ {

7
SAFT_TOFI/src/kernel/precalculateSpeedOfSoundKernel.cuh
View File

@@ -30,15 +30,12 @@ extern texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxel
extern texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth1_preprocess; extern texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth1_preprocess;
extern texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth2_preprocess; extern texture<float4, cudaTextureType3D, cudaReadModeElementType> texTableVoxelToReceiverPathSosBoth2_preprocess;
__global__ void precalculateAverageSpeedOfSoundKernel(cudaArray *deviceSosAttFieldCuArray, __global__ void precalculateAverageSpeedOfSoundKernel(int firstZLayer,
int firstZLayer,
int sosZLayerCount, int sosZLayerCount,
int geometry, int geometry,
int geometryElementCount, int geometryElementCount,
int maxSoSReceiverArrayForTexture, int maxSoSReceiverArrayForTexture,
float *deviceVoxelCountOutputFloat, float3 sosOffset, float3 regionOfInterestOffset, float IMAGE_RESOLUTION, float SOS_RESOLUTION, float debugMode,
float *speedOfSoundSumOutput,
int3 SOSGrid_XYZ, float3 sosOffset, float3 regionOfInterestOffset, float IMAGE_RESOLUTION, float SOS_RESOLUTION, float debugMode,
float debugModeParameter); float debugModeParameter);
__global__ void precalculateAscanIndex_usePathsKernel(int ascanIndexBatchOffset, __global__ void precalculateAscanIndex_usePathsKernel(int ascanIndexBatchOffset,

16
SAFT_TOFI/src/kernel/rayTracing.cuh
View File

@@ -45,9 +45,7 @@ __device__ __forceinline__ void processRayTracedVoxelTextureSosAtt(
float const *currentVoxelFloat, ///< SOS-coordinates of current Voxel for Bresenham in float[3]. float const *currentVoxelFloat, ///< SOS-coordinates of current Voxel for Bresenham in float[3].
float &voxelCount, ///< This argument is written to. Number of voxels in the path so far. // TODO: not necessary here to sum up because it is already known from Bresenham length float &voxelCount, ///< This argument is written to. Number of voxels in the path so far. // TODO: not necessary here to sum up because it is already known from Bresenham length
float &totalSpeed, ///< This argument is written to. Sum Up total SOS float &totalSpeed, ///< This argument is written to. Sum Up total SOS
float &totalAttenuation, ///< This argument is written to. Sum Up total Attenuation float &totalAttenuation ///< This argument is written to. Sum Up total Attenuation
cudaArray *deviceSosAttFieldCuArray, ///< Pointer to cudaArray for SoSATTFieldData // TODO: not necessary if access is used with texture memory
int3 const &SOSGrid_XYZ ///< Size of SOS-Grid in XYZ // TODO: only necessary for boundary-check.
) )
{ {
@@ -59,19 +57,11 @@ __device__ __forceinline__ void processRayTracedVoxelTextureSosAtt(
__device__ __forceinline__ void performRayTracedSpeedAdditionTexture(float &voxelCount, ///< This argument is written to. Number of voxels within the path traced. __device__ __forceinline__ void performRayTracedSpeedAdditionTexture(float &voxelCount, ///< This argument is written to. Number of voxels within the path traced.
float &totalSpeed, ///< This argument is written to. Sum of the speed of sound samples in the path traced. float &totalSpeed, ///< This argument is written to. Sum of the speed of sound samples in the path traced.
float &totalAttenuation, float &totalAttenuation,
float3 const &point1f, ///< Vector array describing the Voxelcoordinates of emitters or receivers. float3 const &point1f, ///< Vector array describing the Voxelcoordinates of emitters or receivers.
dim3 const &point2, ///< Vector array describing the Voxelcoordinates of Voxels. dim3 const &point2, ///< Vector array describing the Voxelcoordinates of Voxels.
float3 &sosOffset)
cudaArray *deviceSpeedOfSoundFieldCuArray, ///< CuArray fuer SOSFieldTextur
int3 const &SOSGrid_XYZ, ///< Size of SOS-Grid in XYZ
float3 &sosOffset, float &SOS_RESOLUTION, float &IMAGE_RESOLUTION, float3 regionOfInterestOffset,
int geometry ///< emitters=0 or receivers=1.
)
{ {
voxelCount = 0.0f; voxelCount = 0.0f;
@@ -156,7 +146,7 @@ __device__ __forceinline__ void performRayTracedSpeedAdditionTexture(float &voxe
int j = 0; int j = 0;
for (j = fastDirectionSteps; j > 0; j--) //(Alle Punkte innerhalb der Schleife berechnen) for (j = fastDirectionSteps; j > 0; j--) //(Alle Punkte innerhalb der Schleife berechnen)
{ {
processRayTracedVoxelTextureSosAtt(pathPoint, voxelCount, totalSpeed, totalAttenuation, deviceSpeedOfSoundFieldCuArray, SOSGrid_XYZ); processRayTracedVoxelTextureSosAtt(pathPoint, voxelCount, totalSpeed, totalAttenuation);
pathPoint[greatestDistanceDim] = pathPoint[greatestDistanceDim] + m_XYZ[greatestDistanceDim]; pathPoint[greatestDistanceDim] = pathPoint[greatestDistanceDim] + m_XYZ[greatestDistanceDim];
pathPoint[slowDim1] = pathPoint[slowDim1] + m_XYZ[slowDim1]; pathPoint[slowDim1] = pathPoint[slowDim1] + m_XYZ[slowDim1];
pathPoint[slowDim2] = pathPoint[slowDim2] + m_XYZ[slowDim2]; pathPoint[slowDim2] = pathPoint[slowDim2] + m_XYZ[slowDim2];

14
SAFT_TOFI/src/kernel/saftPrivate.cu
View File

@@ -228,8 +228,7 @@ void SAFTHandler::precalculateAscanIndex_usePaths(int ascanIndex_i, int aScanWin
CUDA_CHECK(cudaUnbindTexture(&texTableVoxelToReceiverPathSosBoth2_preprocess)); CUDA_CHECK(cudaUnbindTexture(&texTableVoxelToReceiverPathSosBoth2_preprocess));
} }
void SAFTHandler::precalculateAverageSpeedOfSound(int firstZLayer, int sosZLayerCount, int deviceListGeometry, int geometryElementCount, float *deviceVoxelCountOutputFloat, void SAFTHandler::precalculateAverageSpeedOfSound(int firstZLayer, int sosZLayerCount, int deviceListGeometry, int geometryElementCount)
float *deviceSpeedOfSoundSumOutput)
{ {
dim3 threadsPerBlock(SOSGrid_XYZ.x, 1, dim3 threadsPerBlock(SOSGrid_XYZ.x, 1,
1); // max. 512 oder 1024 Threads werden vorgegeben und 1); // max. 512 oder 1024 Threads werden vorgegeben und
@@ -284,14 +283,9 @@ void SAFTHandler::precalculateAverageSpeedOfSound(int firstZLayer, int sosZLayer
} }
} }
precalculateAverageSpeedOfSoundKernel<<<blocksPerGrid, threadsPerBlock>>>(deviceSosAttFieldCuArray, firstZLayer, sosZLayerCount, deviceListGeometry, geometryElementCount, precalculateAverageSpeedOfSoundKernel<<<blocksPerGrid, threadsPerBlock>>>(firstZLayer, sosZLayerCount, deviceListGeometry, geometryElementCount,
maxSoSReceiverArrayForTexture, // maximale Anzahl an Receivern in einem maxSoSReceiverArrayForTexture,
// CUDA Array sosOffset, regionOfInterestOffset, IMAGE_RESOLUTION, SOS_RESOLUTION, debugMode, debugModeParameter);
// deviceVoxelCountOutput,
deviceVoxelCountOutputFloat, deviceSpeedOfSoundSumOutput,
// regionOfInterestOffset,
SOSGrid_XYZ, sosOffset, regionOfInterestOffset, IMAGE_RESOLUTION, SOS_RESOLUTION, debugMode, debugModeParameter);
CUDA_CHECK(cudaGetLastError()); CUDA_CHECK(cudaGetLastError());
CUDA_CHECK(cudaUnbindTexture(&texRefSosAttField)); CUDA_CHECK(cudaUnbindTexture(&texRefSosAttField));

5
SAFT_TOFI/src/processAScans.cpp
View File

@@ -130,12 +130,11 @@ void SAFTHandler::performCoreReconstruction()
// SoS-Pfade für Z-Layer vorberechnen ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ // SoS-Pfade für Z-Layer vorberechnen ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
// Emitter ----------------------------------------------------------------------- Emitter // Emitter ----------------------------------------------------------------------- Emitter
precalculateAverageSpeedOfSound(currentSpeedOfSoundZLayer, maxFeasibleSosZLayerCount, 0, emitter_list_Size, deviceTableVoxelToEmitterPathCountFloat, deviceTableVoxelToEmitterPathSosSum); precalculateAverageSpeedOfSound(currentSpeedOfSoundZLayer, maxFeasibleSosZLayerCount, 0, emitter_list_Size);
// Receiver ----------------------------------------------------------------------- Receiver // Receiver ----------------------------------------------------------------------- Receiver
precalculateAverageSpeedOfSound(currentSpeedOfSoundZLayer, maxFeasibleSosZLayerCount, 1, receiver_list_Size, deviceTableVoxelToReceiverPathCountFloat, precalculateAverageSpeedOfSound(currentSpeedOfSoundZLayer, maxFeasibleSosZLayerCount, 1, receiver_list_Size);
deviceTableVoxelToReceiverPathSosSum);
} }
// Go overall A-Scanblocks ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ // Go overall A-Scanblocks ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
// PrecalAndPerformSAFTAllAscans - Berechne Performanz fuer Durchlauf von allen Ascans ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ // PrecalAndPerformSAFTAllAscans - Berechne Performanz fuer Durchlauf von allen Ascans ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

2
SAFT_TOFI/src/saft.hpp
View File

@@ -364,7 +364,7 @@ class SAFTHandler
// Pre-calculation kernels // Pre-calculation kernels
//------------------------------------------------------------------------ //------------------------------------------------------------------------
void precalculateAverageSpeedOfSound(int firstZLayer, int sosZLayerCount, int deviceGeometry, int geometryElementCount, float *deviceVoxelCountOutputFloat, float *deviceSpeedOfSoundSumOutput); void precalculateAverageSpeedOfSound(int firstZLayer, int sosZLayerCount, int deviceGeometry, int geometryElementCount);
void precalculateAscanIndex(int currentSpeedOfSoundZLayer, ///< First z-layer in the speed of sound grid the pre-calculation is performed for. void precalculateAscanIndex(int currentSpeedOfSoundZLayer, ///< First z-layer in the speed of sound grid the pre-calculation is performed for.
int maxFeasibleSosZLayerCount ///< Number of z-layers in the speed of sound grid the pre-calculation is performed for. int maxFeasibleSosZLayerCount ///< Number of z-layers in the speed of sound grid the pre-calculation is performed for.