AzIntEngineGPU: Fixes

This commit is contained in:
2026-04-22 20:07:53 +02:00
parent 7265bd91f2
commit 92a3c5eb30
2 changed files with 8 additions and 20 deletions
+6 -18
View File
@@ -100,13 +100,13 @@ AzIntEngineGPU::AzIntEngineGPU(const AzimuthalIntegration &integration)
cpu_count_reg(azint_count),
gpu_image(npixel) {
cudaDeviceProp prop;
cudaDeviceProp prop{};
cudaGetDeviceProperties(&prop, 0);
threads = 128;
blocks = 4 * prop.multiProcessorCount;
shared_size = prop.sharedMemPerBlock;
shared_needed = azint_bins * sizeof(float) + azint_bins * sizeof(float) + azint_bins * sizeof(uint32_t);
shared_needed = azint_bins * (2 * sizeof(float) + sizeof(uint32_t));
cudaMemcpy(gpu_azint_correction, integration.Corrections().data(), sizeof(float) * npixel,
cudaMemcpyHostToDevice);
@@ -124,25 +124,13 @@ void AzIntEngineGPU::Run(const std::vector<int32_t> &image, AzimuthalIntegration
if (shared_needed < shared_size) {
gpu_azim_shared<<<blocks, threads, shared_needed, stream>>>(
gpu_pixel_to_bin,
gpu_azint_correction,
gpu_image,
gpu_sum,
gpu_sum2,
gpu_count,
npixel,
azint_bins
gpu_pixel_to_bin,gpu_azint_correction,gpu_image, gpu_sum, gpu_sum2,
gpu_count, npixel, azint_bins
);
} else {
gpu_azim<<<blocks, threads, 0, stream>>>(
gpu_pixel_to_bin,
gpu_azint_correction,
gpu_image,
gpu_sum,
gpu_sum2,
gpu_count,
npixel,
azint_bins
gpu_pixel_to_bin,gpu_azint_correction,gpu_image, gpu_sum, gpu_sum2,
gpu_count, npixel, azint_bins
);
}
+2 -2
View File
@@ -10,8 +10,8 @@ class AzIntEngineGPU : public AzIntEngine {
CudaStream stream;
int threads;
int blocks;
int shared_needed;
int shared_size;
size_t shared_needed;
size_t shared_size;
CudaDevicePtr<float> gpu_azint_correction;
CudaDevicePtr<uint16_t> gpu_pixel_to_bin;