summaryrefslogtreecommitdiffstats
path: root/cuda
Commit message (Collapse)AuthorAgeFilesLines
* Half-precision back-/forward-projection for parallel geometryhalfSuren A. Chilingaryan2022-07-263-2/+49
|
* Add transferConstants functions for FP kernels for consistencyWillem Jan Palenstijn2021-12-062-42/+58
|
* Fix potential invalid memory access in par3d_fp with many anglesWillem Jan Palenstijn2021-12-021-2/+2
|
* Handle angular range discretisation betterWillem Jan Palenstijn2021-12-011-2/+5
|
* Adjust Parker weights to angular rangeWillem Jan Palenstijn2021-12-011-5/+16
| | | | | | | | This fixes a global scaling in shortscan reconstruction. Also add some debugging logs for the Parker weighting and a warning if the angular range is too small. Issue #229
* Fix memleak in error handlingWillem Jan Palenstijn2021-11-263-6/+15
|
* De-duplicate 2D texture object creationWillem Jan Palenstijn2021-11-265-134/+86
|
* Replace texref by texobj in par_bpWillem Jan Palenstijn2021-11-261-25/+44
|
* Replace texref by texobj in fan_bpWillem Jan Palenstijn2021-11-261-27/+49
|
* De-duplicate 3D texture object creationWillem Jan Palenstijn2021-11-265-99/+38
|
* Replace texref by texobj in cone_bpWillem Jan Palenstijn2021-11-261-34/+42
|
* Replace texref by texobj in par3d_bpWillem Jan Palenstijn2021-11-261-30/+37
|
* Replace texref by texobj in cone_fpWillem Jan Palenstijn2021-11-261-36/+42
|
* Replace texref by texobj in par3d_fpWillem Jan Palenstijn2021-11-261-35/+44
|
* Replace texref by texobj in fan_fpWillem Jan Palenstijn2021-11-261-25/+31
|
* Replace texref by texobj in par_fpWillem Jan Palenstijn2021-11-261-26/+33
|
* Fix (harmless) duplicationWillem Jan Palenstijn2021-11-171-1/+1
|
* Switch cuda memhandle checks to checkCudaWillem Jan Palenstijn2021-11-171-26/+12
|
* Fix non-padded GPULink memory handling in FP3D kernelsWillem Jan Palenstijn2021-11-162-0/+11
| | | | | | | | | | This would fail silently if the output projection data object was not padded to a multiple of 32 pixels, potentially corrupting the start of projection rows. 3D GPU memory allocated by ASTRA itself is always padded by cudaMalloc3D and therefore not affected. GPULink allows bypassing this, possibly triggering this bug.
* Remove fft.cu custom cuda error handling macrosWillem Jan Palenstijn2021-11-161-42/+21
|
* Remove unnecessary costly syncs in FFTWillem Jan Palenstijn2021-11-161-2/+8
|
* Add missing synchronize before free in FFTWillem Jan Palenstijn2021-11-161-2/+10
|
* Add checkCufft functionWillem Jan Palenstijn2021-11-161-23/+22
|
* Remove cudaTextForceKernelsCompletionWillem Jan Palenstijn2021-11-1614-114/+83
|
* Remove ASTRA_CUDA_ASSERTWillem Jan Palenstijn2021-11-162-80/+33
|
* Remove reportCudaError functionWillem Jan Palenstijn2021-11-162-23/+6
|
* Add new checkCuda function to simplify error handlingWillem Jan Palenstijn2021-11-161-0/+9
|
* Add specialization to BP3D kernels for single slice volumesWillem Jan Palenstijn2021-11-162-13/+24
|
* Update copyright yearsWillem Jan Palenstijn2021-10-2728-56/+56
|
* Fix memory leak in CUDA 2D fan_bpFerdia Sherry2021-01-271-0/+2
|
* Fix supersampling version of cone_bp and add testWillem Jan Palenstijn2021-01-221-4/+4
|
* Fix memory leak in CUDA 2D par_fpFerdia2020-09-111-0/+4
| | | Arrays were dynamically allocated in the function convertAndUploadAngles, but were not deallocated, resulting in a memory leak. At the end of this function I now delete [] these arrays, fixing the memory leak.
* Fix potential geometry constant corruption in par3d_bpWillem Jan Palenstijn2020-02-051-1/+1
|
* Fix memory leaks in 3D BPWillem Jan Palenstijn2020-02-052-0/+6
|
* Add noteWillem Jan Palenstijn2019-09-251-0/+3
|
* Remove unmaintained, out of date 'STANDALONE' cuda codeWillem Jan Palenstijn2019-09-2515-1822/+0
|
* Adjust par3d adjoint scaling, and clean upWillem Jan Palenstijn2019-09-252-88/+49
|
* Fix part of non-cube cone scalingWillem Jan Palenstijn2019-09-252-2/+5
|
* Small clean up of factorsWillem Jan Palenstijn2019-09-253-13/+11
|
* Remove obsolete DensityWeighting optionWillem Jan Palenstijn2019-09-251-2/+2
|
* Improve adjoint matching for fan/cone BP functions, and clean upWillem Jan Palenstijn2019-09-253-223/+251
|
* Fix fan-beam FBP scalingWillem Jan Palenstijn2019-09-252-8/+9
|
* Adjust SART to line integral scalingWillem Jan Palenstijn2019-09-253-9/+10
|
* Clean up outputscale naming confusion in cuda::algoWillem Jan Palenstijn2019-09-253-19/+22
|
* Adjust FBP to line integral scalingWillem Jan Palenstijn2019-09-253-11/+11
|
* Adjust adjoint to line integral scalingWillem Jan Palenstijn2019-09-254-28/+72
|
* Adjust linear/cuda kernels to line integral scalingWillem Jan Palenstijn2019-09-252-7/+6
|
* Replace signal-based abort handling by query-based handlingWillem Jan Palenstijn2019-08-149-38/+6
| | | | | | | | The abort handling is currently only used to process Ctrl-C from Matlab. Since Matlab R2019a, it appears that calling utIsInterruptPending() from a thread other than the main thread will crash. The previous approach of checking utIsInterruptPending() in a thread, and then signalling the running algorithm was therefore broken.
* Add minimal GPU Array interfaceWillem Jan Palenstijn2018-11-013-8/+167
| | | | | This extension (only) allows creating a CUDA 3D array, copying projection data into it, performing a BP from the array, and freeing the array.
* Reorganize more filter size functionsWillem Jan Palenstijn2018-07-172-14/+4
|
generated by cgit v1.2.1 (git 2.18.0) at 2024-05-18 01:46:09 +0000