CUDA pipeline for computing APR - #185

Open

krzysg wants to merge 91 commits into

Commits on Aug 1, 2022

Bspline filters fixed for CUDA pipeline
krzysg
committed
e6aa9c9
Debug messages turned off
krzysg
committed
b563da4
Fixed Inv Bspline in X direction (CUDA pipeline)
krzysg
committed
3db510f

Commits on Aug 2, 2022

Inverse Bspline pipeline for CUDA fixed
krzysg
committed
18fce44

Commits on Aug 3, 2022

Downsample and downsample gradient corrected to match GPU
krzysg
committed
ad5f194

Commits on Aug 9, 2022

GPU pipeline fixes - Full Gradient test is working now
krzysg
committed
557eff3
Merge branch 'develop' into cuda
krzysg
committed
57765a7
Merge branch 'develop' into cuda
krzysg
committed
3da13ba

Commits on Aug 10, 2022

Commits on Nov 14, 2022

Added new test file for LIS CUDA, GPU now handles boundary (without padding), still float number differences between CPU and GPU
krzysg
committed
b050e07

Commits on Jan 31, 2023

Local Intensity Scale (LIS) not works in X-dir as expected. GPU and CPU gives same results.
krzysg
committed
570ab20

Commits on Feb 1, 2023

Local Intensity Scale (LIS) now works in Z-dir as expected. GPU and CPU gives same results.
krzysg
committed
17e5d8e

Commits on Feb 17, 2023

Commits on Feb 24, 2023

LIS in Z-dir redesigned so code is clearer and faster. Also new test added.
krzysg
committed
521d826

Commits on Mar 13, 2023

Local Intensity Scale (LIS) now works in Y-dir as expected. GPU and CPU gives same results.
krzysg
committed
b297adf

Commits on Mar 16, 2023

Whole LIS pipeline is matching exactly CPU implementation + tests updated
krzysg
committed
2cdf3fe
Quick fix of linking error
krzysg
committed
e093c01
maximum error diff. GPU vs CPU for compute gradient set to 0
krzysg
committed
053380d

Commits on Mar 17, 2023

Commits on Mar 20, 2023

Commits on Mar 24, 2023

full pipeline tests fixed
krzysg
committed
6a5db35

Commits on Jul 20, 2023

Changes from old branches added + modified to GenInfo instead of APRAccess
krzysg
committed
4088e9d

Commits on Jul 21, 2023

Added debug printout to GenInfo
krzysg
committed
b8f2504

Commits on Aug 11, 2023

Moved old CUDA tests to new file
krzysg
committed
6400a9a
Moved old CUDA tests to new file
krzysg
committed
4b35b8e

Commits on Oct 30, 2023

Added CUDA_ARCHITECTURES set to OFF (keep current behaviour) to suppress warning
krzysg
committed
1ed5d4f

Commits on Nov 8, 2023

Temporary test updated to print particles using LinearAccess iterator
krzysg
committed
93ac120
Merge branch 'master' into cuda
krzysg
committed
09bf86a

Commits on Nov 10, 2023

Merge branch 'master' into cuda
krzysg
committed
b7ae1cb
Merge branch 'master' into cuda
krzysg
committed
6181da6

Commits on Nov 13, 2023

Merge branch 'master' into cuda
krzysg
committed
ed09686

Commits on Nov 30, 2023

TODO about some problems with edge case
krzysg
committed
70543d2

Commits on Dec 6, 2023

Fixed test where out of range idx was given
krzysg
committed
dd3d448

Commits on Dec 13, 2023

Pulling Scheme tests (and OVPC on CPU) finished.
krzysg
committed
1a112ec

Commits on Dec 14, 2023

Fixes for tests
krzysg
committed
64ca641

Commits on Jan 9, 2024

Fixed OVPC - clamping values of input levels is necessary
krzysg
committed
9f31bfd

Commits on Feb 5, 2024

Updated OVPC (PS) for CUDA - now it gives correct ans same results as PS CPU ver.
krzysg
committed
2707207

Commits on Feb 16, 2024

PullingSchemeCudaTest finished, added init file for LinearAcccess test
krzysg
committed
3cb4529

Commits on Feb 21, 2024

Finished LinearAccess tests (for linear structure only), added draft for LinearAccessCuda
krzysg
committed
027e52a

Commits on Feb 23, 2024

Check also total_number_particles in LinearAccess test
krzysg
committed
e83b952

Commits on Aug 2, 2024

LinearAccessCuda implemented (it is not used yet in CUDA pipeline)
krzysg
committed
2cc5bca
Compiler warnings fixed
krzysg
committed
e1b63d7

Commits on Aug 6, 2024

Commits on Aug 8, 2024

Commits on Aug 9, 2024

Fix for bsplineYdir for very small input images + test for full pipeline updated
krzysg
committed
bb3b3f4

Commits on Aug 14, 2024

Fixed Local Intensity Scale (LIS) for super small inputs
krzysg
committed
a8c4d77

Commits on Aug 19, 2024

ParticleCellTreeCuda is now main stuff for CUDA
krzysg
committed
e6e4327

Commits on Aug 20, 2024

computeOvpcCuda now using 'stream' instead of hardcoded values
krzysg
committed
00aac97
ParticleCellTreeCuda moved and handle now cpu2gpu transfer
krzysg
committed
1fba1bc
LinearAccessCuda is now using ParticleCellTreeCuda
krzysg
committed
3474250

Commits on Aug 21, 2024

OVPC added to GpuTask
krzysg
committed
1d4e549
Full GPU pipeline works1
krzysg
committed
9ff0580
Some debug prints removed
krzysg
committed
c10225d
Test for full pipeline cleaned up
krzysg
committed
6b7a87d
doAll() removed from Gpu pipeline
krzysg
committed
3c601be

Commits on Aug 22, 2024

GPU pipeline now works for APRConverter!
krzysg
committed
d2fd1d0

Commits on Mar 17, 2025

Linear acces now is using correct cuda stream, bspline params are computed in constructor and memory for them is preallocated
krzysg
committed
9604c63

Commits on Mar 19, 2025

added error handling for bspline y-dir
krzysg
committed
9572e10
Removed (most of the) warnings
krzysg
committed
514c03c

Commits on Jun 25, 2025

added error handling for bspline x-dir, other steps temporarily blocked
krzysg
committed
005a4ba
added error handling for bspline z-dir
krzysg
committed
5534860

Commits on Aug 1, 2025

Stream operations on GPU are working now as expected. All tests are fixed. Still APRConverter for streams must be fixed/improved since there is a draft code.
krzysg
committed
1b54cd0

Commits on Aug 5, 2025

Reverting APRCOnverter type to previoius value
krzysg
committed
7f6e2d3
Fixed CUDA-streams sync issues when copying back to CPU
krzysg
committed
b01df31

Commits on Aug 6, 2025

Little bit cleanup of CUDA in APRConverter, Added move assignment operator to VectorData
krzysg
committed
bf4cdb0

Commits on Aug 11, 2025

Fixed move construtor/assignment - pinned memory is now also moved.
krzysg
committed
cd7d594

Commits on Aug 19, 2025

Fixes needed by CUDA 13.0 - now code compiles
krzysg
committed
0c702f3

Commits on Aug 21, 2025

Initial impl. of CUDA multistreams, it takes many images but STILL only one ARP object - use it only for speed for now
krzysg
committed
1f27876

Commits on Aug 29, 2025

Fixes needed by CUDA 13.0 - now code compiles also on CUDA 12.x
krzysg
committed
0ba2f45

Commits on Sep 4, 2025

CUDA version of getMinMax added (for finding bspline offset)
krzysg
committed
0a44736

Commits on Sep 9, 2025

Bspline Offset is now computed on GPU + copy of original image for sampling
krzysg
committed
4093552

Commits on Sep 11, 2025

Added some useful methods to GenInfo (size and dimension)
krzysg
committed
e686349

Commits on Sep 12, 2025

Sampling on GPU added - THIS CODE is to being cleaned up - it is pushed for backup reasons only
krzysg
committed
1db4abf

Commits on Nov 10, 2025

Reverting CUDA-pipeline related changes
krzysg
committed
8415a52
Initial draft of Example_get_multiapr
krzysg
committed
629f806

Commits on Nov 26, 2025

Added Example_get_multiapr for CUDA parallel processing
krzysg
committed
3a11834

Commits on Nov 28, 2025

localScaleTemp is not anymore passed to GpuProcessinTask, CPU memory init in APRConverter for GPU not needed anymore
krzysg
committed
c20fac6

Commits on Dec 9, 2025

Removed unnecessary comments
krzysg
committed
890283c

Commits on Dec 17, 2025

get_multiapr can now repeat processing of input data, skip saving APRs and limit log messages for benchmarking purposes
krzysg
committed
d9527e6

Commits on Dec 19, 2025

Final (yupi :-) ) version of multistream computing on GPU
krzysg
committed
2682562

Commits on Jan 6, 2026

Parallel copying to pinned buffer
krzysg
committed
9592d98

Commits on Jan 20, 2026

Added support for multi-GPU
krzysg
committed
5f62f14