Code

Add GPU support to MRIReco.jl#182

Code

Merged

tknopp merged 66 commits intomasterMagneticResonanceImaging/MRIReco.jl:masterfrom nh/gpuSupportMagneticResonanceImaging/MRIReco.jl:nh/gpuSupportCopy head branch name to clipboardAug 13, 2024

+800-635

Commits

Commits on Apr 17, 2024

• Update RegLS calls

  nHackel

  committed
  72641ee

  Copy full SHA for 72641ee

• Add S kwarg to SparseOp

  nHackel

  committed
  b0af5eb

  Copy full SHA for b0af5eb

• Fix NFFTOp call, unsure if trajectory is passed on correctly

  nHackel

  committed
  ed506cb

  Copy full SHA for ed506cb

• Add :arrayType to param dict

  nHackel

  committed
  5557aaa

  Copy full SHA for 5557aaa

• Init working GPU reconstruction_simple

  nHackel

  committed
  8ecdc66

  Copy full SHA for 8ecdc66

Commits on Apr 18, 2024

• Remove CompositeOp

  nHackel

  committed
  a3a5202

  Copy full SHA for a3a5202

• Fix NFFTOp nodes retrieval from Trajectory

  nHackel

  committed
  bc1cf80

  Copy full SHA for bc1cf80

• Fix SparseOp

  nHackel

  committed
  7f88b81

  Copy full SHA for 7f88b81

• Fix missing type conversion for default Cartesian3D trajectory

  nHackel

  committed
  4cfc217

  Copy full SHA for 4cfc217

• Remove CompOp include

  nHackel

  committed
  fd86c4e

  Copy full SHA for fd86c4e

• Init ExplicitOp for GPU

  nHackel

  committed
  ef307cf

  Copy full SHA for ef307cf

• ctprod for ExplicitOp on GPU

  nHackel

  committed
  7cc8cd1

  Copy full SHA for 7cc8cd1

Commits on Apr 22, 2024

• Add GPU support to SubspaceOp

  nHackel

  committed
  1070052

  Copy full SHA for 1070052

• Add S kwarg to MapSliceOp

  nHackel

  committed
  6da90c9

  Copy full SHA for 6da90c9

• Add GPU support to circulaShutter!

  nHackel

  committed
  4669dc9

  Copy full SHA for 4669dc9

• Migrate DiagOp to LinearOpCollection, "overwrite" products with @Floop loops for dense vectors

  nHackel

  committed
  4ed271c

  Copy full SHA for 4ed271c

• Add S kwarg to EncodingOps

  nHackel

  committed
  da79555

  Copy full SHA for da79555

Commits on Apr 23, 2024

• Chose @floop executor based on arraytype

  nHackel

  committed
  a29fd35

  Copy full SHA for a29fd35

• Add GPU support for SensitivityOp

  nHackel

  committed
  23f9d44

  Copy full SHA for 23f9d44

• Fix missing commas in EncodingOps

  nHackel

  committed
  c399fe0

  Copy full SHA for c399fe0

• Adapt to function name change for DiagOp

  nHackel

  committed
  44a3d08

  Copy full SHA for 44a3d08

Commits on Apr 24, 2024

• Add operatore copy function as kwarg

  nHackel

  committed
  c98b0b3

  Copy full SHA for c98b0b3

• Pass along FFTParams to operators based on array type

  nHackel

  committed
  8daf6c1

  Copy full SHA for 8daf6c1

• Fix GPU FFT params

  nHackel

  committed
  de51024

  Copy full SHA for de51024

• Hide copyOpsFn for encoding params

  nHackel

  committed
  aa4b9a9

  Copy full SHA for aa4b9a9

• Remove debug for-loop limits for multi_coil reco

  nHackel

  committed
  4152422

  Copy full SHA for 4152422

Commits on Apr 25, 2024

• Impore arrayType param handling

  nHackel

  committed
  856f85d

  Copy full SHA for 856f85d

• Remove (broken) default Wavelet sparsityTrafo

  nHackel

  committed
  e362d3d

  Copy full SHA for e362d3d

• Use copyto! for better GPU performance

  nHackel

  committed
  f934502

  Copy full SHA for f934502

Commits on May 22, 2024

• Remove log message

  nHackel

  committed
  a9e39a3

  Copy full SHA for a9e39a3

Commits on Jun 5, 2024

• Init updating FieldmapNFFTOp constructor for GPU

  nHackel

  committed
  b5486bb

  Copy full SHA for b5486bb

• Reduce clutter in ExplicitOp constructor

  nHackel

  committed
  a757463

  Copy full SHA for a757463

• Init updating MRIOperators tests

  nHackel

  committed
  53b4dc4

  Copy full SHA for 53b4dc4

Commits on Jun 6, 2024

• Add GPU support for FieldmapNFFTOp (not very optimized)

  nHackel

  committed
  f214369

  Copy full SHA for f214369

• Add cuda tests for MRIOperators

  nHackel

  committed
  30e0534

  Copy full SHA for 30e0534

• Update fieldmapop copy

  nHackel

  committed
  762bd5a

  Copy full SHA for 762bd5a

• Add arrayTypes to MRIReco tests

  nHackel

  committed
  b0c13ba

  Copy full SHA for b0c13ba

• Adapt to RegLS Kwargs renaming

  nHackel

  committed
  32e109d

  Copy full SHA for 32e109d

Commits on Jun 11, 2024

• Update DirectReco for GPU

  nHackel

  committed
  2164c8f

  Copy full SHA for 2164c8f

• Fix copy for FieldmapNFFTOp

  nHackel

  committed
  ffc9499

  Copy full SHA for ffc9499

• Update arrayType parameter handling (lessen restriction to accept functions)

  nHackel

  committed
  02f0780

  Copy full SHA for 02f0780

• Move L_inv and SenseMaps to GPU

  nHackel

  committed
  7de09ac

  Copy full SHA for 7de09ac

• Init not-error-ing tests (still some failing)

  nHackel

  committed
  838773f

  Copy full SHA for 838773f

• Fix bug in ExplicitOp

  nHackel

  committed
  deabe4b

  Copy full SHA for deabe4b

Commits on Jul 3, 2024

• Add CUDA buildkite CI

  nHackel

  committed
  b9b39cf

  Copy full SHA for b9b39cf

• Add AMD tests (inactive until NFFT runs on AMD)

  nHackel

  committed
  55b813b

  Copy full SHA for 55b813b

Commits on Jul 4, 2024

• Update RegLS and LinearOperatorColl. compat

  nHackel

  committed
  5d54fb1

  Copy full SHA for 5d54fb1

• Readd JLArray tests to MRIOperators

  nHackel

  committed
  01185c0

  Copy full SHA for 01185c0

• Add JLArray test to MRIReco

  nHackel

  committed
  e9e2039

  Copy full SHA for e9e2039

Commits on Jul 9, 2024

• Attempt to run buildkite for MRIOperators

  nHackel

  committed
  348fcba

  Copy full SHA for 348fcba

• Merge branch 'master' into nh/gpuSupport

  nHackel

  committed
  0499df3

  Copy full SHA for 0499df3

• prepare project.toml for releases

  nHackel

  committed
  beea135

  Copy full SHA for beea135

Commits on Jul 31, 2024

• Fix MRIOperators dep in MRISimulation

  nHackel

  committed
  747730f

  Copy full SHA for 747730f

• Try fix CUDA buildkite

  nHackel

  committed
  65c25ab

  Copy full SHA for 65c25ab

• Try to fix CI by dev'ing all packages in one call

  nHackel

  committed
  9c7e769

  Copy full SHA for 9c7e769

• Remove CuNFFT from CUDA test

  nHackel

  committed
  d81ded7

  Copy full SHA for d81ded7

• Update CUDA and ROCM test files for GPU support

  nHackel

  committed
  a06cad3

  Copy full SHA for a06cad3

• Merge branch 'master' into nh/gpuSupport

  nHackel

  authored
  32ff016

  Copy full SHA for 32ff016

• Merge branch 'nh/gpuSupport' of github.com:MagneticResonanceImaging/MRIReco.jl into nh/gpuSupport

  nHackel

  committed
  c845d6b

  Copy full SHA for c845d6b

• Ignore empty plans in FieldMapNFFTOp constructor

  nHackel

  committed
  6d165d4

  Copy full SHA for 6d165d4

• Use Adapt for MRIOperator constructors

  nHackel

  committed
  70bfd37

  Copy full SHA for 70bfd37

• Try fixing MRIOperators buildkite

  nHackel

  committed
  fda6fb7

  Copy full SHA for fda6fb7

Commits on Aug 1, 2024

• Update MRIOperators import in runtests.jl

  nHackel

  committed
  5be9623

  Copy full SHA for 5be9623

• Use MRIOperators test env in buildkite

  nHackel

  committed
  52c8a81

  Copy full SHA for 52c8a81

• Activate AMD MRIOperators buildkite step

  nHackel

  committed
  b79c650

  Copy full SHA for b79c650

• Fix whitespace in pipeline.yml

  nHackel

  committed
  f0e29bc

  Copy full SHA for f0e29bc