Repository navigation

Polar splines on the CuPy backend - #720

Merged

max-models merged 106 commits into

develfrom

polar-splines-cupy

Oct 13, 2026

Merged

Polar splines on the CuPy backend#720
max-models merged 106 commits into
develfrom
polar-splines-cupy

Commits on Oct 7, 2026
- CUDA version of the linear_vlasov_ampere accumulation
  
  max-models
  and
  claude
  committed
  1ad4cd6
- Fuse the pusher preamble into one CUDA kernel
  max-models
  committed
  7de7285
- cupy array fixes
  max-models
  committed
  48c446b
- Compile the CUDA kernels in the setup stage (Simulation.compile_cuda_kernels)
  
  max-models
  and
  claude
  committed
  0352dd2
- Marker exchange between ranks on the device (CUDA-aware MPI)
  
  max-models
  and
  claude
  committed
  8d1c824
- CUDA version of the vlasov_maxwell accumulation
  
  max-models
  and
  claude
  committed
  e693095
- MHD equilibria on the CuPy backend
  
  max-models
  and
  claude
  committed
  14510d2
- Hand-written CUDA vs code generation: port bstar_parallel_3form both ways (decision for #690)
  
  max-models
  and
  claude
  committed
  b51e11f
- Polar splines on the CuPy backend
  
  max-models
  and
  claude
  committed
  c55990f
- Mark every column of the test hole so accumulation kernels skip it
  
  max-models
  and
  claude
  committed
  1979ea9
- Merge cuda-bugfixes (#709) into cuda-fused-pusher-preamble
  
  max-models
  and
  claude
  committed
  1b5ddb1
- Merge cuda-fused-pusher-preamble (#708) into compile-cuda-kernels-setup
  
  max-models
  and
  claude
  committed
  cff3b7f
- Merge compile-cuda-kernels-setup (#711) into cuda-aware-marker-exchange
  
  max-models
  and
  claude
  committed
  8d3ade3
- Merge cuda-aware-marker-exchange (#712) into cuda-linear-vlasov-ampere
  
  max-models
  and
  claude
  committed
  c414599
- Merge cuda-linear-vlasov-ampere (#705) into cuda-vlasov-maxwell
  
  max-models
  and
  claude
  committed
  d7d3f2a
- Merge cuda-vlasov-maxwell (#713) into mhd-equilibria-cupy
  
  max-models
  and
  claude
  committed
  9c368bb
- Merge mhd-equilibria-cupy (#714) into cuda-codegen-spike
  
  max-models
  and
  claude
  committed
  88853ab
- Merge cuda-codegen-spike (#718) into polar-splines-cupy
  
  max-models
  and
  claude
  committed
  2a58677
Commits on Oct 8, 2026
- formatting
  max-models
  committed
  256d02d
- Marker exchange between ranks on the device (CUDA-aware MPI)
  
  max-models
  and
  claude
  authored
  f2e8676
- CUDA version of the linear_vlasov_ampere accumulation
  
  max-models
  and
  claude
  committed
  1636822
- Mark every column of the test hole so accumulation kernels skip it
  
  max-models
  and
  claude
  committed
  9da31a6
- Merge cuda-linear-vlasov-ampere (#705) into cuda-vlasov-maxwell
  max-models
  committed
  144e362
- Merge cuda-vlasov-maxwell (#713) into mhd-equilibria-cupy
  max-models
  committed
  2e92d25
- Merge mhd-equilibria-cupy (#714) into cuda-codegen-spike
  max-models
  committed
  b319210
- Merge cuda-codegen-spike (#718) into polar-splines-cupy
  max-models
  committed
  02cde3d
- Merge branch 'devel' into cuda-bugfixes
  max-models
  committed
  e27878e
- Merge branch 'cuda-bugfixes' into cuda-fused-pusher-preamble
  max-models
  committed
  047c20b
- Merge branch 'cuda-fused-pusher-preamble' into compile-cuda-kernels-setup
  max-models
  committed
  5226e01
- Merge branch 'compile-cuda-kernels-setup' into cuda-aware-marker-exchange
  max-models
  committed
  fb49941
- Merge branch 'cuda-linear-vlasov-ampere' into cuda-aware-marker-exchange
  max-models
  committed
  18ac9cc
- Merge branch 'cuda-linear-vlasov-ampere' into cuda-vlasov-maxwell
  max-models
  committed
  88386b7
- Merge branch 'cuda-vlasov-maxwell' into mhd-equilibria-cupy
  max-models
  committed
  2e30c40
- Merge branch 'mhd-equilibria-cupy' into cuda-codegen-spike
  max-models
  committed
  024461e
- Merge branch 'cuda-codegen-spike' into polar-splines-cupy
  max-models
  committed
  1848f6d
- GVEC: marker evaluation, cunumpy 0.6.2 host helpers, robust fake-CuPy child tests
  
  max-models
  and
  claude
  committed
  b11ee18
- Merge mhd-equilibria-cupy (#714) into cuda-codegen-spike
  max-models
  committed
  b682bc8
- Merge cuda-codegen-spike (#718) into polar-splines-cupy
  max-models
  committed
  0e6850e
Commits on Oct 9, 2026
- Remove the numba code-generation prototype (#690 decided: hand-written CUDA only)
  
  max-models
  and
  claude
  committed
  7d18d8c
- Merge cuda-codegen-spike into polar-splines-cupy
  max-models
  committed
  4bd43af
- Merge remote-tracking branch 'origin/devel' into cuda-bugfixes
  max-models
  committed
  1eb47d3
- Merge cuda-bugfixes into cuda-fused-pusher-preamble
  max-models
  committed
  3370372
- Rename first_init_idx to first_pusher_idx
  
  max-models
  and
  claude
  committed
  0e74e15
- Merge parent into compile-cuda-kernels-setup
  max-models
  committed
  f4be957
- Merge parent into cuda-linear-vlasov-ampere
  max-models
  committed
  7ed75a6
- Merge parent into cuda-aware-marker-exchange
  max-models
  committed
  72b33d3
- Rename first_init_idx to first_pusher_idx in the code new on this branch
  
  max-models
  and
  claude
  committed
  7e92296
- Merge parent into mhd-equilibria-cupy
  max-models
  committed
  d269195
- Merge parent into cuda-vlasov-maxwell
  max-models
  committed
  12bf231
- Skip the new GVEC tests by default (with_gvec=False), like the other GVEC tests
  
  max-models
  and
  claude
  committed
  8223abf
- Merge parent into cuda-codegen-spike
  max-models
  committed
  49cb3c4
- Rename first_init_idx to first_pusher_idx in the code new on this branch
  
  max-models
  and
  claude
  committed
  703a947
- Merge parent into polar-splines-cupy
  max-models
  committed
  ce5affe
- Use xp.mpi.exchange
  max-models
  committed
  05bc8d9
- Merge branch 'cuda-aware-marker-exchange' into cuda-linear-vlasov-ampere
  max-models
  committed
  b5a285a
- Merge branch 'cuda-linear-vlasov-ampere' into cuda-vlasov-maxwell
  max-models
  committed
  b5ab891
- Merge branch 'cuda-vlasov-maxwell' into mhd-equilibria-cupy
  max-models
  committed
  39e0e8d
- Merge branch 'mhd-equilibria-cupy' into cuda-codegen-spike
  max-models
  committed
  a05dcf5
- Merge branch 'cuda-codegen-spike' into polar-splines-cupy
  max-models
  committed
  06d7833
- Merge devel into cuda-bugfixes
  max-models
  committed
  028eed6
Commits on Oct 10, 2026
- Fuse the pusher preamble into one CUDA kernel
  max-models
  authored
  a6e1b19
- Rename first_init_idx to first_pusher_idx
  
  max-models
  and
  claude
  authored
  c1e299b
- Compile the CUDA kernels in the setup stage (Simulation.compile_cuda_kernels)
  
  max-models
  and
  claude
  authored
  c3ea640
- formatting
  max-models
  authored
  958c013
- Marker exchange between ranks on the device (CUDA-aware MPI)
  
  max-models
  and
  claude
  authored
  08665fe
- Use xp.mpi.exchange
  max-models
  authored
  d4abdf6
- CUDA version of the linear_vlasov_ampere accumulation
  
  max-models
  and
  claude
  authored
  22ed9e0
- Mark every column of the test hole so accumulation kernels skip it
  
  max-models
  and
  claude
  authored
  6dd36db
- Rename first_init_idx to first_pusher_idx in the code new on this branch
  
  max-models
  and
  claude
  authored
  52bc2b9
- Merge cuda-linear-vlasov-ampere into cuda-vlasov-maxwell
  
  max-models
  and
  claude
  committed
  8612c67
- Merge cuda-vlasov-maxwell into mhd-equilibria-cupy
  
  max-models
  and
  claude
  committed
  dcedf86
- Merge mhd-equilibria-cupy into cuda-codegen-spike
  
  max-models
  and
  claude
  committed
  218885a
- Merge cuda-codegen-spike into polar-splines-cupy
  
  max-models
  and
  claude
  committed
  faec1b0
- Evaluate the spline-based equilibria on the device; batched field-line tracing
  
  max-models
  and
  claude
  committed
  1cad34a
- Merge mhd-equilibria-cupy into cuda-codegen-spike
  max-models
  committed
  f5079de
- Merge cuda-codegen-spike into polar-splines-cupy
  max-models
  committed
  440827f
- Pin cunumpy to the commit with copyable fake cupyx splines (cunumpy#97)
  
  max-models
  and
  claude
  committed
  70557fb
- Merge mhd-equilibria-cupy into cuda-codegen-spike
  max-models
  committed
  7818b8f
- Merge cuda-codegen-spike into polar-splines-cupy
  max-models
  committed
  85b6a89
- Pin cunumpy to devel with copyable fake cupyx splines (cunumpy#97 merged)
  
  max-models
  and
  claude
  committed
  822af64
- Merge mhd-equilibria-cupy into cuda-codegen-spike
  max-models
  committed
  ffe1485
- Merge cuda-codegen-spike into polar-splines-cupy
  max-models
  committed
  13f8b87
- Pin cunumpy to devel (f996393: copyable fake cupyx splines, faster type checking)
  
  max-models
  and
  claude
  committed
  dce63a9
- Merge mhd-equilibria-cupy into cuda-codegen-spike
  max-models
  committed
  8ebb29d
- Merge cuda-codegen-spike into polar-splines-cupy
  max-models
  committed
  645b8f8
Commits on Oct 12, 2026
- cupy array fixes
  max-models
  authored
  8fcb59f
- Fuse the pusher preamble into one CUDA kernel
  max-models
  authored
  8d8603a
- Rename first_init_idx to first_pusher_idx
  
  max-models
  and
  claude
  authored
  30d81a1
- Compile the CUDA kernels in the setup stage (Simulation.compile_cuda_kernels)
  
  max-models
  and
  claude
  authored
  225bcae
- formatting
  max-models
  authored
  d412fc2
- Marker exchange between ranks on the device (CUDA-aware MPI)
  
  max-models
  and
  claude
  authored
  77a25b5
- Use xp.mpi.exchange
  max-models
  authored
  0ea9085
- CUDA version of the linear_vlasov_ampere accumulation
  
  max-models
  and
  claude
  authored
  569904f
- Mark every column of the test hole so accumulation kernels skip it
  
  max-models
  and
  claude
  authored
  0d15e06
- Rename first_init_idx to first_pusher_idx in the code new on this branch
  
  max-models
  and
  claude
  authored
  0b6fc99
- Merge branch 'cuda-linear-vlasov-ampere' into cuda-vlasov-maxwell
  max-models
  committed
  e46b01e
- Merge branch 'cuda-vlasov-maxwell' into mhd-equilibria-cupy
  max-models
  committed
  6ca659b
- Merge branch 'mhd-equilibria-cupy' into cuda-codegen-spike
  max-models
  committed
  707b48d
- Merge branch 'cuda-codegen-spike' into polar-splines-cupy
  max-models
  committed
  4e9f2a2
- Formatting
  max-models
  committed
  85ad5e0
- Fix unit test
  max-models
  committed
  5825e1f
- Merge branch 'devel' into polar-splines-cupy
  max-models
  committed
  46d7451
- ruff check --fix
  max-models
  committed
  1248722
- Complete polar spline GPU support after the Kronecker refactor
  max-models
  committed
  71818ca
- Update strategy
  max-models
  committed
  72954b1
- update docs
  max-models
  committed
  b80b20f

Commits

Commits on Oct 7, 2026

Commits on Oct 8, 2026

Commits on Oct 9, 2026

Commits on Oct 10, 2026

Commits on Oct 12, 2026