Repository navigation

CUDA backend - #219

Draft

rafaelha wants to merge 24 commits into

mainfrom

cuda-statevector-executor

Draft

CUDA backend#219
rafaelha wants to merge 24 commits into
mainfrom
cuda-statevector-executor

Commits on Nov 10, 2026
- feat(statevector): add the host code that GPU backends share
  
  rafaelha
  and
  claude
  committed
  ea4e9e3
- fix(statevector): bound the GPU noise tables, chunk sizes and shot-work model
  
  rafaelha
  and
  claude
  committed
  1c3a3d8
- fix(statevector): draw noise of tiny probabilities from 48-bit skips
  
  rafaelha
  and
  claude
  committed
  44264cb
Commits on Nov 11, 2026
- refactor(statevector): assert the tape dialects' chunk size
  
  rafaelha
  and
  claude
  committed
  0bff456
- refactor(statevector): drop rows.pieces
  
  rafaelha
  and
  claude
  committed
  710ddf2
- refactor(statevector): share the noise split and screen pass rate
  
  rafaelha
  and
  claude
  committed
  df60b74
- refactor(statevector): share the tape kernel's glue and limits
  
  rafaelha
  and
  claude
  committed
  41503fd
- refactor(statevector): share the GPU samplers' fork warning
  
  rafaelha
  and
  claude
  committed
  f25fcfc
- feat(statevector): sample on CUDA GPUs on the shared GPU host code
  
  rafaelha
  and
  claude
  committed
  fd8beac
- fix(statevector): fix CUDA sampler issues found in review
  
  rafaelha
  and
  claude
  committed
  bfa5063
- refactor(cuda): pack whole shot words, without piece tables
  
  rafaelha
  and
  claude
  committed
  d832e66
- refactor(cuda): use the shared GPU helpers
  
  rafaelha
  and
  claude
  committed
  5b354f6
- refactor(cuda): let the CUDA package choose its sampler
  
  rafaelha
  and
  claude
  committed
  bce6606
- refactor(cuda): name the thread counts and the counts words
  
  rafaelha
  and
  claude
  committed
  23841cd
- refactor(cuda): move the kernels' CUDA C++ to cuda/kernels.py
  
  rafaelha
  and
  claude
  committed
  1d3b8ee
- test(cuda): one parametrized comparison with cat5
  
  rafaelha
  and
  claude
  committed
  e211b59
- docs(cuda): shorten the changelog entry and the strategy docstrings
  
  rafaelha
  and
  claude
  committed
  70e55b8
- Merge remote-tracking branch 'origin/main' into simplify/gpu-shared
  rafaelha
  committed
  8a7f4cf
- Merge remote-tracking branch 'origin/feat/gpu-shared' into simplify/cuda
  rafaelha
  committed
  27b9806
Commits on Nov 12, 2026
- Merge origin/main into the CUDA backend
  rafaelha
  committed
  f9d8591
- feat(cuda): sample with the device argument and Metal's fixes
  
  rafaelha
  and
  claude
  committed
  7562138
- docs: document the CUDA backend beside Metal
  
  rafaelha
  and
  claude
  committed
  9371e02
- perf(cuda): keep leading whole bytes a view of the rows; tie check out of line
  
  rafaelha
  and
  claude
  committed
  86bd80b
- Merge origin/main (1.1.0, #296) into the CUDA backend
  
  rafaelha
  and
  claude
  committed
  6f99566

Commits

Commits on Nov 10, 2026

Commits on Nov 11, 2026

Commits on Nov 12, 2026