Repository navigation

Start on batched SVD support - #567

Open

kshyatt wants to merge 16 commits into

mainfrom

ksh/batched_svd2

Open

Start on batched SVD support#567
kshyatt wants to merge 16 commits into
mainfrom
ksh/batched_svd2

Commits on Oct 16, 2026
- Start on batched SVD support
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  c1a82e6
- Offload a lot to MAK
  kshyatt
  committed
  54ad087
- Update Project.toml
  kshyatt
  committed
  124b293
- Get rid of unneeded bypass
  kshyatt
  committed
  44db5c2
- Force the batched path to actually run and use QRIteration for AMD + isometric
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  8c60919
- Update Project.toml
  kshyatt
  committed
  df50087
- Fix blocksectors handling
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  9199a1f
- Try the select_inds fixer
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  58c17cd
- More CUDA tests and drop sources
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  4e008be
- Diagonal + Jacobi doesn't work
  kshyatt
  committed
  1cd5673
- Another gesvdx painpoint on AMD
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  ce2a4e6
- Make a specialization to avoid Diagonal batching for now
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  8e762ea
- Use MAK main
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  5f0075c
- Respond to comments
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  2d61b0b
Commits on Oct 17, 2026
- Update Project.toml
  kshyatt
  authored
  975ab5a
Commits on Oct 19, 2026
- Work around ROCm bug in batched QRIteration
  Katharine Hyatt
  committed
  d165e58