Repository navigation

Start on batched SVD support - #567

Open

kshyatt wants to merge 12 commits into

mainfrom

ksh/batched_svd2

Commits on Oct 9, 2026
- Start on batched SVD support
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  1e69a96
- Offload a lot to MAK
  kshyatt
  committed
  ceb69ef
- Update Project.toml
  kshyatt
  committed
  b6db5fc
- Get rid of unneeded bypass
  kshyatt
  committed
  7fe1aa5
- Force the batched path to actually run and use QRIteration for AMD + isometric
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  b2ae3f7
- Update Project.toml
  kshyatt
  committed
  b739743
- Fix blocksectors handling
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  7dbd99a
- Try the select_inds fixer
  
  Katharine Hyatt
  authored and
  kshyatt
  committed
  e3f7cb7
- More CUDA tests and drop sources
  Katharine Hyatt
  committed
  d07bd79
- Diagonal + Jacobi doesn't work
  kshyatt
  committed
  537024d
Commits on Oct 12, 2026
- Another gesvdx painpoint on AMD
  Katharine Hyatt
  committed
  94ff845
- Make a specialization to avoid Diagonal batching for now
  Katharine Hyatt
  committed
  ee44be2