GPU-friendly truncation implementations#349

Merged

lkdvos merged 30 commits intomainfrom

Jan 28, 2026

Commits on Jan 22, 2026

try to make truncation GPU-friendly
lkdvos
committed
Temporarily fix StridedViews version

kshyatt
authored and
lkdvos
committed
Revert "Temporarily fix StridedViews version"

kshyatt
authored and
lkdvos
committed
Small update for diagonal pullbacks

kshyatt
authored and
lkdvos
committed
Fix last error

kshyatt
authored and
lkdvos
committed
Reenable truncated CUDA tests

kshyatt
authored and
lkdvos
committed
make truncation run on GPU
lkdvos
committed
bypass scalar indexing by specializing
lkdvos
committed
convenience overloads
lkdvos
committed
gpu-friendly copies
lkdvos
committed
retain storagetype in extended_S
lkdvos
committed
avoid GPU issues with truncated adjoint tensormaps
lkdvos
committed
various utility improvements
lkdvos
committed
complete rewrite of implementation
lkdvos
committed
GPU doesn't like `trues`
lkdvos
committed
remove CUDA specializations and temporarily add missing MatrixAlgebraKit thingies
lkdvos
committed
better dimension testing
lkdvos
committed
fix unbound type parameter
lkdvos
committed
add missing import
lkdvos
committed
be careful about double method definitions
lkdvos
committed

Commits on Jan 26, 2026

Commits on Jan 27, 2026