Issues

• NKI 0.3.0 simulator: nisa.nc_matmul accumulate=True broken for K_max>1 inner affine_range iterations

  trnsci/trnsparse#28 · Filter by author scttfrdmn opened on Apr 29, 2026

• On-chip iterative solvers over BSR (SBUF-resident A)

  chemistryenhancementneuron
  trnsci/trnsparse#22 · Filter by author scttfrdmn opened on Apr 13, 2026
  3

• Phase 5 — trn2 DMA bandwidth exploitation for SpMM

  phase-5-generation
  trnsci/trnsparse#17 · Filter by author scttfrdmn opened on Apr 13, 2026

• Phase 4 — sharded sparse matrices across chips

  phase-4-multichip
  trnsci/trnsparse#16 · Filter by author scttfrdmn opened on Apr 13, 2026

• Phase 3 — nnz-bucketing SpMM + streaming large-sparse + NEFF cache

  phase-3-perf
  trnsci/trnsparse#15 · Filter by author scttfrdmn opened on Apr 13, 2026
  1