Repository navigation

arm64: NEON BFDOT sbgemm kernel for VORTEX (Apple M1-M3), k-pair interleaved packing - #6145

Open

0tomhoyt wants to merge 7 commits into

OpenMathLib:developfrom

0tomhoyt:sbgemm-bfdot

Open

arm64: NEON BFDOT sbgemm kernel for VORTEX (Apple M1-M3), k-pair interleaved packing#6145
0tomhoyt wants to merge 7 commits into
OpenMathLib:developfrom
0tomhoyt:sbgemm-bfdot

Commits on Oct 20, 2026
- arm64: SME2 bf16 sbgemm kernel for VORTEXM4
  0tomhoyt
  committed
  cf50902
Commits on Oct 21, 2026
- arm64: keep the shared SBGEMM P/Q/R defaults for VORTEXM4
  0tomhoyt
  committed
  0bc31e8
- arm64: drop stray SBGEMM P/Q/R undefs picked up by other cpu blocks
  0tomhoyt
  committed
  c151d85
Commits on Oct 22, 2026
- arm64: fix the SME2 sbgemm kernel's ZA handoff and BFMOPA predicates
  0tomhoyt
  committed
  bce51f1
- arm64: NEON BFDOT sbgemm kernel for VORTEX (Apple M1-M3)
  0tomhoyt
  committed
  37c8967
- arm64: pack the sbgemm panels k-pair interleaved
  0tomhoyt
  committed
  e73ab68
- arm64: vectorize the k-pair interleaved packing
  0tomhoyt
  committed
  5dea71f