feat[fastlanes]: add optimized 1024-bit transpose implementations#6135

Closed

joseph-isaacs wants to merge 15 commits intodevelopfrom

claude/bitpacking-transpose-optimization-tM1U4

Commits on Mar 9, 2026

feat[fastlanes]: add optimized 1024-bit transpose implementations

claude
authored and
robert3005
committed
chore[fastlanes]: add perf measurement examples for transpose

claude
authored and
robert3005
committed
feat[fastlanes]: add BMI2 PEXT/PDEP transpose and fix GFNI implementations

claude
authored and
robert3005
committed
feat[fastlanes]: add scalar_fast and ARM64 NEON transpose implementations

claude
authored and
robert3005
committed
perf[fastlanes]: fully unroll BMI2 transpose for 12% performance gain

claude
authored and
robert3005
committed
test[fastlanes]: add verification against fastlanes crate transpose

claude
authored and
robert3005
committed
feat[fastlanes]: add AVX-512 VBMI transpose with 7.5x speedup

claude
authored and
robert3005
committed
feat[fastlanes]: add dual-block VBMI transpose for 10% more throughput

claude
authored and
robert3005
committed
feat[fastlanes]: add 4-block VBMI transpose for 7% additional speedup

claude
authored and
robert3005
committed
docs[fastlanes]: add transpose optimization plan and results

claude
authored and
robert3005
committed
use upstream
robert3005
committed
more
robert3005
committed
moves
robert3005
committed
remove
robert3005
committed
more
robert3005
committed