perf: AMX BF16 dispatch in EuclideanDistances::computeABt#3564

Open

napetrov wants to merge 21 commits into

uxlfoundation:mainfrom

napetrov:feature/amx-bf16-knn-euclidean

Commits on Mar 28, 2026

perf: AMX BF16 dispatch in EuclideanDistances::computeABt
napetrov
committed
perf: address review comments for AMX BF16 dispatch in computeABt
napetrov
committed
perf: add Float32MatmulPrecision API and refactor AMX-BF16 dispatch
napetrov
committed
perf: separate HW capability from precision hint in matmul precision API
napetrov
committed
docs: document Float32MatmulPrecision levels in cpu_type.h
napetrov
committed
refactor: move Float32MatmulPrecision to daal::internal (experimental)
napetrov
committed
docs: fix Float32MatmulPrecision level descriptions
napetrov
committed
refactor: rename Float32MatmulPrecision levels to strict/allow_bf16
napetrov
committed
fix: correct namespace for CpuFeature::amx_bf16 in env_detect_precision.cpp
napetrov
committed

Commits on Apr 1, 2026

Commits on Apr 16, 2026

Commits on Apr 17, 2026

style: format BF16 GEMM dispatcher
napetrov
committed

Commits on Apr 27, 2026

feat: add required BF16 matmul precision mode
napetrov
committed

Commits on Apr 30, 2026

Merge branch 'main' into feature/amx-bf16-knn-euclidean
napetrov
authored

Commits on May 4, 2026

style: format AMX BF16 precision helper
napetrov
committed