[WIP] GEMM: add support for mixed input scalar types#1557

Draft

mzuzek wants to merge 26 commits intokokkos:developfrom NexGenAnalytics:TODO/gemm-mixed-scalars

+5,192-4,690

Commits on Sep 28, 2022

GEMM: Move Serial, Team and TeamVector implementations to KokkosBlas
Mikołaj Zuzek
committed
GEMM: move unit tests to Blas
Mikołaj Zuzek
committed
GEMM: move MKL implementation of SerialGemm to dedicated TPL header
Mikołaj Zuzek
committed
GEMM: connect TeamVectorGemm to the selective interface
Mikołaj Zuzek
committed
GEMM: implicit MemberType
Mikołaj Zuzek
committed
GEMM: bring back batched interfaces for backward compatibility
Mikołaj Zuzek
committed
MKL: move utils to common header + fix macro duplication
Mikołaj Zuzek
committed
GEMM: fix "batched" in names
Mikołaj Zuzek
committed
GEMM: refactor crossing of A/B matrix transposes
Mikołaj Zuzek
committed
GEMM: implement ConjTranspose
Mikołaj Zuzek
committed
TeamGemv: move {Team,TeamVector}Internal to KokkosBlas2_team_gemv_internal.hpp
Mikołaj Zuzek
committed
TeamGemv: rename impl header
Mikołaj Zuzek
committed
TeamGemv: remove unused headers
Mikołaj Zuzek
committed
Gemv: move functor-level interfaces to the top-level header
Mikołaj Zuzek
committed
Gemv: implicit MemberType
Mikołaj Zuzek
committed
Merge branch 'gemm-transpose-refactoring' into gemv-transpose-refactoring
Mikołaj Zuzek
committed
Merge branch 'fix-gemv-blas-headers' into gemv-transpose-refactoring
Mikołaj Zuzek
committed
GEMV: refactor A matrix transpose
Mikołaj Zuzek
committed
Merge branch 'gemm-add-conjtranspose' into threadvector-kernels
Mikołaj Zuzek
committed
Merge branch 'gemv-transpose-refactoring' into threadvector-kernels
Mikołaj Zuzek
committed
SET/SCAL: add ThreadVector implementations and unit tests
Mikołaj Zuzek
committed
GEMV: add ThreadVector implementation and unit test
Mikołaj Zuzek
committed
GEMM: add ThreadVector implementation and unit test
Mikołaj Zuzek
committed

Commits on Sep 29, 2022

GEMM: Support for mixed scalar types in functor-level kernels
Mikołaj Zuzek
committed
GEMM: fix compilation errors in InnerGemmFixB
Mikołaj Zuzek
committed
GEMM: add mixed scalar support to InnerGemmFix{A,B}
Mikołaj Zuzek
committed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[WIP] GEMM: add support for mixed input scalar types#1557

Commits on Sep 28, 2022

GEMM: Move Serial, Team and TeamVector implementations to KokkosBlas

GEMM: move unit tests to Blas

GEMM: move MKL implementation of SerialGemm to dedicated TPL header

GEMM: connect TeamVectorGemm to the selective interface

GEMM: implicit MemberType

GEMM: bring back batched interfaces for backward compatibility

MKL: move utils to common header + fix macro duplication

GEMM: fix "batched" in names

GEMM: refactor crossing of A/B matrix transposes

GEMM: implement ConjTranspose

TeamGemv: move {Team,TeamVector}Internal to KokkosBlas2_team_gemv_internal.hpp

TeamGemv: rename impl header

TeamGemv: remove unused headers

Gemv: move functor-level interfaces to the top-level header

Gemv: implicit MemberType

Merge branch 'gemm-transpose-refactoring' into gemv-transpose-refactoring

Merge branch 'fix-gemv-blas-headers' into gemv-transpose-refactoring

GEMV: refactor A matrix transpose

Merge branch 'gemm-add-conjtranspose' into threadvector-kernels

Merge branch 'gemv-transpose-refactoring' into threadvector-kernels

SET/SCAL: add ThreadVector implementations and unit tests

GEMV: add ThreadVector implementation and unit test

GEMM: add ThreadVector implementation and unit test

Commits on Sep 29, 2022

GEMM: Support for mixed scalar types in functor-level kernels

GEMM: fix compilation errors in InnerGemmFixB

GEMM: add mixed scalar support to InnerGemmFix{A,B}