[Dev] Potentially improve performance through block reduction#63

Merged

LeiWang1999 merged 38 commits intomicrosoft:mainfrom LeiWang1999:lei/splitk

Jun 30, 2024

+1,102-49

Commits on May 21, 2024

improve e4m3 decoding.
LeiWang199
committed

Commits on May 23, 2024

Merge branch 'main' of https://github.com/microsoft/BitBLAS into main
LeiWang199
committed

Commits on May 25, 2024

append fp16xint1
LeiWang199
committed

Commits on Jun 1, 2024

Update submodule commit reference
LeiWang199
committed
chore: Update shared memory scope for float32 output dtype
LeiWang199
committed

Commits on Jun 2, 2024

BUGFIX: UINT8/INT8 Decoding
LeiWang199
committed

Commits on Jun 5, 2024

feat: Add rasterization options for roller module
LeiWang199
committed
Refactor tensorcore_legalization method to optimize tensor core usage
LeiWang199
committed
feat: Add function to collect variables from expression, improve for splitk
LeiWang199
committed
chore: Update typing import in __init__.py
LeiWang199
committed
chore: Refactor CPU execution of operators
LeiWang199
committed
Refactor matmul implementation for splitk layout
LeiWang199
committed
Refactor matmul implementation for splitk layout
LeiWang199
committed
Refactor matmul implementation for splitk layout
LeiWang199
committed
chore: Update version to 0.0.1.dev8
LeiWang199
committed
chore: Enable debug output in bitblas.set_debug_level()
LeiWang199
committed
Refactor Linear module matmul implementation for splitk layout
LeiWang199
committed
Refactor matmul implementation for splitk layout
LeiWang199
committed
Merge branch 'main' of https://github.com/microsoft/BitBLAS into lei/splitk
LeiWang199
committed
Refactor CUDA kernel launch string for dynamic symbolic set
LeiWang199
committed
Bumpt version to v0.0.1.dev9
LeiWang199
committed
Merge branch 'main' of https://github.com/microsoft/BitBLAS into lei/splitk
LeiWang199
committed

Commits on Jun 29, 2024

fix the typo
LeiWang199
committed
Refactor CUDA kernel launch string for dynamic symbolic set
LeiWang199
committed
Refactor CUDA kernel launch string for dynamic symbolic set
LeiWang199
committed
Refactor CUDA kernel launch string for dynamic symbolic set
LeiWang199
committed