Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[DEV][TL] Support AMD Matrix Code Implementation #237

Merged
merged 23 commits into from
Nov 7, 2024

Conversation

LeiWang1999
Copy link
Contributor

This pull request includes significant changes to the bitblas package, primarily focusing on restructuring the initialization process and improving the layout functions. The most important changes include updating submodule references, refactoring the initialization script, and adding new layout functions.

Submodule Update:

  • Updated the tvm submodule to a new commit. (3rdparty/tvm)

Initialization Refactor:

  • Major refactoring of the bitblas/__init__.py file to streamline environment variable setup and module imports. This includes removing redundant code and reorganizing the import statements. [1] [2]

Logging Improvements:

  • Improved the logging setup by adjusting the formatter and ensuring consistent string formatting. (bitblas/__init__.py)

Import Path Updates:

  • Updated import paths in several files to reflect the new location of the mma_macro_generator module.
    • bitblas/ops/general_matmul/tilelang/dense/matmul_tensorcore.py
    • bitblas/ops/general_matmul/tilelang/dense/matmul_tensorcore_s4.py
    • bitblas/ops/general_matmul/tilelang/dequantize/finegrained_primitive_tensorcore.py
    • bitblas/ops/general_matmul/tilelang/dequantize/finegrained_primitive_tensorcore_s4.py
    • bitblas/ops/general_matmul/tilelang/dequantize/ladder_weight_transform_tensorcore.py
    • bitblas/ops/general_matmul/tilelang/dequantize/ladder_weight_transform_tensorcore_s4.py
    • bitblas/tl/__init__.py

New Layout Functions:

  • Added new layout functions for shared to local memory mapping in bitblas/tl/base_layout.py and bitblas/tl/mfma_layout.py. These functions facilitate efficient memory access patterns for tensor operations.
    • bitblas/tl/base_layout.py
    • bitblas/tl/mfma_layout.py

@LeiWang1999 LeiWang1999 merged commit ad19317 into microsoft:main Nov 7, 2024
3 of 4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant