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.

Sign up for GitHub

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

Jump to bottom

M1 AMX #496

Open

breuera wants to merge 3 commits into libxsmm:hfp_main from breuera:m1_amx

Contributor

breuera commented Jul 23, 2021

The PR contains a revised M1-AMX kernel generation doing C+=A*B^T

Supports FP16, FP32, FP64.
Supports generic block sizes in the micro-kernel (multiples of vec-length).
Supports leading dimensions not matching M, N, K.
Supports wrapping one micro-kernel config for FP16, FP32 and FP64 respectively. This allows for multiples of 64 in M; 32 (FP16) or 16 (FP32) or 8 (FP64) for N and arbitrary K.

breuera added 2 commits

July 23, 2021 20:40


          Added aarch64 orr with shifted register.

179e757


          Added revised AMX-kernel generation for M1.

7ce4ec4

breuera mentioned this pull request

Added support for first M1-AMX kernel #492

Closed


          Replaced add-imm12 with add-meta for address offsets in M1-AMX kernel.

788cb1f

Contributor

FreddieWitherden commented Feb 22, 2022

Is there anything in particular blocking this PR?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment