For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).
Mojo module
mma_amd
AMD CDNA Matrix Cores implementation for matrix multiply-accumulate operations.
This module provides MMA implementations for AMD CDNA2, CDNA3 and CDNA4 data
center GPUs using the MFMA (Matrix Fused Multiply-Add) instructions, and for
CDNA5 (gfx1250), whose matrix cores take the v_wmma instructions instead.
Reference: https://gpuopen.com/learn/amd-lab-notes/amd-lab-notes-matrix-cores-readme/
Functionsโ
- โ
wmma_cdna: Computes one CDNA5 WMMA,d = a * b + c, with the given reuse hints. - โ
wmma_scale_cdna: Computes one block-scaled CDNA5 WMMA,d = a * b * scale + c.