For the complete Mojo documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /docs/manual/basics.md).
mma_amd_rdna
AMD RDNA3/4 WMMA implementation for matrix multiply-accumulate operations.
This module provides MMA implementations for AMD RDNA3 and RDNA4 consumer GPUs using the WMMA (Wave Matrix Multiply Accumulate) instructions.
Reference: https://gpuopen.com/learn/wmma_on_rdna3/
Functions
-
load_matrix_a_amd_rdna16x16x16: Loads 16×16×16 matrix A tile for RDNA (Wave32) architecture. -
load_matrix_b_amd_rdna16x16x16: Loads 16×16×16 matrix B tile for RDNA (Wave32) architecture.