rocWMMA: rocWMMA (rocWMMA)
rocWMMA:
rocWMMA: rocWMMA is C++ library for accelerating mixed-precision matrix
rocWMMA: multiply-accumulate (MMA) operations leveraging AMD GPU hardware.
rocWMMA: rocWMMA makes it easier to break down MMA problems into fragments and
rocWMMA: distribute block-wise MMA operations in parallel across GPU
rocWMMA: wavefronts. The API consists of a header library, that can be used to
rocWMMA: compile MMA acceleration directly into GPU kernel device code. This
rocWMMA: can benefit from compiler optimization in the generation of kernel
rocWMMA: assembly, and does not incur additional overhead costs of linking to
rocWMMA: external runtime libraries or having to launch separate kernels.