Raman Jana

icon indicating a website link https://ramjana.github.io icon indicating an email [email protected]

icon company AMD,inc icon location Austin,TX icon location Machine Learning, hardware-software co-design, GPU kernel development, ML performance modeling.

Results 2 issues of Raman Jana

Async DMA support for Transpose Data Layout

7 comment

Support to move directly to LDS support for Transpose Data layout Added feature constraints for GLVW>DWORD

NoCI

[CK_TILE] LDS prefetch pipeline support for FlashAttention

1 comment

Added support for LDS prefetch pipeline , enabled for one of the sources of GEMM sourced from LDS. could be applicable for both sources.