Back to feed

b8355

Mar 15, 2026
Meta/llama.cppCLIvb8355

cuda : add RDNA4-specific MMVQ parameter table for bs=1 decode (#19478)

  • mmvq: add RDNA3/RDNA4-specific parameter table (nwarps=8, rows=1)

  • mmvq: add dedicated RDNA3 parameter table

  • mmvq: exclude RDNA3.5 (gfx1150/1151) from RDNA3 table

macOS/iOS:

Linux:

Windows:

openEuler: