llama.cpp
cuda : add RDNA4-specific MMVQ parameter table for bs=1 decode
#19478
Merged

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

JoursBleu
JoursBleu JoursBleu requested a review from JohannesGaessler JohannesGaessler 237 days ago
github-actions github-actions added Nvidia GPU
github-actions github-actions added ggml
IMbackK
IMbackK requested changes on 2026-02-11
JoursBleu JoursBleu force pushed 236 days ago
JohannesGaessler
JohannesGaessler commented on 2026-02-12
JohannesGaessler
JohannesGaessler
JoursBleu JoursBleu force pushed 228 days ago
JohannesGaessler
JohannesGaessler commented on 2026-02-19
JoursBleu JoursBleu force pushed 228 days ago
JoursBleu
JohannesGaessler
JoursBleu JoursBleu force pushed 222 days ago
JoursBleu JoursBleu force pushed to 2df360d9 221 days ago
JoursBleu
JohannesGaessler
JoursBleu
JoursBleu JoursBleu force pushed from 2df360d9 to f98218de 217 days ago
JohannesGaessler
JohannesGaessler commented on 2026-03-02
JoursBleu JoursBleu force pushed from f98218de to c10e46ba 217 days ago
JoursBleu
IMbackK IMbackK requested a review from IMbackK IMbackK 216 days ago
IMbackK IMbackK assigned IMbackK IMbackK 216 days ago
JohannesGaessler
JohannesGaessler commented on 2026-03-03
JohannesGaessler
IMbackK
JoursBleu
JoursBleu mmvq: add RDNA3/RDNA4-specific parameter table (nwarps=8, rows=1)
462b5757
JoursBleu mmvq: add dedicated RDNA3 parameter table
8765fdd1
JoursBleu mmvq: exclude RDNA3.5 (gfx1150/1151) from RDNA3 table
9bb430bc
JoursBleu JoursBleu force pushed from c10e46ba to 9bb430bc 213 days ago
JoursBleu
JohannesGaessler
JohannesGaessler approved these changes on 2026-03-15
JohannesGaessler JohannesGaessler merged 617db241 into master 204 days ago
meven3000
IMbackK
meven3000
JoursBleu
JoursBleu

Login to write a write a comment.

Login via GitHub

Assignees
Labels
Milestone