llama.cpp
vulkan: optimizations for direct convolution
#14933
Merged

vulkan: optimizations for direct convolution #14933

0cc4m merged 7 commits into ggml-org:master from jeffbolznv:conv_opt
jeffbolznv
jeffbolznv vulkan: optimizations for direct convolution
9c12ef79
jeffbolznv jeffbolznv requested a review from 0cc4m 0cc4m 102 days ago
github-actions github-actions added Vulkan
github-actions github-actions added ggml
Green-Sky
jeffbolznv
Green-Sky
Green-Sky
etasnadi
jeffbolznv
Green-Sky
jeffbolznv
Green-Sky
Green-Sky
jeffbolznv
jeffbolznv
jeffbolznv Three tiles sizes for CONV_2D, and a heuristic to choose
136ecfbc
jeffbolznv
Green-Sky
etasnadi
daniandtheweb
netrunnereve
etasnadi
jeffbolznv
jeffbolznv reallow collectives for pre-Turing
95ee61ac
jeffbolznv
etasnadi
jeffbolznv
0cc4m
jeffbolznv
0cc4m
jeffbolznv
jeffbolznv
jeffbolznv
etasnadi
etasnadi
etasnadi
etasnadi commented on 2025-07-30
netrunnereve
netrunnereve
jeffbolznv make SHMEM_PAD a spec constant
7d3553fa
jeffbolznv
0cc4m
0cc4m
jeffbolznv
jeffbolznv
jeffbolznv fixes for intel perf - no shmem padding, placeholder shader core count
44566496
jeffbolznv shader variants with/without unrolling
e8643c0f
jeffbolznv
0cc4m
jeffbolznv 0cc4m's fixes for AMD perf
d2a65ece
0cc4m
0cc4m approved these changes on 2025-08-02
0cc4m 0cc4m merged a9f7541e into master 98 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone