PR #16623 metal: TRI, FILL, EXPM1, SOFTPLUS

metal: TRI, FILL, EXPM1, SOFTPLUS #16623

ggerganov merged 12 commits into ggml-org:master from gabe-l-hart:ggml-cumsum-tri

gabe-l-hart requested a review from

slaren 261 days ago

gabe-l-hart requested a review from

CISC 261 days ago

gabe-l-hart requested a review from

ggerganov 261 days ago

github-actions added testing

github-actions added Nvidia GPU

github-actions added examples

github-actions added ggml

github-actions added Apple Metal

feat(wip): Port initial TRI impl from pervious work

a9c92440

fix: Remove argument for constant val override

2a7bbc77

feat: Move the ttype conditional to templating to avoid conditional i…

f2ad8877

gabe-l-hart force pushed from 5071fbd5 to f2ad8877 213 days ago

gabe-l-hart changed the title ~~ggml: CUMSUM and TRI (CPU, Metal, CUDA)~~ metal: TRI 213 days ago

ggerganov commented on 2025-12-03

fix: Type fixes

6a270503

feat: Add softplus for metal

7cbbff74

feat: Add EXPM1 for metal

434ec074

feat: Add FILL for metal

1496afdf

gabe-l-hart changed the title ~~metal: TRI~~ metal: TRI, FILL, EXPM1, SOFTPLUS 213 days ago

refactor: Branchless version of tri using _ggml_vec_tri_cmp as a mask

76908089

ggerganov approved these changes on 2025-12-04

fix: Remove unused arguments

60fe39b0

Merge remote-tracking branch 'origin/master' into ggml-cumsum-tri

c195af2b

Merge remote-tracking branch 'origin/master' into ggml-cumsum-tri

d006f357

refactor: Use select instead of branch for softplus non-vec

338acb32

ggerganov merged bde188d6 into master 212 days ago

gabe-l-hart deleted the ggml-cumsum-tri branch 212 days ago

Reviewers

ggerganov

slaren

CISC

Assignees

No one assigned

Labels

testing Nvidia GPU examples ggml Apple Metal

Milestone

No milestone

llama.cpp metal: TRI, FILL, EXPM1, SOFTPLUS #16623 Merged

metal: TRI, FILL, EXPM1, SOFTPLUS #16623

llama.cpp
metal: TRI, FILL, EXPM1, SOFTPLUS
#16623

Merged