llama.cpp
CUDA Graph Compute Function Refactor (precursor for performance improvements)
#11042

Merged

CUDA Graph Compute Function Refactor (precursor for performance improvements) #11042

slaren merged 11 commits into ggml-org:master from aendk:akieslinger/refactor_cuda_backend

github-actions added Nvidia GPU

github-actions added ggml

aendk marked this pull request as draft 282 days ago

aendk force pushed from 3998c0dc to 004ec3ad 277 days ago

Refactor: Moves cuda graph executable update step to separate function.

ba053310

Refactor: Moves cuda graph update check to separate function.

22c24294

Refactor: Moves cuda graph maintenance (update or adjusting copy para…

eb3ea698

Fix: Adds missing reference to maintain_cuda_graph() definition.

ed10ff58

Refactor: Improves structure and abstractions by moving CUDA graph ev…

37518b7d

Refactor: Moves node graph checks and copy ops into individual functi…

0cdc1339

Refactor: Removes code permanently excluded from compilation to incre…

dd95edfc

Style: Adds missing newline

98d4e55f

aendk force pushed from 004ec3ad to 98d4e55f 275 days ago

aendk marked this pull request as ready for review 275 days ago

slaren commented on 2025-01-10

Style: Consolidates several neighboring '#ifdef USE_CUDA_GRAPH' into …

fcd62d9d

Refactor: Makes 'cuda_graph_update_required' a local variable

62f2f624

slaren approved these changes on 2025-01-13

remove double lines between functions

5226732f

slaren merged 39509fb0 into master 271 days ago

Reviewers

slaren

Assignees

No one assigned

Labels

Nvidia GPU ggml

Milestone

No milestone