metal : allow ops to run concurrently #15929
ggerganov
force pushed
to
417df406
116 days ago
metal : run graphs ops concurrently
2fb15521
cont : add flags for debugging and disabling concurrency
a3519fd5
cont : refactor and handle fusing
74d2961e
cont : simplify - no need to use GPU address
acd1404a
cont : prepare mem ranges for reuse + add ggml-metal-common.cpp
f7aeab93
cont : avoid redundant keywords in cpp [no ci]
1c9d3f3b
metal : reorder graph for better concurrency
89cca2ad
metal : fix race on mem pool buffers
a3f17d61
cont : add env GGML_METAL_GRAPH_OPTIMIZE_DISABLE
0b586365
cont : refactor, optimize, add comments
907616de
cont : refactor ggml-metal.m
faffbec3
ggerganov
force pushed
from
17cf93d2
to
faffbec3
115 days ago
minor : update logs [no ci]
e502db1f
ggerganov
merged
f161463a
into master 115 days ago
ggerganov
deleted the gg/metal-concurrent-graphs branch 115 days ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub