metal: only encode in one command buffer

Browse source

It's advised a program should only have one command buffer. This slow
inference by ~1 ms on 33B model, but we may avoid it by reusing
previous command queue.

...

This commit is contained in:

lshzh-ww 2023-07-21 11:17:48 -04:00

parent d924522a46

commit c8e6ef1846

1 changed files with 517 additions and 548 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download patch file Download diff file Expand all files Collapse all files

1065

ggml-metal.m

View file

File diff suppressed because it is too large Load diff