Reduce some unnecessary allocations in DMA handler (#2886)

Browse Source

* experimental changes to try and reduce allocations in kernel threading and DMA handler

* Simplify the changes in this branch to just 1. Don't make unnecessary copies of data just for texture-texture transfers and 2. Add a fast path for 1bpp linear byte copies

* forgot to check src + dst linearity in 1bpp DMA fast path. Fixes the UE4 regression.

* removing dev log I left in

* Generalizing the DMA linear fast path to cases other than 1bpp copies

* revert kernel changes

* revert whitespace

* remove unneeded references

* PR feedback

Co-authored-by: Logan Stromberg <lostromb@microsoft.com>
Co-authored-by: gdk <gab.dark.100@gmail.com>

...

This commit is contained in:

Logan Stromberg 2022-07-14 11:45:56 -07:00 committed by GitHub

parent c5bddfeab8

commit 6eb85e846f

No known key found for this signature in database

GPG Key ID: 4AEE18F83AFDEB23

1 changed files with 33 additions and 14 deletions

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

Show Stats Download Patch File Download Diff File Expand all files Collapse all files