faisalkindi/DLSS5oneclick v0.13.2

v0.13.2 — Feeder 0.15.0 fixes the sRGB work_resolution bug; the workaround expires with it

on GitHub

DLSS5-Feeder 0.15.0 is out, and it fixes the Dying Light bug — so the workaround for it now expires by itself.

Upstream v0.15.0 is the first stable release in a while and it fixes #85, reported from #74 here:

Every work_resolution below 100% failed on an sRGB swapchain. The staging copy was created in the backbuffer's exact ..._UNORM_SRGB format and then viewed as ..._UNORM; a D3D11 view must match its resource unless the resource is typeless, so that call could never succeed. Any D3D11 game with an sRGB swapchain was affected.

Since v0.12.8 this tool has read work-resolution staging SRV failed out of a game's own log and pinned work_resolution to 100 for that game. Left as it was, it would hold that setting down forever, on evidence written by a Feeder build that the same install then replaces.

The hold now expires with the build that logged it: a failure from anything older than 0.15.0 is ignored, and one from 0.15.0 or newer is still honoured, where it would mean a genuine refusal. So updating through the tool gets you 0.15.0 and your reduced work resolution back in the same run.

Diagnose also stops calling a current Feeder out of date (CURRENT_FEEDER 0.12.0 → 0.15.0).

Two other fixes worth knowing about in that release, both from other people's reports:

• Feature-level 10 games could never build resources (64-bit). The add-on compiled its blit and FSR 1 shaders at Shader Model 5, which a feature level 10_x device rejects — every such game died at blit shader creation failed 0x80070057. If you have one that has never worked, retry it.
• OptiScaler DLSS-NR is now a supported neural consumer of the feed, installed as a winmm.dll proxy, 64-bit only.

SHA-256 of dlss5oneclick.exe: 4c09483d61d4bce59ae3ca817d8dc482185cf3bfb97b7c9f255281e25ba94c19