DLSS 5 Autopilot 2.0.5

Updating from 2.0.x with update now downloads this build, checks its
SHA-256 against the SHA256SUMS.txt published with the release, and offers
a one-click restart. Your settings, library, hidden games and profiles carry
over.

Fixed

• Shared results filed "it worked" as a failure (#414). share the
  result wrote "worked" only when the diagnosis said "Working." Where the
  logs cannot show the outcome - on renodx, native and bridge, which log no
  frames, and in some optiscaler and feeder sessions - every result shared
  went into the compatibility list as a failure, whatever the game did. 27
  of the first 230 shared results were filed that way, and the list is what
  the tool reads before every install to say which route works. When the logs
  cannot tell, share the result now asks you whether the DLSS 5 picture
  showed, and the record says the answer is yours. Old results with an
  outcome the tool could not see are left out of the list instead of being
  counted as failures. Six whose authors rewrote the title to say it
  worked are counted as worked. A result where Windows recorded the game crashing counts as a
  failure on any route.
• "Look in the overlay" when the game had crashed (#412). The report
  form asks "did the game start?". A game that closed itself or never
  started now replaces an "open the overlay" or "Working." verdict with what
  to take out first: a runtime swapped on the dlss page, then everything of
  ours, then another route. It shows in the report, in the window and on the
  game's card.
• Unreal Engine 1 and 2, and DirectX 8 (#403). POSTAL 2 was read as
  DirectX 11/12 and given a dxgi.dll it never loads. Unreal Engine 1/2
  folders are now read from their own ini (RenderDevice=) and renderer
  modules. DirectX 8 goes through DXVK like DirectX 9 does. Neither has been
  run on a real game here: the Unreal reading was checked against the
  report, and DX8 is marked experimental.
• "DXVK ran and ReShade did not" (#400, #348, #238). A 32-bit DirectX 9
  game on the feeder gets ReShade as a per-user Vulkan layer, and Windows'
  Vulkan loader skips per-user layers in a process that runs as
  administrator. The report says so when this tool runs elevated or the
  game's executable is set to run as administrator. The
  diagnosis names the executable's setting as the cause when it is there,
  and otherwise names elevation as one possibility and no longer says
  "install again". play and autopilot do not start a Vulkan-layer
  game while this tool runs as administrator. Not reproduced here.
• OptiScaler route: a Streamline crash dump went unread, and other DLSS
  add-ons went unnamed (#420, #20). The game's own Streamline writing a
  crash dump the session ended on is read as a crash. A ReShade left in the
  folder with DLSS add-ons is named, and so are the add-ons.
• "Check it is switched on" when it was (#390). On the neural-upstream
  route, switched on and hooked with no frame, the diagnosis now says the
  game never called DLSS: turn DLSS on in the game's settings, or use the
  optiscaler route.
• The sidebar overlapped itself at high display scaling. At 175% on a
  1080p screen, at 250%, or in a short window, the lower pages were drawn
  over watch and help. The sidebar now tightens, then shows icons
  only (hover for the name), then moves pages it cannot fit behind "more".

New

• Hide sidebar pages (#293). Right-click video, remix or
  vr in the sidebar and choose to hide it. games > view >
  sidebar pages brings them back.
• wilsjo2's fork + RTX 40 MFG in optiscaler build (#286): the
  package with the pre-upscaler pass and multi-frame generation together,
  installed with its RTX 40 unlock switched on ([DLSSG] AdaMfgUnlock=true;
  the package ships it off). The unlock acts on the game's own DLSS frame
  generation; the install warns when the game ships none or when FSR frame
  generation is on.
  It refuses to install on a card it knows is not an RTX 40, and is never
  chosen automatically.
• A search that finds nothing offers to add the game by hand (#374).