Skip to content

Model L as a model choice: registry row, front-end WG_MODEL_L, model lookup and refusals - #155

Merged
phantoslayer21 merged 2 commits into
mainfrom
model-l/frontend-choice-pr
Nov 14, 2026
Merged

phantoslayer21 merged 2 commits into
mainfrom
model-l/frontend-choice-pr

Conversation

@phantoslayer21

Copy link
Copy Markdown
Owner

Refs #89

Model L as a model choice in the shared settings and the front end, off by default.

  • Registry: sr_model value L and the model row (runtime option model=L, build flag WRONGGPU_MODEL_L);
    regenerated header, ini text and REGISTRY.md. The tests' append-only baseline (tests/data/registry.prev.json)
    moves to release 0.6.0's registry: Model L's entry 0x5747000000000003 is appended after the released M and K
    entries.
  • Front end: CMake option WG_MODEL_L (default OFF, src/model_l.h). With it: the hosted list entry
    "WrongGPU Model L", sr_model=L / model=L passed to the runtime, the model lookup's <place>\L\model.wgcm, and a
    clear refusal when Model L is chosen and no converted Model L exists. Without it: model=L / sr_model=L are
    refused by name in the log and Model M runs, as today.
  • Tests: the front-end unit tests built with and without Model L; registry tests for the third entry.

Model M when the option is off: the same front end (the new code is behind WG_MODEL_L).

Tested: the integration build with WG_MODEL_L=ON carried the "WrongGPU Model L" entry; in Cyberpunk 2077 and
Resident Evil Requiem the host chose it through OptiScaler's [WrongGPU] UpscalerIndex=1, the front end found
<game>\WrongGPU\model\L\model.wgcm and passed it on, and Model L ran on the D3D12 executor.

…end WG_MODEL_L, model lookup and refusals

- registry: sr_model value L and the model row (runtime option model=L, build flag WRONGGPU_MODEL_L);
  regenerated header, ini text and REGISTRY.md. The append-only baseline of the tests
  (tests/data/registry.prev.json) is release 0.6.0's registry: Model L's entry 0x...03 is appended
  after the released M and K entries; a model put before a released one is refused.
- front end: CMake option WG_MODEL_L (default OFF, src/model_l.h). With it: the entry WrongGPU Model L,
  sr_model=L / model=L passed to the runtime, the model lookup looks for <place>\L\model.wgcm, and a
  create with Model L chosen and no converted Model L is refused with the reason (the runtime is not
  asked; AMD takes over as for any failed create). Without it: model=L (wronggpu.ini runtime_options,
  WG_RUNTIME_OPTIONS) and sr_model=L are refused by name in the log and Model M runs.
- tests: the front-end unit tests also built with Model L (test_frontend_model_l); new tests for both
  builds; the registry tests for the third entry.

Refs #89

Signed-off-by: phantoslayer21 <156617979+phantoslayer21@users.noreply.github.com>
@phantoslayer21
phantoslayer21 marked this pull request as ready for review November 14, 2026 08:13
Signed-off-by: phantoslayer21 <156617979+phantoslayer21@users.noreply.github.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant