fix(vk): LONGSHOT_FSR=1 builds fail since #1989: the FSR 4 warm calls make_context with 3 of its 4 arguments #2122

Closed
opened 2026-11-21 21:01:12 +00:00 by dyst · 1 comment
Member

Since #1989 (d76d202f5, merged 2026-11-21 18:41 UTC), a build with LONGSHOT_FSR=1 stops at gpu/vk/vk_fsr4.cpp:

gpu/vk/vk_fsr4.cpp:1536:21: error: no matching function for call to 'make_context'

Each PR built on its own, but they conflict:

  • #1886 (69769ac96, 17:17 UTC) added the preparation's warm call, make_context(width, height, mode != 0) (line 1536).
  • #1989 then gave make_context a fourth parameter with no default, Fsr4Model want (line 1168). It updated the per-frame call (line 1330) but not the warm call.

The code is inside #ifdef LONGSHOT_FSR, so builds without AMD FSR 4 are unaffected. CI sets the flag: .forgejo/workflows/build-linux.yml:248 and .forgejo/workflows/vulkan-dlss.yml:64. Those jobs should be failing on main from d76d202f5 on; I haven't looked at their runs.

Likely fix: pass the model the per-frame path uses (prepare, line 1320), so the preparation warms the pipelines of the model the player picked:

s.context = make_context(width, height, mode != 0, fsr4_model());

How it was found: reading the code, then a syntax-only compile of the one file on main e7579b0a4 with the renderer's include folders and -DLONGSHOT_FSR (llvm-mingw clang++, --target=x86_64-w64-mingw32). That one error is the only one. No full build, and no run.

Since #1989 (`d76d202f5`, merged 2026-11-21 18:41 UTC), a build with `LONGSHOT_FSR=1` stops at `gpu/vk/vk_fsr4.cpp`: ``` gpu/vk/vk_fsr4.cpp:1536:21: error: no matching function for call to 'make_context' ``` Each PR built on its own, but they conflict: - #1886 (`69769ac96`, 17:17 UTC) added the preparation's warm call, `make_context(width, height, mode != 0)` (line 1536). - #1989 then gave `make_context` a fourth parameter with no default, `Fsr4Model want` (line 1168). It updated the per-frame call (line 1330) but not the warm call. The code is inside `#ifdef LONGSHOT_FSR`, so builds without AMD FSR 4 are unaffected. CI sets the flag: `.forgejo/workflows/build-linux.yml:248` and `.forgejo/workflows/vulkan-dlss.yml:64`. Those jobs should be failing on main from `d76d202f5` on; I haven't looked at their runs. **Likely fix:** pass the model the per-frame path uses (`prepare`, line 1320), so the preparation warms the pipelines of the model the player picked: ```cpp s.context = make_context(width, height, mode != 0, fsr4_model()); ``` **How it was found:** reading the code, then a syntax-only compile of the one file on main `e7579b0a4` with the renderer's include folders and `-DLONGSHOT_FSR` (llvm-mingw clang++, `--target=x86_64-w64-mingw32`). That one error is the only one. No full build, and no run.
Member

Possibly related, for context:

  • PR #1886 perf(vk): warm FSR 4 pipelines during shader preparation — Introduced the FSR 4 warm-up call that now has the stale argument list.
  • PR #1989 feat(vk): add FSR 4 model selection plumbing — Changed make_context's signature, leaving the warm-up call without the required model argument.
Possibly related, for context: - PR #1886 perf(vk): warm FSR 4 pipelines during shader preparation — Introduced the FSR 4 warm-up call that now has the stale argument list. - PR #1989 feat(vk): add FSR 4 model selection plumbing — Changed make_context's signature, leaving the warm-up call without the required model argument.
Sign in to join this conversation.
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
MagicBeanDev/Longshot#2122
No description provided.