Merge pull request #701, translator patch 0014 out so DLSS stops shaking the picture (itaskaev2)

Measured on an RTX 4080 SUPER by the author: shift 0.86 px a frame with the patch, 0.00 without it.

Review note: 0014 alone moves the SPIR-V hash of the game's FXAA shader from 5c43ae18948d52cb, which is in FXAA_FS
(runtime/vk_dlaa.c vk_dlaa_skip_draw), to aad42238801d31d8, which is not. With the initializers removed and the id gaps
closed, the 0014 build's shader disassembles identically to the pre-0014 one. So since c0e66fe the FXAA list matched
nothing, and this merge makes it match again, which also brings back fxaa = 0. The hash miss is the likely cause of the
shake rather than a driver reaction to OpConstantNull, but the path from the missed skip to the screen is not traced and
no RTX run checked it. Not run here: translator build, game run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Valiantsin DzerakhandClaude Opus 5.5 committed 2026-11-04 17:00:45 +03:00
commit 3ae9e8360d
3 files changed
+4 -22

No files matched your search

+1 -1
View File
@@ -88,7 +88,7 @@ Source shorthands:
| 3.9 | **Integer vertex inputs read as raw bits** (bone indices `R8UI`). | Float-declared inputs convert integers. | NXN Mario: stretched characters; re-translated with `uint` inputs. Ours: attribute types passed to `QueryGraphicsState` (`Program.cs`). | NXN |
| 3.10 | **BFE / BFI fields past bit 31** (as PTX `bfe` / `bfi`): a field that runs past bit 31 is cut there; one that starts past it extracts 0 (signed: bit 31 repeated) and inserts nothing. TotK 1.4.3's Depths terrain vertex shaders (42 variants) take the position from a register, `((2y + x) * 8) & 0xff` with 5-bit x and y, so up to 248. | `OpBitField*Extract` / `OpBitFieldInsert` undefined when offset or offset + count exceeds 32; AMD's bitfield instructions take offset and width modulo 32. | Ryujinx passes the fields through. UltraTears saw the Depths walls and ground missing on the Steam Deck and a Linux RX 570 (RADV). Ours: clamped where the field is not a constant inside the word (translator patch 0012). RX 9070 on Windows drew the Depths the same before and after. | PTX ISA `bfe`, `bfi` |
| 3.11 | **SHF by 0** gives the unshifted word (C for `SHF.L`, A for `SHF.R`). | Ryujinx's funnel shifts the other word by 32: undefined in SPIR-V; AMD shifts take the low 5 bits and give A \| C. | yuzu fixed Depths terrain gaps in its funnel shift (PR 10459). Ours: a select on a zero amount (translator patch 0013). In 1.4.3 every register SHF pairs its word with 0, so the result was right already. | — |
| 3.12 | **Registers read before any write** on a path (paths the structurizer adds included) have no defined value in the game's code; UltraTears reports 0 from the NVIDIA driver. | An `OpVariable` without an initializer loads an undefined value; Mesa folds undefined values freely. | Ours: every function-local variable starts at 0 (`OpConstantNull`, translator patch 0014). | — |
| 3.12 | **Registers read before any write** on a path (paths the structurizer adds included) have no defined value in the game's code; UltraTears reports 0 from the NVIDIA driver. | An `OpVariable` without an initializer loads an undefined value; Mesa folds undefined values freely. | Ours: none for now. Translator patch 0014 started every function-local variable at 0 (`OpConstantNull`); on NVIDIA (RTX 4080 SUPER, 1.4.3) that made the whole picture shake by about 0.9 output pixels a frame with DLSS or DLAA (#612; no shake with `TOTK_DLAA_JITTER=0` or without the patch), so it was taken out. A zero start for Mesa only, or for the shaders that need it, is still open. | — |
## 4. Memory model