Outgrown point-motion textures are kept until the context is destroyed
維護者通常 1 天內回覆
還沒有人認領這個 Issue。
評估
- 難度
- 4/5
- 預估耗時
- 3-5 天
- 新手友好度
- 32/100
- Issue 類型
- 缺陷
- 描述清晰度
- 基本清楚
- 活躍度
- 活躍
- 技術堆疊
- cpp
研究方向
Start in clean/runtime/interop/executor_d3d12.cpp at point_texture (around line 404), where the point-motion texture is reallocated and the old one is pushed to retired_, and at the destructor (around line 61), the only place it is released. Check how mpoint_w_ and the render rectangle feed the growth path, and how the parked executor in clean/fsr/src/frontend.h keeps retired_ alive. Done means either a single maximum-size allocation from create, or retired textures released once external_done() reaches their recorded value, verified by a WARP test per #62. Read #62 and #64 first, since the ordering assumption depends on them.
由索引模型根據 Issue 內容生成。
描述
What happens
point_texture (clean/runtime/interop/executor_d3d12.cpp:404) reallocates the point-motion texture when a frame needs a larger one. The old texture is pushed to retired_ (:406), because a recorded list may still use it. The only release is in the destructor (:61).
Growth is monotonic (w = max(w, mpoint_w_), :407), so allocation stops once the largest render rectangle has been seen. The total is bounded by the sum over the first upward sweep.
When it can happen
Only with motion_outres=point, which is not the default (original is, clean/runtime/interop/backend.cpp:174, clean/FORMATS.md:692), and only when the game sets WG_CREATE_MOTION_OUTPUT_RES. The parked executor (clean/fsr/src/frontend.h:156) can keep retired_ alive across context re-creation with the same parameters.
A typical case is a few textures. The worst case is dynamic resolution that rises in many small steps, where the retired textures add up to a noticeable amount of VRAM on a smaller card.
Found by reading the code. It has not been reproduced.
To do
- Allocate
mpoint_once at the context's maximum render size taken from create. This removesretired_. The cost is one maximum-size texture, which the growth path reaches anyway. - Or record
external_recorded()when a texture is retired and release it inrecord()onceexternal_done()reaches that value. This relies on the ordering assumption in #64, which holds for a single context.
Test
WARP, using the executor seam described in #62. Feed frames with render widths 100, 101, 102, and so on, and assert that retired_ stays empty (maximum-size allocation) or drains after external_done() advances.
Related: #62, #64.
- 主要語言
- C++
- 星號
- 12
- 分支
- 2
- 平均合併
- 3 小時 10 分鐘
- 30 天內合併 PR
- 171
環境準備
- 沒有 Dockerfile 或 Docker Compose 檔案
- 有 Pull Request 範本
- 閱讀貢獻指南
從這裡開始
- 先讀完整個 Issue,再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫,在一個分支上完成修改。
- 送出 Pull Request,並在描述裡引用這個 Issue 編號。
phantoslayer21/wronggpu 的其他 Issue
-
難度 2/5 1-3 小時 新手友好度 88/100
phantoslayer21/wronggpu#275 ·
維護者通常 1 天內回覆
-
難度 2/5 1-3 小時 新手友好度 82/100
phantoslayer21/wronggpu#274 ·
維護者通常 1 天內回覆
-
bug
難度 2/5 1-3 小時 新手友好度 88/100
phantoslayer21/wronggpu#241 ·
維護者通常 1 天內回覆
-
bug build
難度 2/5 1-3 小時 新手友好度 74/100
phantoslayer21/wronggpu#234 ·
維護者通常 1 天內回覆
-
bug build
難度 2/5 1-3 小時 新手友好度 80/100
phantoslayer21/wronggpu#233 ·
維護者通常 1 天內回覆
查看 phantoslayer21/wronggpu 的全部 Issue
相似的 Issue
-
ai_p2
難度 2/5 1-3 小時 新手友好度 80/100
ClickHouse/ClickHouse#132519 ·
維護者通常 1 天內回覆
-
難度 2/5 1-3 小時 新手友好度 76/100
juce-framework/JUCE#1810 ·
-
難度 1/5 1 小時以內 新手友好度 82/100
petercorke/robotics-toolbox-python#952 ·
維護者通常 1 天內回覆
-
NatronRenderer: a Python error in the script (-t / project script) doesn't change the exit status未關閉func:engine type:bug
難度 2/5 1-3 小時 新手友好度 72/100
NatronGitHub/Natron#1276 ·
維護者通常 1 天內回覆
-
難度 1/5 1-3 小時 新手友好度 77/100
維護者通常 1 天內回覆