Schedule on Android: verify CoroutineSchedule on ART and measure scheduled SDPA on device
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 5/5
- 预计耗时
- 一周以上
- 新手友好度
- 25/100
- Issue 类型
- 缺陷
- 描述清晰度
- 描述清楚
- 活跃度
- 活跃
- 领域
- mobile, performance
调研方向
Read the Schedule getting started documentation, the Android DirectCpuExecutionContext() actual, and the related issues #1196, #1259, #1262, and #1266. First establish which default schedule Android resolves to, then run the requested [20, 1500, 64] SDPA comparison on a physical arm64 device and add an androidTest in skainet-backend-jni-cpu or skainet-backend-cpu. Done means the docs report measured numbers and the tests demonstrate bit-identity, with no deadlock or oversubscription.
由索引模型根据 Issue 内容生成。
描述
Context
0.54.0 introduced Schedule (SKEEP-005): scaledDotProductAttention runs its (batch, head) units on the context's schedule, and the JVM default is CoroutineSchedule.hardware(). The changelog describes the JVM behaviour and the CI deadlock fix (#1262, #1266); nothing states what the Android compilation resolves to.
An encoder-style workload (Whisper: 20 heads × 1500 frames, prefill-shaped, not decode-shaped) is exactly where head-level parallelism pays, and Android big.LITTLE cores make the pool sizing non-obvious (#1196 documented that sleeping pools lose to one pegged big core for sub-millisecond bursts).
Scope
- State and test what
DirectCpuExecutionContext()uses as its defaultScheduleon Android (androidMainactual, or the JVM one via the shared source set?). - Run the SDPA schedule on a physical arm64 device with prefill shapes (
[20, 1500, 64]) and record speedup vsSchedule.Sequential, on big cores only vs all cores. - Decide whether the JNI row-threading pool (
skainet_row_threads, spin-then-park) andCoroutineScheduleshould coexist per op or whether one should own the cores on Android; document the choice. -
TraceEvent.ScheduleDowngradedvisible on Android when a requested schedule is not honoured. - An
androidTestinskainet-backend-jni-cpu(orskainet-backend-cpu) that exercises a scheduled SDPA on device and asserts bit-identity with sequential.
Acceptance
- The docs' "Schedule getting started" page has an Android paragraph with measured numbers.
- No deadlock or oversubscription when a scheduled op runs inside a JNI-threaded matmul region and vice versa.
Related
- #1259, #1262, #1266 — Schedule introduction and the CI deadlock
- #1196 — threaded packed matmuls and the pool design notes
- 主要语言
- Kotlin
- 星标
- 52
- 派生
- 15
- 平均合并
- 1 天 8 小时
- 30 天内合并 PR
- 37
环境准备
- 没有 Dockerfile 或 Docker Compose 文件
- 没有 Pull Request 模板
- 阅读贡献指南
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
SKaiNET-developers/SKaiNET 的其他 Issue
-
coding good first issue platform size:xs skill:js sub-issue
难度 2/5 1-3 小时 新手友好度 72/100
SKaiNET-developers/SKaiNET#1232 ·
维护者通常 1 天内回复
-
compute-backend dsl/api enhancement size:m skill:kotlin-core skill:numerics
难度 5/5 一周以上 新手友好度 18/100
SKaiNET-developers/SKaiNET#1343 ·
维护者通常 1 天内回复
-
compute-backend dsl/api enhancement size:m skill:kotlin-core skill:numerics
难度 5/5 一周以上 新手友好度 12/100
SKaiNET-developers/SKaiNET#1342 ·
维护者通常 1 天内回复
-
skeep tracking
难度 5/5 一周以上 新手友好度 15/100
SKaiNET-developers/SKaiNET#1331 · 1 条评论 ·
维护者通常 1 天内回复
-
assessment size:s skill:review sub-issue
难度 4/5 1-2 天 新手友好度 18/100
SKaiNET-developers/SKaiNET#1330 ·
维护者通常 1 天内回复
查看 SKaiNET-developers/SKaiNET 的全部 Issue
相似的 Issue
-
afk-ok area:data-quality importer size:S
难度 2/5 1-3 小时 新手友好度 74/100
enorm-labs/event-junkie#3301 ·
维护者通常 1 天内回复
-
MDM: interfaceName cannot be set through managed configurations可能已有人在做 @JulienUsson 今天认领。 未关闭bug
难度 1/5 1 小时以内 新手友好度 82/100
bk138/droidVNC-NG#404 · 1 条评论 ·
-
难度 2/5 1-3 小时 新手友好度 65/100
维护者通常 2 天内回复
-
难度 2/5 1-3 小时 新手友好度 72/100
rustyrazorblade/easy-db-lab#1017 ·
维护者通常 1 天内回复
-
难度 2/5 1-3 小时 新手友好度 62/100
openclaw/openclaw#172218 · 1 条评论 ·
维护者通常 1 天内回复