-
Notifications
You must be signed in to change notification settings - Fork 77
All issues
Issue creation is restricted in this repository
- #545 · zhusy54 opened
on Apr 14, 2026 3 - #995 · ChaoZheng109 opened
on Jun 5, 2026 1 - #1706 · ChaoZheng109 opened
on Aug 6, 2026 4
Issues
is:issue state:open
is:issue state:open
Search results
L2 dispatcher passes xdist options to a child with xdist disabled
bugSomething isn't workingSomething isn't workingcode healthTechnical debt, robustness, code qualityTechnical debt, robustness, code qualityStatus: Open.#1993 In hw-native-sys/simpler;[Feature] Size the hbg graph heap from task lifetimes plus a materialization lookahead cap, not a 256 MB constant
enhancementNew feature or requestNew feature or requestStatus: Open.#1956 In hw-native-sys/simpler;[Feature] Expose an explicit task-to-task dependency (add_dep) on the host-level submit API
enhancementNew feature or requestNew feature or requestStatus: Open.#1954 In hw-native-sys/simpler;[Code Health] Nothing records whether a run paid torch_npu's device autoload
code healthTechnical debt, robustness, code qualityTechnical debt, robustness, code qualityStatus: Open.#1950 In hw-native-sys/simpler;[Bug] SIGTERM during INIT_FAILED publication can escape forked child main
bugSomething isn't workingSomething isn't workingStatus: Open.#1948 In hw-native-sys/simpler;[Code Health] CTRL_FREE hands the chip child a bare device pointer it frees unconditionally
code healthTechnical debt, robustness, code qualityTechnical debt, robustness, code qualityStatus: Open.#1931 In hw-native-sys/simpler;[Code Health] hbg: ready queues are sized by a fixed constant, and one is reachable by more tasks than it has slots
code healthTechnical debt, robustness, code qualityTechnical debt, robustness, code qualityStatus: Open.#1920 In hw-native-sys/simpler;- Status: Open.#1878 In hw-native-sys/simpler;
[Feature] Let consecutive runs queue on the same stream instead of round-tripping through a host sync
enhancementNew feature or requestNew feature or requestStatus: Open.#1853 In hw-native-sys/simpler;[Performance] hbg: host-view register/unregister of every entry tensor costs ~256 ms/run unregistering 38 GiB the orchestrator never reads
performancePerformance regression or optimizationPerformance regression or optimizationStatus: Open.#1848 In hw-native-sys/simpler;[Performance] Orchestrator避免频繁的内存分配和释放
performancePerformance regression or optimizationPerformance regression or optimizationStatus: Open.#1841 In hw-native-sys/simpler;