Timeline for Concurrent Handle Table
Current License: CC BY-SA 4.0
4 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 25, 2022 at 22:18 | comment | added | BlamKiwi | For modifying the CAS operation, I assume acquire/release for success and acquire for failure? Or is relaxed on failure sufficient? | |
| Oct 25, 2022 at 21:22 | comment | added | BlamKiwi | clear() is not "thread-safe" in the sense that calling while threads are active will cause undefined behaviour. A full implementation will need to observe changes from other threads. The intended use case is to clean up after stalled hardware queues (rare but can happen). The acquire fence should be hoisted out of the loop though. | |
| Oct 25, 2022 at 21:09 | comment | added | BlamKiwi | You're probably right about doing cas instead of test-and-cas. The only way this table will be at high occupancy is if there's back pressure on the hardware completions, which is a rate limiting condition regardless. | |
| Oct 25, 2022 at 12:29 | history | answered | G. Sliepen | CC BY-SA 4.0 |