| *** tpb <[email protected]> has joined #openrisc | 00:00 | |
| *** wxwisiasdf <[email protected]> has quit IRC (Quit: leaving) | 02:30 | |
| *** Finde <[email protected]> has quit IRC (Quit: WeeChat 2.3) | 05:36 | |
| *** Finde <[email protected]> has joined #openrisc | 05:36 | |
| zx2c4 | shorne: it seems like memory allocation is really slow | 12:19 |
|---|---|---|
| shorne | zx2c4: thats a good start, the helper_lookup_tb_ptr would be pretty normal it is used a lot for traversing basic blocks in qemu | 17:41 |
| shorne | but I thought there is a way to short circuit that | 17:41 |
| shorne | tlb_flush page will be more interesting as I was hoping that would be reduced, it would be good to see what in the kernel is causing the flushes | 17:42 |
| shorne | that is just perf output it looks like> | 17:42 |
| shorne | ? | 17:42 |
| zx2c4 | Yeah | 17:51 |
| zx2c4 | Just perf top during the slow part of wg tests | 17:52 |
| *** wxwisiasdf <[email protected]> has joined #openrisc | 23:10 | |
| shorne | zx2c4: qemu as a lock profiler (called sync-profiler) | 23:27 |
| shorne | Type Object Call site Wait Time (s) Count Average (us) | 23:27 |
| shorne | -------------------------------------------------------------------------------------------------------- | 23:27 |
| shorne | condvar [ 2] softmmu/cpus.c:423 470.17711 13295 35364.96 | 23:27 |
| shorne | BQL mutex 0x55ee2bf722c0 accel/tcg/cpu-exec.c:681 1.92947 57833060 0.03 | 23:27 |
| shorne | there is one highly contended lock here | 23:27 |
| shorne | Trying to figure out if I can see what "object" is | 23:27 |
| shorne | ... | 23:28 |
| shorne | Type Object Call site Wait Time (s) Count Average (us)(qemu) info sync-profile -n | 23:28 |
| shorne | Type Object Call site Wait Time (s) Count Average (us) | 23:28 |
| shorne | -------------------------------------------------------------------------------------------------------- | 23:28 |
| shorne | condvar 0x55ee2e0c16d0 softmmu/cpus.c:423 95.28372 4287 22226.20 | 23:28 |
| shorne | condvar 0x55ee2e0da640 softmmu/cpus.c:423 60.92801 4126 14766.85 | 23:28 |
| shorne | BQL mutex 0x55ee2bf722c0 accel/tcg/cpu-exec.c:681 1.33362 39048211 0.03 | 23:28 |
| shorne | wxwisiasdf: thanks for your patch, it makes sense, but I am wondering if those attrributes belong in the target implementation, I need to have a look to give better suggestions | 23:29 |
| shorne | ok, object is the "mutex" that is contended | 23:37 |
| shorne | ok, it seems this is just the idle thread, so its just saying 2 CPU's were idle for most of the time | 23:39 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!