1. Fix compilation issue on raspberry pi.
2. Minor rework and optimization on balance code path.
3. Fix wrong nr_max_tries in migrate_pending_tasks.
This is mainly a bug fix and minor optimization release for 4.17. The rework of balance code doesn't go well, it actually make more overhead than current implement. Another rework which based on current implement is still on going, hopefully be included in next release.
Enjoy PDS 0.98s for v4.17 kernel, :)
Code are available at
https://github.com/cchalpha/linux-gc/commits/linux-4.17.y-pds
and also
https://gitlab.com/alfredchen/linux-pds
All-in-one patch is available too.
Builds/boots fine for me.
ReplyDeleteI've also switched to pulling commits from the GitLab, thanks for that.
Boots ok here, on Ryzen 1700 and Ryzen 2400G (kid got a new computer :)) Will try on i7@work as well.
ReplyDeleteSince I'm a crazy person, I'll try full tickless system with 0.98s, will report back with some feeling and/or numbers about system behavior.
First results for 0.98s + nohz_full are as follows (took results from previous tests as well):
D3 location 1: PDS (105 FPS), PDS+nohz_full (115 FPS), MUX+MC (110 FPS), MUX+SMT (55 FPS)
D3 location 2: PDS (67 FPS), PDS+nohz_full (68 FPS), MUX+MC (70 FPS), MUX+SMT (40 FPS)
POE location 1: PDS (120 FPS), PDS+nohz_full (130 FPS), MUX+MC (120 FPS), MUX+SMT (120 FPS)
Please take into consideration that latest nohz_full build is 4.17.5 while previous are 4.17.2, dunno if that does make a diff.
Thanks Alfred.
BR, Eduardo
How is it compared to stock kernel (no PDS, no MUX)?
DeleteI have not installed default 4.17.5, but if I'll remember, I'll do that at home in coming days.
DeleteInteresting. For NOHZ_FULL, when I tested it before, there is some accounting issue from htop as far as can be seen, even in mainline scheduler. But in 4.17, there are code changes from mainline, so give it another try, but looks like the accounting issue has been gone. So NOHZ_FULL, I'd like to said NOHZ_FULL works in PDS but it's not fully tested, if you find something unexpected under NOHZ_FULL, pls let me know.
ReplyDeleteAlfred, correct accounting actually was first nice surprise for me, at least so far it looks good to me as well.
DeletePlayed Witcher3 and compiled one more kernel yesterday with nohz_full, all seemed ok, no stutters, good FPS in wine, kid said some of his games perform a little better. We'll see more in coming days.
BRk Eduardo
@Edurado
DeleteOK, thanks for testing NOHZ_FULL and provide feedback. I'd also check the handling code in PDS.
Would you add a results page with benchmarks like : github.com/mfleming/adrestia, hackbench, ARM Lisa or rt-app ?
ReplyDelete