A Last Minute Linux 4.17 Pull To Help Non-PCID Systems With KPTI Meltdown Performance

[latalante]

And, can you offer some 'real' numbers, with and without, please?

I forgot to add in synthetic micro-benchmarks. For this example, about 50% improvement. How does this translate into real applications? Probably very little.
It's a very old laptop (2006) with Core Duo 2 ("first generation").
This test. https://github.com/paboldin/meltdown-exploit
It does not show that it is vulnerable to meltdown.
and spectre

https://gist.github.com/ErikAugust/724d4a969fb2c6ae1bbd7b2a9e3d4bb6#file-spectre-c

Though it may be.

[nuetzel]

GREAT, I'll send some numbers (Xeon X3470, Lynnfield, Nehalem based) when Alex updated amd-staging-drm-next to 4.17-rc1. Currently I'm running with nopti and nospectre_v2 during Mesa radeonsi/radv benchmarking...

[nuetzel]

I've been testing it for a week. Yes, it advances the results in benchmarks.
perf stat --null --sync --repeat 10 perf bench sched messaging -g 20 -t

It is certainly untested on laptops. It broke down hibernation.

And, can you offer some 'real' numbers, with and without, please?

[latalante]

I've been testing it for a week. Yes, it advances the results in benchmarks.
perf stat --null --sync --repeat 10 perf bench sched messaging -g 20 -t

It is certainly untested on laptops. It broke down hibernation.

[phoronix]

Phoronix: A Last Minute Linux 4.17 Pull To Help Non-PCID Systems With KPTI Meltdown Performance

While the Linux 4.17 kernel merge window is closing today and is already carrying a lot of interesting changes as covered by our Linux 4.17 feature overview, Thomas Gleixner today sent in a final round of x86 (K)PTI updates for Meltdown mitigation with this upcoming kernel release...

http://www.phoronix.com/scan.php?page=news_item&px=KPTI-PCID-Global-Pages-4.17