Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
On Sun, 14 Jun 2020 08:46:48 +0200 Riccardo Mottola via Dng wrote: > Hi, > > On 6/12/20 9:21 PM, Steve Litt wrote: > >> Maybe try ;https://make-linux-fast-again.com/ kerneloptions to > >> disable all spectre and meltdown stuff > > I hope you mean the preceding as a diagnostic test and not as a > > solution. > > > first, diagnostic on what is slowing down and stealing performance in > such a significant way. > > second, it could be also acceptable as a solution: If a system like > laptop is not running multi-user, no virtual machines and you > consider the software running trusted, where are the risks? I think > only in the browser which runs "unknown" and "untrusted" stuff: The > only attack surface would be JavaScript (or Java, Flash or similar). > If, from what I read, the browser removed high resolution timers, > timing attacks needed to exploit these bugs are unlikely. > > Am I simplifying too much and overseeing some holes? In my opinion, yes. What if a browser didn't remove the high resolution timers? What if other software, currently considered "trusted", acts as a vehicle for the exploit? What if the badguys know more about exploits than you? Making sure you use the right browsers and only trusted apps in the face of a virus like this strikes me like wearing a mask and using plenty of hand cleaner and then hanging around in a meat packing plant. Yeah, you've taken all the precautions you know to take, but you've placed yourself right in the flight path of the virus. There's something else too. The OP measured 32% slowdown on one activity. We don't know whether that kind of performance is common to most activities. Removing protection against known exploits to cure a problem with an unknown cause and an unknown range is the ultimate in proceeding on an untested assumption. Until the root cause is known, no solution should be applied, and diagnostic tests should be performed in a safe manner. SteveT Steve Litt May 2020 featured book: Troubleshooting Techniques of the Successful Technologist http://www.troubleshooters.com/techniques ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
On Sun, Jun 14, 2020 at 08:46:48AM +0200, Riccardo Mottola via Dng wrote: > Hi, > > On 6/12/20 9:21 PM, Steve Litt wrote: > > > Maybe try ;https://make-linux-fast-again.com/ kerneloptions to > > > disable all spectre and meltdown stuff > > I hope you mean the preceding as a diagnostic test and not as a > > solution. > > > first, diagnostic on what is slowing down and stealing performance in such a > significant way. > > second, it could be also acceptable as a solution: If a system like laptop > is not running multi-user, no virtual machines and you consider the software > running trusted, where are the risks? I think only in the browser which runs > "unknown" and "untrusted" stuff: The only attack surface would be JavaScript > (or Java, Flash or similar). If, from what I read, the browser removed high > resolution timers, timing attacks needed to exploit these bugs are unlikely. > > Am I simplifying too much and overseeing some holes? Even if you will not run your system airgapped in production, it might be worth running it airgapped without mitigation as a diagnostic to see if mitigation is really the problem. If it is, you know the situation and can restore mitigation before you connect to the world. If it isn't, you know to look elsewhere. -- hendrik > > > Riccardo > > ___ > Dng mailing list > Dng@lists.dyne.org > https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
Hi, On 6/12/20 9:21 PM, Steve Litt wrote: Maybe try ;https://make-linux-fast-again.com/ kerneloptions to disable all spectre and meltdown stuff I hope you mean the preceding as a diagnostic test and not as a solution. first, diagnostic on what is slowing down and stealing performance in such a significant way. second, it could be also acceptable as a solution: If a system like laptop is not running multi-user, no virtual machines and you consider the software running trusted, where are the risks? I think only in the browser which runs "unknown" and "untrusted" stuff: The only attack surface would be JavaScript (or Java, Flash or similar). If, from what I read, the browser removed high resolution timers, timing attacks needed to exploit these bugs are unlikely. Am I simplifying too much and overseeing some holes? Riccardo ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
Hi all, On 6/13/20 12:24 PM, Patrick Erdmann wrote: Just to be sure, try this line as kernel parameters: noibrs noibpb nopti nospectre_v2 nospectre_v1 l1tf=off nospec_store_bypass_disable no_stf_barrier mds=off tsx=on tsx_async_abort=off mitigations=off I tried all "noibrs noibpb nopti nospectre_v2 nospectre_v1 l1tf=off nospec_store_bypass_disable no_stf_barrier mds=off tsx=on tsx_async_abort=off" and here some results. Laptop with i7 (4x parallel build): ascii "standard": 30-31 minutes beowulf "standard": 41 minutes beowulf "fast": 34 minutes Laptop with T7200 (2x parallel build): ascii "standard": 63 minutes ascii "fast": 61 minutes From this we see that at least some mitigations were already "slowing down" ascii already and at least on another CPU I gained some speed back, but this makes the comparison even with beowulf! We see that "fast" still is slower than previous ascii by about 10% (which we can estimate do be more, since I did not test the same configuration). Something is still slowing down... it is not that my life changes with 3minutes, but I hate to know tht just a plain "upgrade" robbed me 10-15% of speed. By the way, informal tests (not on the same hardware) show that this speed gain is the difference between an i5 and an i7, so like to say my CPU was downgraded :-P What could be slowing down? some other kernel "settings" not related to mitigations.. cache handling, file system, scheduling? Riccardo ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
Hi, On 6/13/20 7:45 PM, Harald Arnesen via Dng wrote: Do you compile ArcticFox with the same compiler in both cases? GCC gets slower for each release, it seems... Yes, same version, gcc 6.5 - the highest I can use to compile. I have issues compiling it with gcc 7 (actually, linking it...). Help fixing it welcome! Also, informal tests have shown that there is no such big difference, probably it depends on the code. Riccardo ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
Patrick Erdmann [13.06.2020 12:24]: > My test case is compiling ArcticFox, thus something in the realm of > Firefox: lots of compiler activity, make, disk access, make and python. > But, of course, predominant factor is compilation and linking. > > With ascii, I was consistently (= not just one build) doing a clean > build in about 31 minutes! Quite fast for this small beast and I was happy. > > With beowulf, this number is consistently about 41 minutes. Do you compile ArcticFox with the same compiler in both cases? GCC gets slower for each release, it seems... -- Hilsen Harald ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
On 12.06.20 13:02, d...@d404.nl wrote: > On 12-06-2020 12:49, Riccardo Mottola via Dng wrote: >> Hi all! >> >> I just upgraded from ascii to beowulf a pretty decent laptop, equipped >> with a core i7 and 8GB of RAM. >> >> I upgraded everything in place: so same desktop environment, same >> applications, same hard disk... just "apt-get dist-upgrade" essentially. >> >> I even kept gcc6 because I had it before and I need it. >> >> My test case is compiling ArcticFox, thus something in the realm of >> Firefox: lots of compiler activity, make, disk access, make and python. >> But, of course, predominant factor is compilation and linking. >> >> With ascii, I was consistently (= not just one build) doing a clean >> build in about 31 minutes! Quite fast for this small beast and I was happy. >> >> With beowulf, this number is consistently about 41 minutes. >> >> I say this is a very significant slowdown! Can I gain some speed back? >> Some setting? some spectre/meltdown mitigation? Having latest ascii, I >> think I had at least some of the backports. >> >> >> I am wary thus updating to beowulf on slower machines. >> >> Any ideas? >> Just to be sure, try this line as kernel parameters: noibrs noibpb nopti nospectre_v2 nospectre_v1 l1tf=off nospec_store_bypass_disable no_stf_barrier mds=off tsx=on tsx_async_abort=off mitigations=off >> >> Riccardo >> ___ >> Dng mailing list >> Dng@lists.dyne.org >> https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng > > According Geekbench > https://browser.geekbench.com/v4/cpu/compare/12415873?baseline=12415673 > removing mitigation mainly makes a difference on single core performance. > > But i recently did see some Intel firmware update passing at my system. > So maybe Ascii has not yet received the same updates as Beowulf. > > Grtz. > > Nick > > > > ___ > Dng mailing list > Dng@lists.dyne.org > https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng > ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
On Fri, Jun 12, 2020 at 03:21:28PM -0400, Steve Litt wrote: > On Fri, 12 Jun 2020 15:44:38 +0200 > Bastiaan van den Berg wrote: > > > Maybe try ; https://make-linux-fast-again.com/ kerneloptions to > > disable all spectre and meltdown stuff > > I hope you mean the preceding as a diagnostic test and not as a > solution. It could be a solution on an air-gapped computer. -- hendrik ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
On Fri, 12 Jun 2020 15:44:38 +0200 Bastiaan van den Berg wrote: > Maybe try ; https://make-linux-fast-again.com/ kerneloptions to > disable all spectre and meltdown stuff I hope you mean the preceding as a diagnostic test and not as a solution. SteveT Steve Litt May 2020 featured book: Troubleshooting Techniques of the Successful Technologist http://www.troubleshooters.com/techniques ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
il devuanizzato Bastiaan van den Berg il 12-06-20 15:44:38
ha scritto:
Maybe try ; https://make-linux-fast-again.com/ kerneloptions to disable
all spectre and meltdown stuff
Experimented slowdown me too. Thanks for reminding me this site.
Joking, risky and written immediately one line substitution:
sed -i "s/GRUB_CMDLINE_LINUX_DEFAULT=\"/GRUB_CMDLINE_LINUX_DEFAULT=\"`wget
https://make-linux-fast-again.com/ -O - 2>/dev/null` /g" /etc/default/grub
--
_
< Viverna >
-
\^/^
\ / \ // \
\ |\___/| / \// .\
\ /0 0 \__ /// | \ \ **
/ / \/_/// | \ \ \ |
@_^_@`/ \/_ //| \ \ \/\ \
//_^_/ \/_ // |\\ \ \
( //) |\/// | \ \ | |
( / /) | // | \ _\ | /
( // /) | ; -.|_ _\.-~ / /
(( / / )) |_ *-.|.-~-. .~~
(( // / ))\ / ~-. _ .-~ /
(( /// )) `. }{ /
(( / )) .~-.\\-` .~
///...<\ _ -~
///-._ _ _ _ _ _ _{^ - - - - ~
___
Dng mailing list
Dng@lists.dyne.org
https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
Maybe try ; https://make-linux-fast-again.com/ kerneloptions to disable all spectre and meltdown stuff ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
Hi, On 12/6/20 8:49 pm, Riccardo Mottola via Dng wrote: > Hi all! > > I just upgraded from ascii to beowulf a pretty decent laptop, equipped > with a core i7 and 8GB of RAM. > > I upgraded everything in place: so same desktop environment, same > applications, same hard disk... just "apt-get dist-upgrade" essentially. > > I even kept gcc6 because I had it before and I need it. > > My test case is compiling ArcticFox, thus something in the realm of > Firefox: lots of compiler activity, make, disk access, make and python. > But, of course, predominant factor is compilation and linking. > > With ascii, I was consistently (= not just one build) doing a clean > build in about 31 minutes! Quite fast for this small beast and I was happy. > > With beowulf, this number is consistently about 41 minutes. > > I say this is a very significant slowdown! Can I gain some speed back? > Some setting? some spectre/meltdown mitigation? Having latest ascii, I > think I had at least some of the backports. > > > I am wary thus updating to beowulf on slower machines. Just an idea, but it might be spectre/meltdown mitigations? A. signature.asc Description: OpenPGP digital signature ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
Re: [DNG] Ascii to Beowulf upgrade - 32% slowdown !
On 12-06-2020 12:49, Riccardo Mottola via Dng wrote: > Hi all! > > I just upgraded from ascii to beowulf a pretty decent laptop, equipped > with a core i7 and 8GB of RAM. > > I upgraded everything in place: so same desktop environment, same > applications, same hard disk... just "apt-get dist-upgrade" essentially. > > I even kept gcc6 because I had it before and I need it. > > My test case is compiling ArcticFox, thus something in the realm of > Firefox: lots of compiler activity, make, disk access, make and python. > But, of course, predominant factor is compilation and linking. > > With ascii, I was consistently (= not just one build) doing a clean > build in about 31 minutes! Quite fast for this small beast and I was happy. > > With beowulf, this number is consistently about 41 minutes. > > I say this is a very significant slowdown! Can I gain some speed back? > Some setting? some spectre/meltdown mitigation? Having latest ascii, I > think I had at least some of the backports. > > > I am wary thus updating to beowulf on slower machines. > > Any ideas? > > > Riccardo > ___ > Dng mailing list > Dng@lists.dyne.org > https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng According Geekbench https://browser.geekbench.com/v4/cpu/compare/12415873?baseline=12415673 removing mitigation mainly makes a difference on single core performance. But i recently did see some Intel firmware update passing at my system. So maybe Ascii has not yet received the same updates as Beowulf. Grtz. Nick signature.asc Description: OpenPGP digital signature ___ Dng mailing list Dng@lists.dyne.org https://mailinglists.dyne.org/cgi-bin/mailman/listinfo/dng
