From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ferruh Yigit Subject: Re: AVX512 bug on SkyLake Date: Fri, 9 Nov 2018 09:49:17 +0000 Message-ID: References: <20181023212318.43082-1-yskoh@mellanox.com> <432F92CE-5714-45DC-B72F-CD8771DAFC89@intel.com> <1612642.At0RDolh7h@xps> <9d3f48fc-5a47-c813-1da8-7e1cab6bdd9e@intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: 8bit Cc: Thomas Monjalon , keith.wiles@intel.com, dev , Bruce Richardson , Shahaf Shuler , "Ananyev, Konstantin" , anatoly.burakov@intel.com, stable@dpdk.org, justin.parus@microsoft.com, David Coronel , Josh Powers , Jay Vosburgh , Dan Streetman To: Christian Ehrhardt , yskoh@mellanox.com Return-path: In-Reply-To: Content-Language: en-US List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" On 11/9/2018 6:27 AM, Christian Ehrhardt wrote: > On Fri, Nov 9, 2018 at 12:01 AM Yongseok Koh wrote: >> >> >>> On Nov 8, 2018, at 9:21 AM, Ferruh Yigit wrote: >>> >>> On 11/8/2018 3:59 PM, Thomas Monjalon wrote: >>>> Hi, >>>> >>>> We need to gather more information about this bug. >>>> More below. >>>> > > Thanks Thomas for looping us in! > >>>> 07/11/2018 10:04, Wiles, Keith: >>>>>> On Nov 6, 2018, at 9:30 PM, Yongseok Koh wrote: >>>>>>> On Nov 5, 2018, at 6:06 AM, Wiles, Keith wrote: >>>>>>>> On Nov 2, 2018, at 9:04 PM, Yongseok Koh wrote: >>>>>>>> >>>>>>>> This is a workaround to prevent a crash, which might be caused by >>>>>>>> optimization of newer gcc (7.3.0) on Intel Skylake. >>>>>>> >>>>>>> Should the code below not also test for the gcc version and >>>>>>> the Sky Lake processor, maybe I am wrong but it seems it is >>>>>>> turning AVX512 for all GCC builds >>>>>> >>>>>> I didn't want to check gcc version as 7.3.0 is very new. Only gcc 8 is newly up since then (gcc 8.2). >>>>>> Also, I wasn't able to test every gcc versions and I wanted to be a bit conservative for this crash. >>>>>> Performance drop (if any) by disabling a new (experimental) feature would be less risky than unaccountable crash. >>>>>> And, it does disable the feature only if CONFIG_RTE_ENABLE_AVX512=n. Please refer to v3. >>>>> >>>>> Are you not turning off all of the GCC versions for AVX512. >>>>> And you can test for range or greater then GCC version and >>>>> it just seems like we are turning off every gcc version, is that true? >>>> >>>> Do we know exactly which GCC versions are affected? >>>> >>>>>>> Also bug 97 seems a bit obscure reference, maybe you know >>>>>>> the bug report, but more details would be good? >>>>>> >>>>>> I sent out the report to dev list two month ago. >>>>>> And I created the Bug 97 in order to reference it >>>>>> in the commit message. >>>>>> I didn't want to repeat same message here and there, >>>>>> but it would've been better to have some sort of summary >>>>>> of the Bug, although v3 has a few more words. >>>>>> However, v3 has been merged. >>>>> >>>>> Still this is too obscure if nothing else give a link to >>>>> a specific bug not just 97. >>>> >>>> The URL is >>>> https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fbugs.dpdk.org%2Fshow_bug.cgi%3Fid%3D97&data=02%7C01%7Cyskoh%40mellanox.com%7C90ff6c361faf422b976108d6459eb490%7Ca652971c7d2e4d9ba6a4d149256f461b%7C0%7C0%7C636772945282345908&sdata=2o%2Fg203aWrKCYg16S6oI4BcS41igpLu1DloS%2FrRnknc%3D&reserved=0 >>>> The bug is also pointing to an email: >>>> https://emea01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fmails.dpdk.org%2Farchives%2Fdev%2F2018-September%2F111522.html&data=02%7C01%7Cyskoh%40mellanox.com%7C90ff6c361faf422b976108d6459eb490%7Ca652971c7d2e4d9ba6a4d149256f461b%7C0%7C0%7C636772945282345908&sdata=NCFKxaREd69iZ8eyFKg%2FWBP73CLTXkxrNQQeii%2Bbsao%3D&reserved=0 >>>> >>>> Summary: >>>> - CPU: Intel Skylake >>>> - Linux environment: Ubuntu 18.04 >>>> - Compiler: gcc-7.3 (Ubuntu 7.3.0-16ubuntu3) >>> >>> Is it possible to test a few other gcc versions to check if the issue is >>> specific to this compiler version? >> >> Nothing's impossible but even with my quick search in gcc.gnu.org, >> I could find the following documents mention mavx512f support: >> >> GCC 4.9.0 >> April 22, 2014 (changes, documentation) >> >> GCC 5.1 >> April 22, 2015 (changes, documentation) >> >> GCC 6.4 >> July 4, 2017 (changes, documentation) >> >> GCC 7.1 >> May 2, 2017 (changes, documentation) >> >> GCC 8.1 >> May 2, 2018 (changes, documentation) >> >> We altogether have to put quite large resource to verify all of the versions. >> >> I assumed older than gcc 7 would have the same issue. I know it was a speculation >> but like I mentioned I wanted to be more conservative. I didn't mean this is a permanent fix. >> For two months, we couldn't have any tangible solution (actually nobody cared including myself), >> so I submitted the patch to temporarily disable mavx512f. >> >> I'm still not sure what the best option is... >> > > What I wonder in all of this as I don't understand that part of it yet is this. > I assume you are building on Ubuntu as that is your gcc reference. > FYI: as people asked for bug references, there also is [1] which seems > pretty much the same issue. > > It builds with mostly defaults, that means per > mk/machine/default/rte.vars.mk and similar it sets -march=corei7 > > But when I look at what that implies all avx512 is disabled > $ gcc -Q --help=target -m64 -march=corei7 | grep avx512f > -mavx512f [disabled] This is output is the "corei7" architecture. Defect is for "native" build on Skylake. For "native" arch: gcc -Q --help=target -m64 -march=native | grep avx512f -mavx512f [enabled] Also I can confirm from disassembly output that avx512 instructions are used without '-mno-avx512f' provided. > > So I wonder what/why -mno-avx512f should help at all. > I used the full list of gcc args we have for the build (e.g. [2] of a > 18.05 build), but that doesn't change that (mostly -W, -I and -D). > So I wonder, did people do a custom build and bump up march or enable > -mavx512f on their own to hit that? > Or are we facing a real gcc issue where " -mavx512f [disabled]" is not > the same as -mno-avx512f ? > Maybe someone who hit the bug could clarify that please? > > BTW: per reports I've seen it also seems to apply to the latest > compiler update of the same series - at least it was said to be fully > updated, that would be 7.3.0-27ubuntu1~18.04 > But this is 2nd grade information as I don't have a system with the > right combo MLX5+Skylake available atm, so I can't confirm for sure > :-/ > > [1]: https://bugs.launchpad.net/ubuntu/+source/dpdk/+bug/1799397 > [2]: https://launchpadlibrarian.net/373589345/buildlog_ubuntu-bionic-amd64.dpdk_18.05-1~ubuntu0.18.04.1_BUILDING.txt.gz >