Re: Retire a package from Fedora i686 (not x86_64)
An update on this: Kevin Kofler wrote: > This is only a minor annoyance where a portable fallback exists (we can > just ship the portable C/C++ version in /usr/lib and the SSE2 version in > /usr/lib/sse2, as we are doing with Qt 5 QtDeclarative). The real problem > is software which has no portable fallback at all, such as > Chromium/QtWebEngine or Darktable. This completely screws not only > non-SSE2 x86, but also all secondary architectures, and in some cases such > as Darktable, even the primary architecture ARM. It is unfortunate that > the number of such non- portable software seems increasing lately. > Portability to any and all CPU architectures used to be a big selling > point for Free Software. These days, more and more projects seem happy to > sacrifice this on the altar of performance. :-( I think I have what it takes to get QtWebEngine working on non-SSE2 now: 1. https://bugzilla.redhat.com/show_bug.cgi?id=1293190 2. https://bugzilla.redhat.com/show_bug.cgi?id=1244196#c24 Thankfully, V8 actually does support non-SSE2 x86: Shortly after support was removed from the "ia32" target, it was readded as a separate "x87" one. So only Chromium/QtWebEngine needs patching (because they started requiring SSE2 in their own code). That said, this only helps 32-bit x86. Most secondary architectures are still screwed, because V8 does not have portable C/C++ fallback code (unlike Chromium's own code). Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org http://lists.fedoraproject.org/admin/lists/devel@lists.fedoraproject.org
Re: Retire a package from Fedora i686 (not x86_64)
Am 12.11.2015 um 21:49 schrieb Przemek Klosowski: On the other hand, Harald, I have to say that in this case my irony detector bent the needle. You often voice your displeasure in this forum that some setup you used for years was broken by new stuff, and yet you say you CAN NOT HAVE both - leading edge software and legacy support forever I just see this quoted back at you in the future :) it's a difference replacing working software with some alpha/beta-quality stuff lacking half of the features and waiting years to get useable as before or just drop the support for hardware older than 10 years for the sake of use capabilities of more recent one signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
A clear example of the mentioned problems for Darktable 32 bit http://redmine.darktable.org/issues/10717 -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/11/2015 03:54 PM, Reindl Harald wrote: Am 11.11.2015 um 21:20 schrieb Fred New: On the other hand, my >10 year-old 32-bit system (Fedora 22) is the only place my LaserJet 4L printer works. My 64-bit systems drop control sequences (like they don't flush the output buffer), leaving the last page of every job sitting in printer memory. I reported the bug a few years ago, but it got closed as WONTFIX and now I can't find it in bugzilla we talk about a *bleeding edge distribution* https://en.wikipedia.org/wiki/HP_LaserJet_4#4L_and_4P "The LaserJet 4 series was discontinued in the 1990s" why in the world does somebody install a operating system released in 2015 to drive harwdare realeased 10-15 years ago and *why in the world* should recent hardware feautures ignored forever because of users with such hardware? This should not be a big problem---the printer used to work, so we have a regression. From Fred's description, it looks like a simple issue, actually. I agree with you that the ball is in Fred's court, though: he's the one that wants to keep his stuff running so at least he should help to diagnose and report the problem (which he did), and follow up by tracking it to some satisfactory resolution (which he didn't). It seems to me that Fred should find, reopen and pursue his bugzilla report. On the other hand, Harald, I have to say that in this case my irony detector bent the needle. You often voice your displeasure in this forum that some setup you used for years was broken by new stuff, and yet you say you CAN NOT HAVE both - leading edge software and legacy support forever I just see this quoted back at you in the future :) -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Sat, Nov 7, 2015 at 6:03 AM, Reindl Haraldwrote: > > > Am 07.11.2015 um 05:00 schrieb Luya Tshimbalanga: > >> On 06/11/15 07:29 PM, Ralf Corsepius wrote: >> >>> But if upstream doesn't care, it's going to be a problem. :-( >>> Exactly - That's the actual problem. Upstream does not care and Fedora >>> seems unable to address this issue. >>> >>> Ralf >>> >>> According to my system, it has SSSE3. Is it SSE3? >> Taken from A10-7400P laptop >> > > any system not older than 10 years has SSE3 and frankly systems older than > 10 years are hardly a traget for Fedora at all > > On the other hand, my >10 year-old 32-bit system (Fedora 22) is the only place my LaserJet 4L printer works. My 64-bit systems drop control sequences (like they don't flush the output buffer), leaving the last page of every job sitting in printer memory. I reported the bug a few years ago, but it got closed as WONTFIX and now I can't find it in bugzilla. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 11.11.2015 um 21:20 schrieb Fred New: On Sat, Nov 7, 2015 at 6:03 AM, Reindl Harald> wrote: Am 07.11.2015 um 05:00 schrieb Luya Tshimbalanga: On 06/11/15 07:29 PM, Ralf Corsepius wrote: But if upstream doesn't care, it's going to be a problem. :-( Exactly - That's the actual problem. Upstream does not care and Fedora seems unable to address this issue. Ralf According to my system, it has SSSE3. Is it SSE3? Taken from A10-7400P laptop any system not older than 10 years has SSE3 and frankly systems older than 10 years are hardly a traget for Fedora at all On the other hand, my >10 year-old 32-bit system (Fedora 22) is the only place my LaserJet 4L printer works. My 64-bit systems drop control sequences (like they don't flush the output buffer), leaving the last page of every job sitting in printer memory. I reported the bug a few years ago, but it got closed as WONTFIX and now I can't find it in bugzilla we talk about a *bleeding edge distribution* https://en.wikipedia.org/wiki/HP_LaserJet_4#4L_and_4P "The LaserJet 4 series was discontinued in the 1990s" why in the world does somebody install a operating system released in 2015 to drive harwdare realeased 10-15 years ago and *why in the world* should recent hardware feautures ignored forever because of users with such hardware? you CAN NOT HAVE both - leading edge software and legacy support forever that's the same disucssion as others have with "my PHP crap don't work with anything above PHP 5.2 so why can't i have newest software *but not* recent PHP signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11 November 2015 at 13:20, Fred Newwrote: > > > On Sat, Nov 7, 2015 at 6:03 AM, Reindl Harald > wrote: >> >> >> >> Am 07.11.2015 um 05:00 schrieb Luya Tshimbalanga: >>> >>> On 06/11/15 07:29 PM, Ralf Corsepius wrote: > > But if upstream doesn't care, it's going to be a problem. :-( Exactly - That's the actual problem. Upstream does not care and Fedora seems unable to address this issue. Ralf >>> According to my system, it has SSSE3. Is it SSE3? >>> Taken from A10-7400P laptop >> >> >> any system not older than 10 years has SSE3 and frankly systems older than >> 10 years are hardly a traget for Fedora at all >> > On the other hand, my >10 year-old 32-bit system (Fedora 22) is the only > place my LaserJet 4L printer works. My 64-bit systems drop control sequences > (like they don't flush the output buffer), leaving the last page of every > job sitting in printer memory. I reported the bug a few years ago, but it > got closed as WONTFIX and now I can't find it in bugzilla. > I don't doubt it is WONTFIX :). That printer is old enough to vote and drink in the United States. Depending on how the printer is connected to the computer you would need either a parallel port debugger or whatever monstrosity is converting USB to RS232 (if you are using serial input).. or SCSI? At a certain point this sort of hardware can only be supported if there are owners who can do the kind of work needed to make it work. Which usually means that at some point you move off of the 'cutting/bleeding' edge distributions to something like Debian or CentOS 5. -- Stephen J Smoogen. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald composed on 2015-11-11 22:10 (UTC+0100): > when you and the other 5 users think they can demand support for ancient > hardware forever because these days cheap hardware is too much cost i > tell you that my lifetime for adpot a ton of regressions cuased by > replacing perfect fine working software with half baken replacements is > *much more* worth because you can buy reasonable hardware, i can't buy > additional lifetime wasted over the last years by braindead software > decisions Must be nice to be affluent. This thread proves you simply don't get that not everyone is in position to buy, buy, buy to replace what ain't broke just because marketers and manufacturers have something ostensibly better to offer. Many, many many have no choice but to make do with whatever they have or can acquire without money. They too want to do the best they can with what they have or can get without money, meaning for many, first choice would not be a yesteryear distro. Compatibililty with more blessed peers or for other reasons may be necessary, and dictate at least some "bleeding edge" software. -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 11.11.2015 um 21:54 schrieb Reindl Harald: Am 11.11.2015 um 21:20 schrieb Fred New: On Sat, Nov 7, 2015 at 6:03 AM, Reindl Harald10 year-old 32-bit system (Fedora 22) is the only place my LaserJet 4L printer works. My 64-bit systems drop control sequences (like they don't flush the output buffer), leaving the last page of every job sitting in printer memory. I reported the bug a few years ago, but it got closed as WONTFIX and now I can't find it in bugzilla we talk about a *bleeding edge distribution* https://en.wikipedia.org/wiki/HP_LaserJet_4#4L_and_4P "The LaserJet 4 series was discontinued in the 1990s" why in the world does somebody install a operating system released in 2015 to drive harwdare realeased 10-15 years ago and *why in the world* should recent hardware feautures ignored forever because of users with such hardware? you CAN NOT HAVE both - leading edge software and legacy support forever that's the same disucssion as others have with "my PHP crap don't work with anything above PHP 5.2 so why can't i have newest software *but not* recent PHP and to make it clear: when you and the other 5 users think they can demand support for ancient hardware forever because these days cheap hardware is too much cost i tell you that my lifetime for adpot a ton of regressions cuased by replacing perfect fine working software with half baken replacements is *much more* worth because you can buy reasonable hardware, i can't buy additional lifetime wasted over the last years by braindead software decisions signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11 November 2015 at 14:10, Reindl Haraldwrote: > > Am 11.11.2015 um 21:54 schrieb Reindl Harald: >> >> >> Am 11.11.2015 um 21:20 schrieb Fred New: >>> >>> On Sat, Nov 7, 2015 at 6:03 AM, Reindl Harald >> any system not older than 10 years has SSE3 and frankly systems >>> older than 10 years are hardly a traget for Fedora at all >>> >>> On the other hand, my >10 year-old 32-bit system (Fedora 22) is the only >>> place my LaserJet 4L printer works. My 64-bit systems drop control >>> sequences (like they don't flush the output buffer), leaving the last >>> page of every job sitting in printer memory. I reported the bug a few >>> years ago, but it got closed as WONTFIX and now I can't find it in >>> bugzilla >> >> >> we talk about a *bleeding edge distribution* >> >> https://en.wikipedia.org/wiki/HP_LaserJet_4#4L_and_4P >> "The LaserJet 4 series was discontinued in the 1990s" >> >> why in the world does somebody install a operating system released in >> 2015 to drive harwdare realeased 10-15 years ago and *why in the world* >> should recent hardware feautures ignored forever because of users with >> such hardware? >> >> you CAN NOT HAVE both - leading edge software and legacy support forever >> >> that's the same disucssion as others have with "my PHP crap don't work >> with anything above PHP 5.2 so why can't i have newest software *but >> not* recent PHP > > > and to make it clear: > > when you and the other 5 users think they can demand support for ancient Harald, he didn't demand anything. He just stated what he was using and why. You need to back the angry sysadmin a bit :). Save it for when someone wants to know why they can't get their Leading Edge i386 to work with Fedora anymore and what the developers are going to do about it or else he will post nasty notes to slashdot. > hardware forever because these days cheap hardware is too much cost i tell > you that my lifetime for adpot a ton of regressions cuased by replacing > perfect fine working software with half baken replacements is *much more* > worth because you can buy reasonable hardware, i can't buy additional > lifetime wasted over the last years by braindead software decisions > > > -- > devel mailing list > devel@lists.fedoraproject.org > https://admin.fedoraproject.org/mailman/listinfo/devel > Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct -- Stephen J Smoogen. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 11.11.2015 um 22:35 schrieb Felix Miata: Reindl Harald composed on 2015-11-11 22:10 (UTC+0100): when you and the other 5 users think they can demand support for ancient hardware forever because these days cheap hardware is too much cost i tell you that my lifetime for adpot a ton of regressions cuased by replacing perfect fine working software with half baken replacements is *much more* worth because you can buy reasonable hardware, i can't buy additional lifetime wasted over the last years by braindead software decisions Must be nice to be affluent. This thread proves you simply don't get that not everyone is in position to buy, buy, buy to replace what ain't broke just because marketers and manufacturers have something ostensibly better to offer. Many, many many have no choice but to make do with whatever they have or can acquire without money. They too want to do the best they can with what they have or can get without money, meaning for many, first choice would not be a yesteryear distro. Compatibililty with more blessed peers or for other reasons may be necessary, and dictate at least some "bleeding edge" software it's not a matter of be affluent if buying every decade a new computer is too uch just don't demand bleeding edge software runs on yours and use a LTS distribution - is that really so hard to understand? signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald composed on 2015-11-11 22:44 (UTC+0100): > Felix Miata composed: >> Must be nice to be affluent. This thread proves you simply don't get that not >> everyone is in position to buy, buy, buy to replace what ain't broke just >> because marketers and manufacturers have something ostensibly better to >> offer. Many, many many have no choice but to make do with whatever they have >> or can acquire without money. They too want to do the best they can with what >> they have or can get without money, meaning for many, first choice would not >> be a yesteryear distro. Compatibililty with more blessed peers or for other >> reasons may be necessary, and dictate at least some "bleeding edge" software > it's not a matter of be affluent Affluence is relative. If you don't have money, you don't buy new stuff, especially to replace working stuff. > if buying every decade a new computer is too uch just don't demand What's magical about a decade? Nothing. It's arbitrary. People last more than a decade. Cars last more than a decade. Houses last more than a decade. Refrigerators last more than a decade. No good reason a PC can't last more than a decade before polluting a landfill. > bleeding edge software runs on yours and use a LTS distribution - is > that really so hard to understand? You seem unable to understand that there are far more than a few easily classifiable use cases, or that bleeding edge doesn't imply being first to tell computer users (or let them find out the hard way) theirs are too old for continuing to use the distro they've been using and are used to and/or prefer and/or shared with those they need or want to maintain maximum compatibility with. -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 10.11.2015 um 00:52 schrieb Alexander Ploumistos: On Tue, Nov 10, 2015 at 1:34 AM, Reindl Haraldwrote: about money: with the energy costs i saved going away from P4 computers many years ago my current one is already paied Nobody is saying that a newer computer won't save you money on the power bill. What they have repeatedly stated and for some (first world, perhaps?) reason you refuse to acknowledge, is that you can't trade future power use for a new computer; you have to pay up front, it's a sort of investment, that not everyone is affluent enough to make. Not every person or institution can afford to get a new Phenom or Xeon or whatever every year, or even every X years. What exactly do you suggest these people do? Get LFS, Gentoo or Arch and clear all incompatible flags? Or should they apply for a loan to their bank, detailing the ROI from reduced power usage? i would suggest NOT use a fast moving distribution like Fedora for such antique hardware why do you ask again when i statet *more than once* that i have zero understaind for combine more than 10 years old hardware with a bleeding-edge distribution and then demand the bleeding-edge distribution to support that hardware the next 20 years instead move forward signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 2015-11-09, Reindl Haraldwrote: > server spin "maybe you are using the wrong operating system" and frankly > for schools you typically use thin-cients one machine > Unfortunatelly recent desktop environments require OpenGL. Recent web pages require hardware-accelareted video decoding. Not so recent widget toolkits started to use pixmaps instead of X server-side rendering. Users want to use sound output and USB devices. This makes thin clients really hard to implement. Especially if all the network-oriented features Linux deskop had are becoming obsolete. What could work is a diskless thick client. But these old computers usually have 100-Mbps NIC. That's 5 times slower than local ATA disk. This does not matter when everything is loaded, but it hurts when starting programs. Especially in classes where all students log in and to the same stuff at the same time. -- Petr -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Please leave this thread alone. Since numerous days, nothing about the original subject has been added. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Kalev Lember wrote: > I wonder if it might be a good time to bump the i686 minimal > requirements for F24 to include sse2. Not for performance reasons, but > for compatibility: almost all developers are on x86_64 these days and > i686 is pretty much just limping along. If we include sse2 on i686, that > removes another difference between x86_64 and i686 and makes things > easier for us as a downstream. I don't think this difference is a real issue. > Apparently quite a few upstreams these days consider sse2 as a minimal > requirement and it seems more and more that Fedora packagers need to > work this around if we don't follow the lead. This is more of a problem. This is only a minor annoyance where a portable fallback exists (we can just ship the portable C/C++ version in /usr/lib and the SSE2 version in /usr/lib/sse2, as we are doing with Qt 5 QtDeclarative). The real problem is software which has no portable fallback at all, such as Chromium/QtWebEngine or Darktable. This completely screws not only non-SSE2 x86, but also all secondary architectures, and in some cases such as Darktable, even the primary architecture ARM. It is unfortunate that the number of such non- portable software seems increasing lately. Portability to any and all CPU architectures used to be a big selling point for Free Software. These days, more and more projects seem happy to sacrifice this on the altar of performance. :-( I think dropping support for non-SSE2 in all of Fedora would send entirely the wrong message to those upstreams and only contribute to the surge of non-portable software. That said, we need a formal policy to deal with such software, which IMHO should be treated just like an ExcludeArch, except that you should not actually ExcludeArch the package as long as it supports SOME CPUs of that architecture. It should still be put on the relevant ExcludeArch tracker though, and if a fix/workaround exists, applying it should be required. > I don't think this is worth doing for performance reasons. I don't think > a few percentage of possible gain in micro benchmarks is worth it, but > if it saves some hundreds of hours of developer time for people who are > working on Fedora and don't have to work around missing sse2, I'd say > it's very well worth the cost of losing ancient CPU support. That would unfortunately decrease the usefulness of the i686 distribution a lot. Given that we want to push users of 64-bit-capable machines to the x86_64 version, it would leave basically only one generation of x86 machines for the i686 distribution: the Pentium 4 Northwood generation, including one or two generation(s) of AMD CPUs from the same time. Add to that some newish low-end Atoms where they removed x86_64 support "to save power" (actually to introduce an artificial market segmentation), no idea how popular those are. Everything else that has SSE2 also has 64-bit. Unfortunately, that prerogative ("all 64-bit-capable machines should use 64- bit binaries") is not universally shared by upstreams. For example, Qt sees the primary target of their 32-binaries as being modern machines that would also support x86_64, but where the user elected to use 32-bit for whatever reason, not old 32-bit-only machines as we see it. That is the primary reason for the different stance on non-SSE2 support. (That said, Qt, outside of QtWebEngine, does not actually REQUIRE SSE2. They just enable its use by default.) My argument that the default option ought to be the safe, portable one (non-SSE2) was rejected on the grounds of performance on what they think are the vast majority of machines using 32-bit Qt (whereas I argued that it was only really one generation, but as I wrote above, they have a completely different conception of the purpose of 32-bit binaries). Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Rex Dieter wrote: > Qt5 (Gui and Declarative) pretty much requires sse2, and I/kde-sig asked > about pushing the minimum i686 fedora spec to include sse2, but fesco was > against that idea at the time. For what it's worth, I was against it. Using /usr/lib/sse2 as we do now works just fine as a solution to this problem. People without SSE2 at least get a distribution that works for them, and even Qt 5 builds (except the soon-to-be-packaged QtWebEngine module, sadly, unless/until something happens at its upstream to fix the V8 non-portability) that work, albeit slowly. LXQt should even work at an acceptable speed, given that they use QtWidgets for their UI, which does not need the QML 2 interpreter or JIT (so you don't get the non-JIT interpreter). And people with SSE2 get SSE2- optimized .so binaries where it matters. The only thing the duplication costs is some disk space, but that is the price of portability. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 09.11.2015 um 22:31 schrieb Samuel Sieb: On 11/07/2015 10:23 AM, Reindl Harald wrote: these days (and i talk about hardware from 2011) you can virtualize things fast, easy and efficient and consolidate machines on more or less cheap hardware For servers, sure, but that doesn't work for desktops. how many physical desktops does one use? talking about a rapid moving distribution like Fedora and at the same time hestitate to use technology available for a decade is strange - i guess people plan to run their hardware for 10, 12, 15 years are not using typically a distribution like Fedora I manage all the IT for a private school. There is a computer lab for the students to use. They don't have the money to buy new computers, the computers they have were donated. They are P4 desktops from around 2005 and they have SSE2. I install Fedora on all the computers and laptops there that I manage. (The older students manage their own laptops, but even some of them want Fedora installed as well.) I upgraded all the P4 desktops to 2GB RAM and with F22 using a non-3D WM they work perfectly for student use. So yes, I'm using 10 year old hardware with the latest Fedora and it works great. (I'll update them to F23 next time I'm there.) SSE3 was introduced in 2004 well, i tell you know the same which was told me all the time before the server spin "maybe you are using the wrong operating system" and frankly for schools you typically use thin-cients one machine with the agrumentations of this thread hardware manufactuers could stop develop new hardware capabilities if we wait 20 years to consider make them default - in the IT 10 years is virtually forever and *yes* capable hardware these days is cheap if you don't need that much performance signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/07/2015 10:23 AM, Reindl Harald wrote: these days (and i talk about hardware from 2011) you can virtualize things fast, easy and efficient and consolidate machines on more or less cheap hardware For servers, sure, but that doesn't work for desktops. talking about a rapid moving distribution like Fedora and at the same time hestitate to use technology available for a decade is strange - i guess people plan to run their hardware for 10, 12, 15 years are not using typically a distribution like Fedora I manage all the IT for a private school. There is a computer lab for the students to use. They don't have the money to buy new computers, the computers they have were donated. They are P4 desktops from around 2005 and they have SSE2. I install Fedora on all the computers and laptops there that I manage. (The older students manage their own laptops, but even some of them want Fedora installed as well.) I upgraded all the P4 desktops to 2GB RAM and with F22 using a non-3D WM they work perfectly for student use. So yes, I'm using 10 year old hardware with the latest Fedora and it works great. (I'll update them to F23 next time I'm there.) -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 09.11.2015 um 23:57 schrieb Samuel Sieb: On 11/09/2015 02:22 PM, Reindl Harald wrote: Am 09.11.2015 um 22:31 schrieb Samuel Sieb: So yes, I'm using 10 year old hardware with the latest Fedora and it works great. (I'll update them to F23 next time I'm there.) well, i tell you know the same which was told me all the time before the server spin "maybe you are using the wrong operating system" and frankly for schools you typically use thin-cients one machine I really wonder where you come from. I have never seen a grade school, public or private, that uses thin-clients http://www8.hp.com/us/en/thin-clients/education.html As for "wrong operating system", do you have any suggestions for a different distribution that is as up-to-date as Fedora that would work on these computers? i don't get why i need the most up-to-date software when at the same time i don't care about my hardware and run a P4 heater because that indicates low requirements about money: with the energy costs i saved going away from P4 computers many years ago my current one is already paied I've always run Fedora on my servers as well me too signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/09/2015 02:22 PM, Reindl Harald wrote: Am 09.11.2015 um 22:31 schrieb Samuel Sieb: I manage all the IT for a private school. There is a computer lab for the students to use. They don't have the money to buy new computers, the computers they have were donated. They are P4 desktops from around 2005 and they have SSE2. I install Fedora on all the computers and laptops there that I manage. (The older students manage their own laptops, but even some of them want Fedora installed as well.) I upgraded all the P4 desktops to 2GB RAM and with F22 using a non-3D WM they work perfectly for student use. So yes, I'm using 10 year old hardware with the latest Fedora and it works great. (I'll update them to F23 next time I'm there.) well, i tell you know the same which was told me all the time before the server spin "maybe you are using the wrong operating system" and frankly for schools you typically use thin-cients one machine I really wonder where you come from. I have never seen a grade school, public or private, that uses thin-clients. As for "wrong operating system", do you have any suggestions for a different distribution that is as up-to-date as Fedora that would work on these computers? I've always run Fedora on my servers as well. with the agrumentations of this thread hardware manufactuers could stop develop new hardware capabilities if we wait 20 years to consider make them default - in the IT 10 years is virtually forever and *yes* capable hardware these days is cheap if you don't need that much performance "Cheap"? I might be able to able to build a computer for $300, I haven't actually checked lately, that's what it was a few years ago. But for a computer lab? Why would they want to try to come up with that much money when the existing computers work perfectly? -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald composed on 2015-11-09 23:22 (UTC+0100): > *yes* capable hardware these days is cheap if you don't need that much > performance Unless and until hardware prices drop to zero, there will be users for whom any cost is an insurmountable hurdle. And, there will remain users content with existing capability who only wish to upgrade software in order to maintain security and/or access to software demanded by their needs. Ordinary users don't care when SSE3 was introduced. They just want to keep using what ain't broke as long as possible. Many don't see justification for capability of newer hardware causing software to demand that capability either, particularly when it breaks arch portability or produces no more than a modest, or no, detectable performance improvement, among them, me. -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Tue, Nov 10, 2015 at 1:34 AM, Reindl Haraldwrote: > about money: with the energy costs i saved going away from P4 computers many > years ago my current one is already paied Nobody is saying that a newer computer won't save you money on the power bill. What they have repeatedly stated and for some (first world, perhaps?) reason you refuse to acknowledge, is that you can't trade future power use for a new computer; you have to pay up front, it's a sort of investment, that not everyone is affluent enough to make. Not every person or institution can afford to get a new Phenom or Xeon or whatever every year, or even every X years. What exactly do you suggest these people do? Get LFS, Gentoo or Arch and clear all incompatible flags? Or should they apply for a loan to their bank, detailing the ROI from reduced power usage? -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/09/2015 03:34 PM, Reindl Harald wrote: Am 09.11.2015 um 23:57 schrieb Samuel Sieb: On 11/09/2015 02:22 PM, Reindl Harald wrote: Am 09.11.2015 um 22:31 schrieb Samuel Sieb: So yes, I'm using 10 year old hardware with the latest Fedora and it works great. (I'll update them to F23 next time I'm there.) well, i tell you know the same which was told me all the time before the server spin "maybe you are using the wrong operating system" and frankly for schools you typically use thin-cients one machine I really wonder where you come from. I have never seen a grade school, public or private, that uses thin-clients http://www8.hp.com/us/en/thin-clients/education.html I didn't say they didn't exist. I've certainly heard of them and seen them in various places. I said in all the schools I've been in, I've never seen any actually deployed. As for "wrong operating system", do you have any suggestions for a different distribution that is as up-to-date as Fedora that would work on these computers? i don't get why i need the most up-to-date software when at the same time i don't care about my hardware and run a P4 heater because that indicates low requirements What does low requirements have to do with up-to-date software? Why should I have to use an old version of LibreOffice just because I'm using a 10 year old computer when the latest version works perfectly well? about money: with the energy costs i saved going away from P4 computers many years ago my current one is already paied I would be very surprised if the power savings are really that much. I will do a power check the next time I'm on site. Besides that, electricity is cheap here so it would take a very long time to recoup the cost of new computers. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 08.11.2015 um 11:56 schrieb drago01: On Sun, Nov 8, 2015 at 10:35 AM, Reindl Haraldwrote: There is no such thing as "the complete distribution" ... the flags used to compile libreoffice won't affect http or firefox surely, there is such a thing on a typical machine is running way more than a single application, you have (growing) background-noise, currently plasma-desktop eats 25-30% of a core, X eats some percent and so on That sounds like a bug. Throwing in some compiler flags won't fix that. Fixing that requires 1) finding out what it is doing (profile and/or get stack traces) 2) fix the code software is full of bugs (KDE5 is overwhelmed) all the time and you missed the point because "used to compile libreoffice won't affect http or firefox" is simply wrong on a MULTITASKING system sharing one CPU and so each application will affect others at the end of the day signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 08.11.2015 um 09:47 schrieb drago01: On Sat, Nov 7, 2015 at 9:24 PM, Reindl Haraldwrote: Am 07.11.2015 um 20:36 schrieb drago01: On Sat, Nov 7, 2015 at 7:40 PM, Reindl Harald wrote: the point is compile a single application with new features won't gain that muc 8until you do the same with most libraries used by the software) but having the whole distribution is a summary with a completly different behavior than a single test of software xyz Uh no you just have to compile the software any libraries used by the hot paths of the software you are trying to test. for which one? that don't show the impact on a complete distribution anyways There is no such thing as "the complete distribution" ... the flags used to compile libreoffice won't affect http or firefox surely, there is such a thing on a typical machine is running way more than a single application, you have (growing) background-noise, currently plasma-desktop eats 25-30% of a core, X eats some percent and so on htop shows 7-10% i7-3770 CPU @ 3.40GHz by doing *nothing* expect write that mail and listen to a MP3 file in the background the distribution flags are going to be interesting when each of this pieces is more efficient while you are doing CPU intense tasks also working more efficient signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Felix Miata wrote: > For those without > SSE2, Plasma 5 is completely unusable, but OK with any of the lightweigh . Qt5 (Gui and Declarative) pretty much requires sse2, and I/kde-sig asked about pushing the minimum i686 fedora spec to include sse2, but fesco was against that idea at the time. -- Rex -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/08/2015 10:17 PM, Rex Dieter wrote: > Felix Miata wrote: > >> For those without >> SSE2, Plasma 5 is completely unusable, but OK with any of the lightweigh > > . > > Qt5 (Gui and Declarative) pretty much requires sse2, and I/kde-sig asked > about pushing the minimum i686 fedora spec to include sse2, but fesco was > against that idea at the time. I wonder if it might be a good time to bump the i686 minimal requirements for F24 to include sse2. Not for performance reasons, but for compatibility: almost all developers are on x86_64 these days and i686 is pretty much just limping along. If we include sse2 on i686, that removes another difference between x86_64 and i686 and makes things easier for us as a downstream. Apparently quite a few upstreams these days consider sse2 as a minimal requirement and it seems more and more that Fedora packagers need to work this around if we don't follow the lead. I don't think this is worth doing for performance reasons. I don't think a few percentage of possible gain in micro benchmarks is worth it, but if it saves some hundreds of hours of developer time for people who are working on Fedora and don't have to work around missing sse2, I'd say it's very well worth the cost of losing ancient CPU support. -- Kalev -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Sun, Nov 8, 2015 at 10:20 PM, Reindl Haraldwrote: > > > Am 08.11.2015 um 22:17 schrieb Rex Dieter: >> >> Felix Miata wrote: >> >>> For those without >>> SSE2, Plasma 5 is completely unusable, but OK with any of the lightweigh >> >> >> . >> >> Qt5 (Gui and Declarative) pretty much requires sse2, and I/kde-sig asked >> about pushing the minimum i686 fedora spec to include sse2, but fesco was >> against that idea at the time > > > that was years ago, -msse2 is now the default No as I said before its not. (Note the "i686"). On x86_64 you don't have to pass -msee2. SSE2 is part of the x86_64 ABI. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 08.11.2015 um 22:21 schrieb drago01: On Sun, Nov 8, 2015 at 10:20 PM, Reindl Haraldwrote: Am 08.11.2015 um 22:17 schrieb Rex Dieter: Felix Miata wrote: For those without SSE2, Plasma 5 is completely unusable, but OK with any of the lightweigh Qt5 (Gui and Declarative) pretty much requires sse2, and I/kde-sig asked about pushing the minimum i686 fedora spec to include sse2, but fesco was against that idea at the time that was years ago, -msse2 is now the default No as I said before its not. (Note the "i686"). On x86_64 you don't have to pass -msee2. SSE2 is part of the x86_64 ABI indeed, the defaults even don't contain -msse2 in 2015 ridiculous [builduser@buildserver64:~]$ cat /usr/lib/rpm/rpmrc | grep i686 optflags: i386 -O2 -g -march=i386 -mtune=i686 signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 08.11.2015 um 22:17 schrieb Rex Dieter: Felix Miata wrote: For those without SSE2, Plasma 5 is completely unusable, but OK with any of the lightweigh . Qt5 (Gui and Declarative) pretty much requires sse2, and I/kde-sig asked about pushing the minimum i686 fedora spec to include sse2, but fesco was against that idea at the time that was years ago, -msse2 is now the default that thread is about -msse3 available for a full decade signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Sun, Nov 8, 2015 at 10:35 AM, Reindl Haraldwrote: > > > Am 08.11.2015 um 09:47 schrieb drago01: >> >> On Sat, Nov 7, 2015 at 9:24 PM, Reindl Harald >> wrote: >>> >>> Am 07.11.2015 um 20:36 schrieb drago01: On Sat, Nov 7, 2015 at 7:40 PM, Reindl Harald wrote: > > > the point is compile a single application with new features won't gain > that > muc 8until you do the same with most libraries used by the software) > but > having the whole distribution is a summary with a completly different > behavior than a single test of software xyz Uh no you just have to compile the software any libraries used by the hot paths of the software you are trying to test. >>> >>> >>> >>> for which one? >>> that don't show the impact on a complete distribution anyways >> >> >> There is no such thing as "the complete distribution" ... the flags >> used to compile libreoffice won't affect http or firefox > > > surely, there is such a thing > > on a typical machine is running way more than a single application, you have > (growing) background-noise, currently plasma-desktop eats 25-30% of a core, > X eats some percent and so on That sounds like a bug. Throwing in some compiler flags won't fix that. Fixing that requires 1) finding out what it is doing (profile and/or get stack traces) 2) fix the code. > htop shows 7-10% i7-3770 CPU @ 3.40GHz by doing *nothing* expect write that > mail and listen to a MP3 file in the background The mp3 playback is already making use of SIMD instructions. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Sat, Nov 7, 2015 at 9:24 PM, Reindl Haraldwrote: > > > Am 07.11.2015 um 20:36 schrieb drago01: >> >> On Sat, Nov 7, 2015 at 7:40 PM, Reindl Harald >> wrote: >>> >>> the point is compile a single application with new features won#t gain >>> that >>> muc 8until you do the same with most libraries used by the software) but >>> having the whole distribution is a summary with a completly different >>> behavior than a single test of software xyz >> >> >> Uh no you just have to compile the software any libraries used by the >> hot paths of the software you are trying to test. > > > for which one? > that don't show the impact on a complete distribution anyways There is no such thing as "the complete distribution" ... the flags used to compile libreoffice won't affect http or firefox. >>> if there would be no difference kernel upstream won't invest that much >>> time >>> for runtime-cpu-detection (look at the bootlog on different hardware) >>> >>> here are examples >>> http://www.phoronix.com/scan.php?page=news_item=MTU0MTY >>> http://www.phoronix.com/scan.php?page=news_item=MTI1Njc >> >> >> You missed the point completely. >> >> 1. AVX != (S)SSE3 > > > i know that Then don't cite unrelated links. >> There are lots of workloads that would benefit from AVX but SSE3 >> doesn't add that much (enabling SSE2 on i686 would gain you more for >> instance). > > > enable SSE3 don't diable SSE2 > so "gain you more" is impossible I meant "i686 + SSE2" (which we don't do for the same reason) gains more than "x86_64 + sse3". >> 2. Runtime detection does not have the cost of dropping support for >> specific hardware > > > runtime detection needs to be implemented in every relevant software at it's > own and hence has a *high cost* Not really its a common practice for software that does matter; even glibc does it. > add -msse3 to the default falgs have no costs at all but only an impact of a > unknown amount f *very* outdated hardware which is unlikely running a > bleeding edge distribution As you just wrote in that very paragraph the cost is dropping support for a set of hardware. We have no numbers on how much if at all it would help. > frankly whatever somebody has run on a 10 years old machine can be easily > virtualized and i doubt that many people have a 10 years old computer as > their only device, as far there is something with a core2 or newer in the > house you can virtualize the other machine and save a lot of energy Depends on the device. Virtualizing a laptop for instance won't make much sense. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
A few lines of IRC chat, Freenode #darktable. Hanatos is Darktable project founder. [09:17] ``requiring SSE3 is not really allowed '' [09:17] so much bundled cluelessness :/ [09:19] Germano: re 32-bit [09:19] the sse thing is one thing [09:20] the other is the very limited virtual address space (2G really) [09:20] everybody coding anything half way serious will tell you the same story [09:20] (rawtherapee has the same issues iirc) [09:20] our old cache was allocing one big chunk of memory at startup and maintained it manually [09:20] essentially duplicating a poor man's malloc, specialised for our thumbnail caches [09:21] the new cache is much faster and easier to read [09:21] but based on malloc/free [09:21] which means your virtual address space (not the physical one mapping to your ram) [09:21] will get fragmented and you quickly start addressing blocks above the 10G range [09:22] which may not be a problem, even on systems with only 2G of physical ram, because blocks have been freed in between. it's just on 32-bit systems you can't address it any more and die [09:22] basically, at this point, DT makes no sense on x86, except maybe dt-cli [09:22] which is a similar argument as the sse3 is. [09:22] it's just not a worthwhile experience running this software on this kind of hardware [09:22] boucman: yes, that. [09:23] hanatos_: I think jmalloc is also more clever than glibc malloc wrt fragmentation. that's why blender uses it, afaik [09:24] *je [09:24] Germano: so i'd like to contradict the `upstream doesn't care' bit [09:25] upstream does care. [09:25] just not about random principles and guidelines [09:25] but about how well darktable runs [09:25] Artefact2: jemalloc you mean [09:25] hanatos_: yes [09:25] yes, it's mostly multithreaded/ [09:25] block per thread [09:25] might be worthwhile when running many threads for thumbnail gen [09:25] but honestly i doubt it [09:26] it speeds up another piece of code i wrote [09:26] which uses many 10s of 1000s of malloc calls per second.. [09:26] we don't do that in dt [09:26] (or tcmalloc for that matter) [09:26] simple enough to try with an LD_PRELOAD [09:26] oh yeah. calling malloc this many times is a bad idea anyway [09:32] the alternative would have been allocate ridiculous amounts of memory up front [09:32] bad idea, too [09:32] but if you have a better solution i'd sure like to hear it :) [09:32] the problem is to construct a binary search tree [09:32] i'm not a memory guru, sadly :| [09:32] in parallel [09:32] so you start at the root and push the children as new jobs (malloc job_t) [09:32] and so on [09:33] it's millions of nodes total, so you don't want to allocate them up front [09:33] maybe a compromise. allocate a pool that can store, say 10 jobs at a time [09:34] (and yes, i would agree.. calling malloc is almost always a bad idea, unless you can't avoid it) [09:34] but see.. that pool per thread.. that's exactly what jemalloc/tcmalloc do [09:34] this way you reduce the allocator load by a factor of 10, while still not allocating huge amounts of contiguous memory [09:35] maybe the issue is elsewhere. what are you doing millions of? is it possible to make "bigger" jobs and have less of them? ie a smaller tree [09:38] nope, can't touch the tree [09:38] its some spatial acceleration structure for ray tracing [09:39] it's been optimised for fast ray tracing for many years [09:39] are we still talking about darktable? didn't know it needed a raytracer [09:40] no, different piece of code.. as i said, i don't think darktable needs thread-cached malloc [10:11] Germano: also feel free to refer those guys here to us if they have questions. seems to me that some direct contact may be better. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 16:01 schrieb Felix Miata: Kevin Kofler composed on 2015-11-07 14:05 (UTC+0100): Reindl Harald wrote: come on and don't tell me 99% of i686 users have machines older than 10 years https://en.wikipedia.org/wiki/SSE3 because "ntel introduced SSE3 in early 2004 with the Prescott revision of their Pentium 4 CPU" Any demarcation that is calendar based is purely arbitrary. 64 bit was introduced far more than 10 years ago. There is plenty of 32 bit hardware perfectly capable of doing what needs doing regardless of age. Not everyone needs, or wants, the bloat that newer enables. Not everyone is "speed" sensitive, while most are sensitive to the security that keeping current provides and these are the Fedora relevant users? that machines are fare better served with CentOS (now there is even a i686 one while RHEL7 does not exist of 32bit) instead bleeding edge while i don't see a compelling reason that a bleeding edge distribution in 2015 ignores SSE3 while in the meantime SSE4.2/AVX/AVX2 where introduced for a handful of users "I have approximately 40 Fedora installations on 32 bit CPUs" by one real computer and put them all inside virtual machines on top of it - problem solved - the power savings alone wil pay it! our new DL380 has two Xeon(R) CPU E5-2643 v3 @ 3.40GHz, 128 GB RAM, runs the whole company 8 with a second node for failover) and consumes 115W power (average over 24 hours) signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald composed on 2015-11-07 16:12 (UTC+0100): > Felix Miata composed: >> Kevin Kofler composed on 2015-11-07 14:05 (UTC+0100): >>> Reindl Harald wrote: come on and don't tell me 99% of i686 users have machines older than 10 years https://en.wikipedia.org/wiki/SSE3 because "ntel introduced SSE3 in early 2004 with the Prescott revision of their Pentium 4 CPU" >> Any demarcation that is calendar based is purely arbitrary. 64 bit was >> introduced far more than 10 years ago. There is plenty of 32 bit hardware >> perfectly capable of doing what needs doing regardless of age. Not everyone >> needs, or wants, the bloat that newer enables. Not everyone is "speed" >> sensitive, while most are sensitive to the security that keeping current >> provides > and these are the Fedora relevant users? Surely there are people who need relativly recent or current versions of software other than web browsers who aren't concerned with speed or replacing hardware at an arbitrary chronological age. Photographers using Darktable may well be among them. Under any given roof, it makes sense to keep all machines on the same OS when practical. Who's to say a whole family or business should be using CentOS because one machine is past its prime? -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Ralf Corsepius wrote: > Exactly - That's the actual problem. Upstream does not care and Fedora > seems unable to address this issue. I think the only way to really fix this package would be to track down and add patches reverting all commits like these: http://redmine.darktable.org/projects/darktable/repository/revisions/a8f31a96385564303d98215f2c67313e69798bdd http://redmine.darktable.org/projects/darktable/repository/revisions/2807ac99c0f415a174f6b6444cc7f4f154f7a108 … Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Kevin Kofler composed on 2015-11-07 14:05 (UTC+0100): > Reindl Harald wrote: >> come on and don't tell me 99% of i686 users have machines older than 10 >> years https://en.wikipedia.org/wiki/SSE3 because "ntel introduced SSE3 >> in early 2004 with the Prescott revision of their Pentium 4 CPU" Any demarcation that is calendar based is purely arbitrary. 64 bit was introduced far more than 10 years ago. There is plenty of 32 bit hardware perfectly capable of doing what needs doing regardless of age. Not everyone needs, or wants, the bloat that newer enables. Not everyone is "speed" sensitive, while most are sensitive to the security that keeping current provides. FWIW, Prescott introduction roughly coincides with PCIe video slots replacing AGP slots. Any motherboard with AGP likely is incapable of having a 64 bit Intel CPU installed and supported by its BIOS. >> those machines would not be able to run a recent Fedora due all the >> bloat introduced over the years - try a system update with 512 MB and >> shake your hands with the OOM-killer > I have a computer without SSE3. It runs Fedora just fine. I initially had 1 > GiB of RAM in it, I upgraded it to 3 GiB. I have approximately 40 Fedora installations on 32 bit CPUs. About half of those are on Prescotts. I'm unable to notice that those with more than 1G RAM run any better than those with more RAM. I am able to notice that the older, slower ones that do have SSE2 are able to run, but with Plasma 5, those are very slow. Those do run respectably with lighter WMs/DEs. For those without SSE2, Plasma 5 is completely unusable, but OK with any of the lightweights. -- "The wise are known for their understanding, and pleasant words are persuasive." Proverbs 16:21 (New Living Translation) Team OS/2 ** Reg. Linux User #211409 ** a11y rocks! Felix Miata *** http://fm.no-ip.com/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 16:40 schrieb Felix Miata: Reindl Harald composed on 2015-11-07 16:12 (UTC+0100): Felix Miata composed: Any demarcation that is calendar based is purely arbitrary. 64 bit was introduced far more than 10 years ago. There is plenty of 32 bit hardware perfectly capable of doing what needs doing regardless of age. Not everyone needs, or wants, the bloat that newer enables. Not everyone is "speed" sensitive, while most are sensitive to the security that keeping current provides and these are the Fedora relevant users? Surely there are people who need relativly recent or current versions of software other than web browsers who aren't concerned with speed or replacing hardware at an arbitrary chronological age. Photographers using Darktable may well be among them. Under any given roof, it makes sense to keep all machines on the same OS when practical. Who's to say a whole family or business should be using CentOS because one machine is past its prime? well, who's to say that we stay forever on a level of CPU feature-support while there are instruction sets available for a whole decade which improve performance, save power in case you need fewer instructions doing the same work? not that i say Fedora should go ahead and build with -mavx but a discussion about SSE3 in 2015 is really odd signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 05:41 schrieb Ralf Corsepius: On 11/07/2015 05:03 AM, Reindl Harald wrote: Am 07.11.2015 um 05:00 schrieb Luya Tshimbalanga: On 06/11/15 07:29 PM, Ralf Corsepius wrote: But if upstream doesn't care, it's going to be a problem. :-( Exactly - That's the actual problem. Upstream does not care and Fedora seems unable to address this issue. According to my system, it has SSSE3. Is it SSE3? That's what my question actually was about. https://en.wikipedia.org/wiki/SSE3#CPUs_with_SSE3 I see ssse3 on all my x86_64ers, I see sse3 on my 32bit atoms, but I don't know if ssse3 implies sse3 and I don't know what -msse3 actually does to the instruction set when being used on x86_64ers and which impact it has https://en.wikipedia.org/wiki/SSE3#New_instructions (-msse3 is not part of Fedora's gcc implicit CFLAGS). and that is really sad because it wastes for 99.9% of users performance and hence i override the builds of perormance relevant software since years with optimized falgs while hat -fstack-protector-all in use long before Fedora supported and switched to -fstack-protector-strong in other words: you get the small performance impact of hardening back by better utilize your existing hardware instead castrate your whole software to nostalgic decades any system not older than 10 years has SSE3 and frankly systems older than 10 years are hardly a traget for Fedora at all So, Fedora or Linux as a resort to escape Windows on old HW is not a Fedora target anymore? come on and install a recent Fedora on such old machines, guess what SSE3 or not is your smallest problem signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Germano Massullo wrote: > [09:22] basically, at this point, DT makes no sense on x86, > except maybe dt-cli Well, I think that, in the absence of a real portability fix, building the package (with SSE3, or in general the minimum instruction set it supports) is better than ExcludeArch. The goal of the portability policies is to make our packages useful to an as wide range of users as possible. If we interpret them as meaning that ExcludeArch is preferred to shipping a package that requires something like SSE3, they become counterproductive. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald wrote: > come on and don't tell me 99% of i686 users have machines older than 10 > years https://en.wikipedia.org/wiki/SSE3 because "ntel introduced SSE3 > in early 2004 with the Prescott revision of their Pentium 4 CPU" > > those machines would not be able to run a recent Fedora due all the > bloat introduced over the years - try a system update with 512 MB and > shake your hands with the OOM-killer I have a computer without SSE3. It runs Fedora just fine. I initially had 1 GiB of RAM in it, I upgraded it to 3 GiB. > Fedora incompatible? > > many software including the kernel has binary code for AVX and even AVX2 > and is using CPU runtime detection - using SSE3 instructions in a > software is not bound to compiler flags and when a upstream developer > don't care about 11 year old hardware you have to suck taht downstrem What he means is that Fedora policies DO NOT allow REQUIRING those instruction sets. Runtime detection is of course fine. The problem we are having now is upstream projects that do not support ANY kind of fallback, not even at compile time. (Those are also inherently non- portable because they don't work on any other CPU architecture.) Chromium/V8 is another such problem program. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Oh, and I forgot (sorry for replying to the same mail twice): Ralf Corsepius wrote: > ACK. Does anybody know if there are any x86_64/64bit-CPUs which do not > support sse3? The first Athlon64 CPUs did not support SSE3. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
07.11.2015, 12:19, "Germano Massullo": > [10:11] Germano: also feel free to refer those guys here to > us if they have questions. seems to me that some direct contact may be > better. What those guys? The only one who wants to kill off darktable is you, Germano. You almost succeeded last time and had it removed, before the community stepped in. Why are you trying to repeat it again? Seriously. Just let darktable be. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 20:36 schrieb drago01: On Sat, Nov 7, 2015 at 7:40 PM, Reindl Haraldwrote: the point is compile a single application with new features won#t gain that muc 8until you do the same with most libraries used by the software) but having the whole distribution is a summary with a completly different behavior than a single test of software xyz Uh no you just have to compile the software any libraries used by the hot paths of the software you are trying to test. for which one? that don't show the impact on a complete distribution anyways if there would be no difference kernel upstream won't invest that much time for runtime-cpu-detection (look at the bootlog on different hardware) here are examples http://www.phoronix.com/scan.php?page=news_item=MTU0MTY http://www.phoronix.com/scan.php?page=news_item=MTI1Njc You missed the point completely. 1. AVX != (S)SSE3 i know that There are lots of workloads that would benefit from AVX but SSE3 doesn't add that much (enabling SSE2 on i686 would gain you more for instance). enable SSE3 don't diable SSE2 so "gain you more" is impossible 2. Runtime detection does not have the cost of dropping support for specific hardware runtime detection needs to be implemented in every relevant software at it's own and hence has a *high cost* add -msse3 to the default falgs have no costs at all but only an impact of a unknown amount f *very* outdated hardware which is unlikely running a bleeding edge distribution frankly whatever somebody has run on a 10 years old machine can be easily virtualized and i doubt that many people have a 10 years old computer as their only device, as far there is something with a core2 or newer in the house you can virtualize the other machine and save a lot of energy signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Sat, Nov 7, 2015 at 7:40 PM, Reindl Haraldwrote: > > > Am 07.11.2015 um 19:27 schrieb drago01: >> >> On Sat, Nov 7, 2015 at 7:23 PM, Reindl Harald >> wrote: >>> >>> maybe you did not get "not that i say Fedora should go ahead and build >>> with >>> -mavx" but we talk about SSE3 >> >> >> Compiling everything with sse3 does have a cost (dropping support for >> some hardware) but what does it gain us? >> Better performance in theory? To convince anyone you have to at least >> provide numbers. Otherwise the discussion is a waste of time > > > how do you expect to get such numbers? > > even if you test whatever application with -msse2 versus -msse3 mosdt parts > of the operating system and the whole libraries are still built with only > SSE2 support > > the point is compile a single application with new features won#t gain that > muc 8until you do the same with most libraries used by the software) but > having the whole distribution is a summary with a completly different > behavior than a single test of software xyz Uh no you just have to compile the software any libraries used by the hot paths of the software you are trying to test. > if there would be no difference kernel upstream won't invest that much time > for runtime-cpu-detection (look at the bootlog on different hardware) > > here are examples http://www.phoronix.com/scan.php?page=news_item=MTU0MTY > http://www.phoronix.com/scan.php?page=news_item=MTI1Njc You missed the point completely. 1. AVX != (S)SSE3 There are lots of workloads that would benefit from AVX but SSE3 doesn't add that much (enabling SSE2 on i686 would gain you more for instance). 2. Runtime detection does not have the cost of dropping support for specific hardware. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Felix Miata wrote: >> I have a computer without SSE3. It runs Fedora just fine. I initially had >> 1 GiB of RAM in it, I upgraded it to 3 GiB. > > I have approximately 40 Fedora installations on 32 bit CPUs. About half of > those are on Prescotts. I'm unable to notice that those with more than 1G > RAM run any better than those with more RAM. There are mainly 2 reasons I had to upgrade the RAM: 1. Akonadi / KMail 2. 2. "Modern" websites with massive abuse of inline images. Both of those eat lots and lots of RAM. > I am able to notice that the older, slower ones that do have SSE2 are able > to run, but with Plasma 5, those are very slow. Those do run respectably > with lighter WMs/DEs. The main limiting factor there is probably not the CPU, but the GPU. If your GPU does not support OpenGL 2 (OpenGL 1.x is not enough because GLSL shaders are required), then QML 2 and thus Plasma 5 will fall back to software rendering through the llvmpipe, which of course on an old CPU is not impressive speed-wise. What might help is disabling all the animations that can be disabled. > For those without SSE2, Plasma 5 is completely unusable, but OK with any > of the lightweights. Without SSE2, you additionally also lose the QML/JavaScript JIT in QML 2. (There is no implementation of the JIT for the 387 FPU instructions, only for the SSE2 ones.) So you also get an interpreter in C++ there. (At least there IS a fallback, unlike in V8 and thus Chromium/QtWebEngine.) And of course the llvmpipe will also be slower without SSE2 instructions. And they will contend the CPU. There is not much Fedora can do about that, it is a result of the design choices in upstream Qt. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 18:30 schrieb Kevin Kofler: Felix Miata wrote: I have a computer without SSE3. It runs Fedora just fine. I initially had 1 GiB of RAM in it, I upgraded it to 3 GiB. I have approximately 40 Fedora installations on 32 bit CPUs. About half of those are on Prescotts. I'm unable to notice that those with more than 1G RAM run any better than those with more RAM. There are mainly 2 reasons I had to upgrade the RAM: 1. Akonadi / KMail 2. 2. "Modern" websites with massive abuse of inline images. Both of those eat lots and lots of RAM not to forget that we now have many cases where GTK2/GTK3/QT4/QT5 are loaded at the same time, with that memory not so long ago a whole graphical operating system including applications is working fine _ below 1.5 GB RAM no way at all frankly we recently upgraded our server-hardware and changed the clustermode to sandybridge, all seemed to work fine (except https://bugzilla.redhat.com/show_bug.cgi?id=1269014 which is a very interesting bug) there are tiny backup VM's with 1024 Mb RAM over years for each production machine running a mysqld-slave and hourly rsync of the userdata for later offsite-backups from there day 1 1:16: one of them crashed and got restarted by VMware HA day 2 1:16: one of them crashed and got restarted by VMware HA well, the screenshot vmware makes from the VT showed page allocation errors and finally i realized that the once per day added --delay-updates rsync param on the new hardware for whatever reason triggered that errors increased *any* virtual machine to at least 1.5 GB RAM - no problems for 2 months now signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 19:26 schrieb Neal Gompa: On Sat, Nov 7, 2015 at 1:23 PM, Reindl Harald
Re: Retire a package from Fedora i686 (not x86_64)
On Sat, Nov 7, 2015 at 7:23 PM, Reindl Haraldwrote: > > > Am 07.11.2015 um 19:01 schrieb Kevin Kofler: >> >> Reindl Harald wrote: >>> >>> well, who's to say that we stay forever on a level of CPU >>> feature-support while there are instruction sets available for a whole >>> decade which improve performance, save power in case you need fewer >>> instructions doing the same work? >>> >>> not that i say Fedora should go ahead and build with -mavx but a >>> discussion about SSE3 in 2015 is really odd >> >> >> Sorry, but hardware simply does not get replaced instantly. This is a >> matter >> of both: >> * cost – hardware is not free (as in beer) > > > that *really* depends > > these days (and i talk about hardware from 2011) you can virtualize things > fast, easy and efficient and consolidate machines on more or less cheap > hardware > > besides it brings new CPU capabilities it saves a lot of energy: > > * one machine instead 2,3,4 or 10 machines > * the new hardware needs so much less power > * your UPS (in case you use one) can be cheaper or lasts longer > * cooling is easier, cheaper and not that loud > > some numbers in that context: > > in early 2010 we planned a UPS systems to last for around 3 hours, with the > current hardware and after consolidation the same UPS lasts 9 hours now and > the air conditioner has lower load and longer lifetime > > the same applies to consumer hardware, my 365/7 running homeserver from 2011 > is a power beast and the whole IT needs around 50W while the previous > hardware needed 120W - that's money, that's heat on summer days, that's less > noise and hardddisks don't need to replaced that often because they are > cooler > >> * ecology – there are huge landfills in China, India and Africa full of >> electronic trash from Europe and the USA; that gets mined for >> recyclable materials in an extremely polluting way that not >> only >> ruins the environment, but also damages the health of the >> people >> doing the mining (and the most dangerous materials are >> handled >> by children). Materials used for electronic components are >> precious, someone has to recycle them or they would run out >> pretty soon. > > > that is true but see above > >> As a developer who writes mathematical software for a living, I'd love >> being >> able to require AVX-512 right now (https://en.wikipedia.org/wiki/AVX-512 – >> finally a way to specify the rounding mode per operation >> (https://en.wikipedia.org/wiki/EVEX_prefix) rather than through the >> extremely expensive stateful fesetround operations that reset the whole >> prefetch queue, interval arithmetic should become MUCH faster with that), >> but I have to deal with real-world CPUs that are here NOW (so it might >> make >> sense to have runtime detection for AVX-512 only in 1 or 2 years, and >> requiring it is simply not possible within the next decade) > > > maybe you did not get "not that i say Fedora should go ahead and build with > -mavx" but we talk about SSE3 Compiling everything with sse3 does have a cost (dropping support for some hardware) but what does it gain us? Better performance in theory? To convince anyone you have to at least provide numbers. Otherwise the discussion is a waste of time. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Sat, Nov 7, 2015 at 1:23 PM, Reindl Haraldwrote: > > > Am 07.11.2015 um 19:01 schrieb Kevin Kofler: > >> Reindl Harald wrote: >> >>> well, who's to say that we stay forever on a level of CPU >>> feature-support while there are instruction sets available for a whole >>> decade which improve performance, save power in case you need fewer >>> instructions doing the same work? >>> >>> not that i say Fedora should go ahead and build with -mavx but a >>> discussion about SSE3 in 2015 is really odd >>> >> >> Sorry, but hardware simply does not get replaced instantly. This is a >> matter >> of both: >> * cost – hardware is not free (as in beer) >> > > that *really* depends > > these days (and i talk about hardware from 2011) you can virtualize things > fast, easy and efficient and consolidate machines on more or less cheap > hardware > > besides it brings new CPU capabilities it saves a lot of energy: > > * one machine instead 2,3,4 or 10 machines > * the new hardware needs so much less power > * your UPS (in case you use one) can be cheaper or lasts longer > * cooling is easier, cheaper and not that loud > > some numbers in that context: > > in early 2010 we planned a UPS systems to last for around 3 hours, with > the current hardware and after consolidation the same UPS lasts 9 hours now > and the air conditioner has lower load and longer lifetime > > the same applies to consumer hardware, my 365/7 running homeserver from > 2011 is a power beast and the whole IT needs around 50W while the previous > hardware needed 120W - that's money, that's heat on summer days, that's > less noise and hardddisks don't need to replaced that often because they > are cooler > > * ecology – there are huge landfills in China, India and Africa full of >> electronic trash from Europe and the USA; that gets mined for >> recyclable materials in an extremely polluting way that not >> only >> ruins the environment, but also damages the health of the >> people >> doing the mining (and the most dangerous materials are >> handled >> by children). Materials used for electronic components are >> precious, someone has to recycle them or they would run out >> pretty soon. >> > > that is true but see above > > As a developer who writes mathematical software for a living, I'd love >> being >> able to require AVX-512 right now (https://en.wikipedia.org/wiki/AVX-512 >> – >> finally a way to specify the rounding mode per operation >> (https://en.wikipedia.org/wiki/EVEX_prefix) rather than through the >> extremely expensive stateful fesetround operations that reset the whole >> prefetch queue, interval arithmetic should become MUCH faster with that), >> but I have to deal with real-world CPUs that are here NOW (so it might >> make >> sense to have runtime detection for AVX-512 only in 1 or 2 years, and >> requiring it is simply not possible within the next decade) >> > > maybe you did not get "not that i say Fedora should go ahead and build > with -mavx" but we talk about SSE3 > > talking about a rapid moving distribution like Fedora and at the same time > hestitate to use technology available for a decade is strange - i guess > people plan to run their hardware for 10, 12, 15 years are not using > typically a distribution like Fedora > > > The flip side of it is that Fedora is the foundation for RHEL and other distributions. Most of the others don't matter so much, but RHEL does, as that has to support servers and such. But even there, I would suggest to say that we could say that RHEL 8 shouldn't have a problem bumping up, since RHEL 7 supports those systems fine and will be supported for 9 more years. -- 真実はいつも一つ!/ Always, there's only one truth! -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 19:27 schrieb drago01: On Sat, Nov 7, 2015 at 7:23 PM, Reindl Haraldwrote: maybe you did not get "not that i say Fedora should go ahead and build with -mavx" but we talk about SSE3 Compiling everything with sse3 does have a cost (dropping support for some hardware) but what does it gain us? Better performance in theory? To convince anyone you have to at least provide numbers. Otherwise the discussion is a waste of time how do you expect to get such numbers? even if you test whatever application with -msse2 versus -msse3 mosdt parts of the operating system and the whole libraries are still built with only SSE2 support the point is compile a single application with new features won#t gain that muc 8until you do the same with most libraries used by the software) but having the whole distribution is a summary with a completly different behavior than a single test of software xyz if there would be no difference kernel upstream won't invest that much time for runtime-cpu-detection (look at the bootlog on different hardware) here are examples http://www.phoronix.com/scan.php?page=news_item=MTU0MTY http://www.phoronix.com/scan.php?page=news_item=MTI1Njc signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 19:01 schrieb Kevin Kofler: Reindl Harald wrote: well, who's to say that we stay forever on a level of CPU feature-support while there are instruction sets available for a whole decade which improve performance, save power in case you need fewer instructions doing the same work? not that i say Fedora should go ahead and build with -mavx but a discussion about SSE3 in 2015 is really odd Sorry, but hardware simply does not get replaced instantly. This is a matter of both: * cost – hardware is not free (as in beer) that *really* depends these days (and i talk about hardware from 2011) you can virtualize things fast, easy and efficient and consolidate machines on more or less cheap hardware besides it brings new CPU capabilities it saves a lot of energy: * one machine instead 2,3,4 or 10 machines * the new hardware needs so much less power * your UPS (in case you use one) can be cheaper or lasts longer * cooling is easier, cheaper and not that loud some numbers in that context: in early 2010 we planned a UPS systems to last for around 3 hours, with the current hardware and after consolidation the same UPS lasts 9 hours now and the air conditioner has lower load and longer lifetime the same applies to consumer hardware, my 365/7 running homeserver from 2011 is a power beast and the whole IT needs around 50W while the previous hardware needed 120W - that's money, that's heat on summer days, that's less noise and hardddisks don't need to replaced that often because they are cooler * ecology – there are huge landfills in China, India and Africa full of electronic trash from Europe and the USA; that gets mined for recyclable materials in an extremely polluting way that not only ruins the environment, but also damages the health of the people doing the mining (and the most dangerous materials are handled by children). Materials used for electronic components are precious, someone has to recycle them or they would run out pretty soon. that is true but see above As a developer who writes mathematical software for a living, I'd love being able to require AVX-512 right now (https://en.wikipedia.org/wiki/AVX-512 – finally a way to specify the rounding mode per operation (https://en.wikipedia.org/wiki/EVEX_prefix) rather than through the extremely expensive stateful fesetround operations that reset the whole prefetch queue, interval arithmetic should become MUCH faster with that), but I have to deal with real-world CPUs that are here NOW (so it might make sense to have runtime detection for AVX-512 only in 1 or 2 years, and requiring it is simply not possible within the next decade) maybe you did not get "not that i say Fedora should go ahead and build with -mavx" but we talk about SSE3 talking about a rapid moving distribution like Fedora and at the same time hestitate to use technology available for a decade is strange - i guess people plan to run their hardware for 10, 12, 15 years are not using typically a distribution like Fedora signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald wrote: > well, who's to say that we stay forever on a level of CPU > feature-support while there are instruction sets available for a whole > decade which improve performance, save power in case you need fewer > instructions doing the same work? > > not that i say Fedora should go ahead and build with -mavx but a > discussion about SSE3 in 2015 is really odd Sorry, but hardware simply does not get replaced instantly. This is a matter of both: * cost – hardware is not free (as in beer), * ecology – there are huge landfills in China, India and Africa full of electronic trash from Europe and the USA; that gets mined for recyclable materials in an extremely polluting way that not only ruins the environment, but also damages the health of the people doing the mining (and the most dangerous materials are handled by children). Materials used for electronic components are precious, someone has to recycle them or they would run out pretty soon. As a developer who writes mathematical software for a living, I'd love being able to require AVX-512 right now (https://en.wikipedia.org/wiki/AVX-512 – finally a way to specify the rounding mode per operation (https://en.wikipedia.org/wiki/EVEX_prefix) rather than through the extremely expensive stateful fesetround operations that reset the whole prefetch queue, interval arithmetic should become MUCH faster with that), but I have to deal with real-world CPUs that are here NOW (so it might make sense to have runtime detection for AVX-512 only in 1 or 2 years, and requiring it is simply not possible within the next decade). Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald wrote: > if there would be no difference kernel upstream won't invest that much > time for runtime-cpu-detection (look at the bootlog on different hardware) Exactly BECAUSE performance-critical software normally does runtime detection (except where upstream is really really lazy), hardcoding a requirement for newer instruction sets does not buy us anything. The absence of a portable fallback also makes the software unusable on our secondary architectures, and in some cases (such as Darktable), even on the PRIMARY architecture ARM. (Now, whether ARM should be a primary architecture to begin with is a different matter, but it currently is.) So I really don't understand those upstreams that provide ONLY platform- specific vector code with no portable version. Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald wrote: > frankly whatever somebody has run on a 10 years old machine can be > easily virtualized and i doubt that many people have a 10 years old > computer as their only device, as far there is something with a core2 or > newer in the house you can virtualize the other machine and save a lot > of energy Virtualization is not a solution for most home users: * It assumes you have multiple machines to replace with one VM host and VMs. * Virtualization also does not magically provide you with monitors and keyboards, which are essential parts of a home system. And the different seats (where there is more than one computer to begin with) are often in different rooms. This cannot trivially be replaced by a VM-based setup. You'd need at least thin clients in addition to the VMs. As for the "core2 or newer" part: The average Core 2 system is going to run out of RAM very quickly if you try to run VMs on it. My Core 2 Duo notebook can do hardware virtualization, and I use that to test live ISOs, but if I start one Plasma 4 VM in addition to the Plasma 4 host, the RAM (4 GiB) is already full (and Plasma 5 is reported to be no better and probably worse there). (I need to allocate 2 GiB for the VM or Plasma will just not work in it. If you count the same for the host, it's no surprise that the 4 GiB of RAM end up full.) You need to think outside of the company scenario you are used to (and even your home is not a typical home user environment). Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Reindl Harald wrote: > frankly whatever somebody has run on a 10 years old machine can be > easily virtualized and i doubt that many people have a 10 years old > computer as their only device, as far there is something with a core2 or > newer in the house you can virtualize the other machine and save a lot > of energy Virtualization is not a solution for most home users: * It assumes you have multiple machines to replace with one VM host and VMs. * Virtualization also does not magically provide you with monitors and keyboards, which are essential parts of a home system. And the different seats (where there is more than one computer to begin with) are often in different rooms. This cannot trivially be replaced by a VM-based setup. You'd need at least thin clients in addition to the VMs. As for the "core2 or newer" part: The average Core 2 system is going to run out of RAM very quickly if you try to run VMs on it. My Core 2 Duo notebook can do hardware virtualization, and I use that to test live ISOs, but if I start one Plasma 4 VM in addition to the Plasma 4 host, the RAM (4 GiB) is already full (and Plasma 5 is reported to be no better and probably worse there). (I need to allocate 2 GiB for the VM or Plasma will just not work in it. If you count the same for the host, it's no surprise that the 4 GiB of RAM end up full.) You need to think outside of the company scenario you are used to (and even your home is not a typical home user environment). Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 08.11.2015 um 00:56 schrieb Kevin Kofler: Reindl Harald wrote: if there would be no difference kernel upstream won't invest that much time for runtime-cpu-detection (look at the bootlog on different hardware) Exactly BECAUSE performance-critical software normally does runtime detection (except where upstream is really really lazy), hardcoding a requirement for newer instruction sets does not buy us anything. define performance critical our cms systems got 5 times faster within 6 months by micro-optimizig each alone would not had any effect The absence of a portable fallback also makes the software unusable on our secondary architectures, and in some cases (such as Darktable), even on the PRIMARY architecture ARM. (Now, whether ARM should be a primary architecture to begin with is a different matter, but it currently is.) So I really don't understand those upstreams that provide ONLY platform- specific vector code with no portable version i understand them well - not endless time and write effecient software without test on dozens of environments older than 10 years - as C programmer if i would write new code these days i would even go so far and rely on AVX yes, that can't be done for a distribution, but i know that all machines sorrounding me are SandyBridge bases, even the network routers for small offices and so why would i care about anything else? signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On Sat, 7 Nov 2015 04:29:48 +0100 Ralf Corsepiuswrote: > On 11/06/2015 10:00 PM, Kevin Kofler wrote: > > Germano Massullo wrote: > >> For example, SSE3 instructions set is one of the minimum > >> requirements and 99% of 32 bit only CPUs do not support it. Only > >> Pentium 4 >= Prescott architecture supports it. > > > > I think building it with SSE3 is better than excluding the > > architecture entirely. And FYI, requiring SSE3 is not really > > allowed even for x86_64 (you are allowed to assume only SSE2), so > > by that argument, there is no architecture left. > ACK. Does anybody know if there are any x86_64/64bit-CPUs which do > not support sse3? > > I don't see sse3 in /proc/cpuinfo of any machine I have, but I also > don't see any runtime error/warning from darktable. > SSE3 is reported as 'pni' (Prescott New Instructions) > > Of course, a better solution would be to fix the package to work on > > all CPUs you support. > ACK. > > > But if upstream doesn't care, it's going to be a problem. :-( > Exactly - That's the actual problem. Upstream does not care and > Fedora seems unable to address this issue. > > Ralf > -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/07/2015 04:55 AM, Reindl Harald wrote: Am 06.11.2015 um 20:11 schrieb Germano Massullo: For example, SSE3 instructions set is one of the minimum requirements and 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott architecture supports it seriously? Probably. with that argumentation Fedora would drop i686 at all No. darktable handles the situation on sse3-less CPUs gracefully at run-time. The actually problem is darktable wanting to be compiled with -msse3, which means being compiled with a Fedora incompatible instruction set, on both i386 and x86_64. well, i don't have any non x86_64 machien but the argumentation is bous You are right, these days in professional PC-usage, i686ers likely are rare and hard to find. But in personal (private) PC-usage, the situation is different. There, 5/6-year old 32bit PCs are still wide spread and will be supported by Windows for many years. Also consider, people are using Linux on such machines to extend these machines life time. Feel free to drop the i386, if you want drop the end-user/desktop market and to leave this market to Ubuntu, Debian and other distributions. Ralf -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/07/2015 05:03 AM, Reindl Harald wrote: Am 07.11.2015 um 05:00 schrieb Luya Tshimbalanga: On 06/11/15 07:29 PM, Ralf Corsepius wrote: But if upstream doesn't care, it's going to be a problem. :-( Exactly - That's the actual problem. Upstream does not care and Fedora seems unable to address this issue. Ralf According to my system, it has SSSE3. Is it SSE3? That's what my question actually was about. I see ssse3 on all my x86_64ers, I see sse3 on my 32bit atoms, but I don't know if ssse3 implies sse3 and I don't know what -msse3 actually does to the instruction set when being used on x86_64ers and which impact it has (-msse3 is not part of Fedora's gcc implicit CFLAGS). any system not older than 10 years has SSE3 and frankly systems older than 10 years are hardly a traget for Fedora at all So, Fedora or Linux as a resort to escape Windows on old HW is not a Fedora target anymore? Ralf -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Il 07/11/2015 04:29, Ralf Corsepius ha scritto: > I don't see sse3 in /proc/cpuinfo of any machine I have, but I also > don't see any runtime error/warning from darktable. Because SSE3 is shown as "pni", Prescott New Instructions [1] [2] [1]: https://en.wikipedia.org/wiki/SSE3 [2]: http://unix.stackexchange.com/questions/43539/what-do-the-flags-in-proc-cpuinfo-mean -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 07.11.2015 um 05:00 schrieb Luya Tshimbalanga: On 06/11/15 07:29 PM, Ralf Corsepius wrote: But if upstream doesn't care, it's going to be a problem. :-( Exactly - That's the actual problem. Upstream does not care and Fedora seems unable to address this issue. Ralf According to my system, it has SSSE3. Is it SSE3? Taken from A10-7400P laptop any system not older than 10 years has SSE3 and frankly systems older than 10 years are hardly a traget for Fedora at all signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 06/11/15 07:29 PM, Ralf Corsepius wrote: > On 11/06/2015 10:00 PM, Kevin Kofler wrote: >> Germano Massullo wrote: >>> For example, SSE3 instructions set is one of the minimum >>> requirements and >>> 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott >>> architecture supports it. >> >> I think building it with SSE3 is better than excluding the architecture >> entirely. And FYI, requiring SSE3 is not really allowed even for >> x86_64 (you >> are allowed to assume only SSE2), so by that argument, there is no >> architecture left. > ACK. Does anybody know if there are any x86_64/64bit-CPUs which do not > support sse3? > > I don't see sse3 in /proc/cpuinfo of any machine I have, but I also > don't see any runtime error/warning from darktable. > >> Of course, a better solution would be to fix the package to work on >> all CPUs >> you support. > ACK. > >> But if upstream doesn't care, it's going to be a problem. :-( > Exactly - That's the actual problem. Upstream does not care and Fedora > seems unable to address this issue. > > Ralf > According to my system, it has SSSE3. Is it SSE3? Taken from A10-7400P laptop. -- Luya Tshimbalanga Graphic & Web Designer E: l...@fedoraproject.org W: http://www.coolest-storm.net -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/06/2015 09:50 PM, Stuart Gathman wrote: On 11/06/2015 02:11 PM, Germano Massullo wrote: For example, SSE3 instructions set is one of the minimum requirements and 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott architecture supports it. For that kind of thing, add a runtime test for required CPU features at startup. That's what they actually do. cf. https://bugzilla.redhat.com/show_bug.cgi?id=1278064#c14 $ darktable [dt_init] unfortunately we depend on SSE3 instructions at this time. [dt_init] please contribute a backport patch (or buy a newer processor). That may not be necessary if the app already fails in a friendly way on old CPUs (as opposed to hanging or mysterious crashes). Ralf -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/06/2015 10:00 PM, Kevin Kofler wrote: Germano Massullo wrote: For example, SSE3 instructions set is one of the minimum requirements and 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott architecture supports it. I think building it with SSE3 is better than excluding the architecture entirely. And FYI, requiring SSE3 is not really allowed even for x86_64 (you are allowed to assume only SSE2), so by that argument, there is no architecture left. ACK. Does anybody know if there are any x86_64/64bit-CPUs which do not support sse3? I don't see sse3 in /proc/cpuinfo of any machine I have, but I also don't see any runtime error/warning from darktable. Of course, a better solution would be to fix the package to work on all CPUs you support. ACK. But if upstream doesn't care, it's going to be a problem. :-( Exactly - That's the actual problem. Upstream does not care and Fedora seems unable to address this issue. Ralf -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Am 06.11.2015 um 20:11 schrieb Germano Massullo: For example, SSE3 instructions set is one of the minimum requirements and 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott architecture supports it seriously? with that argumentation Fedora would drop i686 at all well, i don't have any non x86_64 machien but the argumentation is bous signature.asc Description: OpenPGP digital signature -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
For example, SSE3 instructions set is one of the minimum requirements and 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott architecture supports it. -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 06/11/15 14:18, Germano Massullo wrote: I just removed 32bit CPU support from Darktable's spec file due technical reasons. Are there any other things I need to clean up in Fedora infrastructure? Well you should start by reading the relevant guidelines: https://fedoraproject.org/wiki/Packaging:Guidelines#Architecture_Support Tom -- Tom Hughes (t...@compton.nu) http://compton.nu/ -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
06.11.2015, 17:18, "Germano Massullo": > I just removed 32bit CPU support from Darktable's spec file due > technical reasons. It is hard to help if you just say "due to technical reasons". What are they? -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
06.11.2015, 22:11, "Germano Massullo":For example, SSE3 instructions set is one of the minimum requirements and 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott architecture supports it. Prescott is 11 years old. I would expect vast majority of people who are currently running 32 bit Fedora to have newer computers. I would go even as far as to say that nobody who has an older than a 11 year old computer would even try to launch Darktable. And your solution is to just kill it off, because a few ancient computers can't run it? -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
On 11/06/2015 02:11 PM, Germano Massullo wrote: For example, SSE3 instructions set is one of the minimum requirements and 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott architecture supports it. For that kind of thing, add a runtime test for required CPU features at startup. If you don't want to touch the original code, just add an introducer that checks current hardware, then execs the app in /usr/libexec. That may not be necessary if the app already fails in a friendly way on old CPUs (as opposed to hanging or mysterious crashes). -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
Re: Retire a package from Fedora i686 (not x86_64)
Germano Massullo wrote: > For example, SSE3 instructions set is one of the minimum requirements and > 99% of 32 bit only CPUs do not support it. Only Pentium 4 >= Prescott > architecture supports it. I think building it with SSE3 is better than excluding the architecture entirely. And FYI, requiring SSE3 is not really allowed even for x86_64 (you are allowed to assume only SSE2), so by that argument, there is no architecture left. Of course, a better solution would be to fix the package to work on all CPUs you support. But if upstream doesn't care, it's going to be a problem. :-( We have a similar issue with QtWebEngine, because Chromium/V8 upstream requires SSE2 in their JIT and there is no interpreter fallback. (They are now working on an "interpreter" mode, but it is not a real interpreter fallback either, some stuff still uses the JIT.) Kevin Kofler -- devel mailing list devel@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/devel Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct