Summary: | genlop -q cannot find CPU at gentoo.linuxhowtos.org | ||
---|---|---|---|
Product: | Portage Development | Reporter: | Tim Boundy <gigaplex> |
Component: | Tools | Assignee: | Portage Tools Team <tools-portage> |
Status: | CONFIRMED --- | ||
Severity: | normal | CC: | korte, zoltan |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
Tim Boundy
2007-09-11 12:12:57 UTC
This is still an issue apparently, so the bug is technically valid. However looking through the CPU list at http://gentoo.linuxhowtos.org/compiletimeestimator/, it does not seem people have been submitting recent CPU-s to the database (probably neither build times). Does anyone still rely on this online database? - If yes, beyond fixing the CPU naming issue, probably there should be a way to automatically submit results to the database rather then emailing `genlop -ntl` outputs. - If not (more likely this is the case) then rather fixing this, removing the online query feature would make more sense. One more note regarding this feature. Even if we had auto-submit, the dataset would be relatively low confidence at this point. Why? - People use different type of SSD-s which are faster than HDD (not sure though if people still use HDDs as their main drive), - Some people compile in tmpfs, - Some people use distcc, - Cloud computing is now a thing, - Multicore systems are everywhere, users `-j` settings will heavily affect results. Thus genlop compile times are really comparable on a given host configuration not really across many hosts, even if they have the same CPU. So for the previously mentioned reasons as well what is listed above I'd vote for removing the query feature from genlop. Any and all opinions are welcome. Imho the feature should be removed or renewed and the DB should be put on a gentoo server. The database could be created on a (maybe protected) wiki page or a text file in a git repo. When installing gentoo on weak computers, like a RaspberryPi 400, the DB would be of use. I submitted values to gentoo.linuxhowtos.org, but they didn't get enlisted. using tcpdump: /query.php?cpuname=1x%20%20&vcpu=4&packetname=libreoffice The cpustring creation doesn't fit to the Pi400. It should be more like: "4x BCM2835 ARMv8, Raspberry Pi 400 Rev 1.0" or "Raspberry Pi 400 Rev 1.0, 4x BCM2835 ARMv8" details like ccache, distcc, CPU speed (over-/underclocked), (non-)rotational media, RAM, max RAM used, emerge -jx, MAKEOPTS="-jy" would be of use. a manually "compressed" cpuinfo: processor : 0-3 BogoMIPS : 108.00 Features : fp asimd evtstrm crc32 cpuid CPU implementer : 0x41 CPU architecture: 8 CPU variant : 0x0 CPU part : 0xd08 CPU revision : 3 Hardware : BCM2835 Revision : XXXXXX Serial : XXXXXXXXXXXXXXXX Model : Raspberry Pi 400 Rev 1.0 From app-portage/genlop-0.30.10-r2 /usr/bin/genlop line 547 open(my $cmdline, "/proc/cpuinfo"); while (<$cmdline>) { if (m/processor\s*:\s*(\d*)/) { $vcpu = $1 if ($1 > $vcpu); } if (m/model name\s*:\s*(.*)$/) { $modelname = $1; } if (m/cache size\s*:\s*(.*)$/) { $cachesize = $1; } if (m/physical id\s*:\s*(\d*)$/) { $pcpu++ if ($1 != $opcpu); $opcpu = $1; } } $vcpu++; $pcpu = 1 if ($pcpu == 0); my $cpuname = $pcpu . "x $modelname $cachesize"; $cpuname =~ s/ /%20/g; my $retval = LWP::Simple::get("http://gentoo.linuxhowtos.org/query.php?cpuname=$cpuname&vcpu=$vcpu&packetname=$packet"); |