Phoronix IRC Log: 2009-07-25
Kano: hi michaellarabel
michaellarabel: hi kano
Kano: michaellarabel: just found a few missing libboost tests
Kano: for vdrdrift i had to install 2 more packages
Kano: /usr/include/boost/date_time/posix_time/posix_time_types.hpp : libboost-date-time-dev
Kano: /usr/include/boost/regex.hpp : libboost-regex-dev
michaellarabel: hmmm I hadn't run into that problem
Kano: depends on your config of apt-get i think
Kano: i usually use apt-get autoremove to clean up and remove other unused things
Kano: but your libboost check does not verify those files
michaellarabel: I'll add it into Git
Kano: i had libboost-dev, but those are not depending on it
michaellarabel: will add it in Git
Kano: that was b3
Kano: michaellarabel: do you use 2 laps now for vdrift? it takes much longer than before
Kano: and that 3 times...
michaellarabel: how slow is the game running for you?
Kano: on my fgrlx card with 6 fps ;)
Kano: no faster ati card here available
Kano: simple 3450
Kano: didnt you say you changes something in git?
Kano: git pull did notthing
Kano: 7.3 fps result... wow
Kano: maybe somebody should send me a faster passive cooled card
Kano: it is in my low noise box
Kano: michaellarabel: i guess u has apt-get with a setting that installs recommended packages by default. but on debian this is usually off
Kano: therefore vdrift compiled for you
Kano: best disalbe that setting or try with kanotix ;)
Kano: michaellarabel: next bug. when i test on a profile which is existing
Kano: then thin the results overview the new test is not shown only the first test
Kano: below there is the new test
Kano: just not in the overview
michaellarabel: Hmmm, what?
Kano: lets say i save something as 1
Kano: then itest another test with 1
Kano: then it is not shown correctly
Kano: in the overview part
michaellarabel: upload something to global so I can see how it's not shown correctly?
Kano: i wanted to test those only locally
Kano: but can copy you that
michaellarabel: Ah, I see now.
michaellarabel: that overview is all XSL driven and to write an "advanced" overview to handle it properly will be a pain in the ass with XSL. But I do have an image-rendered overview in development.
Kano: so it will only work with the same test now?
michaellarabel: for the overview, yes, well, if the FIRST test in the result file has all of the proper names
michaellarabel: if you had those date stamps used as the result identifiers in the other tests, within the first test, it should order everything fine
michaellarabel: basically it looks at only the first test result file to determine the header
Kano64: that would be the most common use when you use the compare url
Kano64: just when you have to enter always a name for the test to save locally it was more easy just to use alwys the same.
Kano64: if it would be autogenerated by default if none specified that would be better
Kano64: you get bored when you always have to think of a new save name
Kano: hi bkero
Kano: redeeman: still problems with your nv card?
Kano: just played a few h with my card, no problem at all
Kano: just found something weird with my usb mouse
Kano: when i run vdrift benchmark
Kano: then my mouse creates an error in dmesg and does not move anymore
Kano: only a change in the vdrift config file avoids that
redeeman: its not just me having the issues
Kano: that happens only in the benchmark mode, like it is used by pts
Kano: mousegrab = false
Kano: and it will work
Kano: but basically the system is really fast now
Kano: cpu likes fsb 400 ;)
Kano: bkero: did you bench against that yet
bkero: Kano: Not yet, I accidently got drunk last night.
bkero: We went to a grill that was it's own microbrewery.
bkero: and I just HAD to sample(with pints)!
Kano: i would suggest du to use pts git
bkero: Any way to do that Kanotix/
Kano: b3 has still some errors
Kano: apt-get remove --purge phoronix-test-suite
Kano: sudo apt-get update
Kano: sudo apt-get install git-core
bkero: Haven't installed it yet, so that won't be a problem :)
Kano: and then checkout pts, no problem
Kano: that works live
Kano: all tests i have done there are live mode
Kano: live with toram
bkero: I'll have to try this toram option :)
bkero: Will 6GB be enough?
Kano: well for those tests i needed only 4
Kano: you can use
Kano: sudo mount -o remount,size=5G /live/cow
bkero: That only gives me 1GB of RAM then
Kano: to have got a bigger ramdrive, i guess you know that
bkero: yea :)
Kano: ram is never a problem for those apps, only ramdisk can be full soon
Kano: i guess i will get 4 gb extra ram on monday
Kano: then i can test more things with 8 gb
Kano: i do not like to upload results with a 160 gb ide drive ;)
bkero: 1 to 4gb will probably make a difference, 4 to 8 won't (I bet) :)
bkero: I'll be uploading results with a flash drive :P
Kano: no for a complete suite
Kano: those are single tests currently
Kano: i thought of multicore suite or so
Kano: in theory i have got a spare 1 tb hd, but i want to use that for something else
Kano: michaellarabel: did you watch ram usage with parallel bzip
Kano: michaellarabel: at a certain point it is killed due to too much ram used
Kano: at least on a quad with 8 gb ram
Kano: it fulls the complete ram
Kano: so i would say the values are invalid
Kano: even the pre script takes ages
Kano: you create a 2 gb file from urandom, but urandom is really slow to read/copy
Kano: my x2-3800 runs now 4 min and the file is about 1 gb size...
Kano: maybe use a 256 mb or 512 mb file
Kano: povray runs with 1 core only
Kano: is that normal?
Kano: michaellarabel: povray is definitely no mulitcore test
Kano: watch htop while running it
michaellarabel: povray is single-core, but the new version coming is multi-core
Kano: hmm, well in pts git is is single core
Kano: i would change the bzip2 test. doesnt that get killed on your system due to low mem? i dont use swap partitions, but 4 gb ram completely filled is too much
Kano: ./compress-pbzip2: line 4: 29923 Aborted ./pbzip2 -c -p$NUM_CPU_CORES -r -5 ../compressfile > /dev/null 2>&1
Kano: looks like that then
Kano: value useless
Kano: you bench the time till your ram is filled nothing else
Kano: graphics-magic is only partly multicore, a bit more than povray but well not really multi
Kano: one thread 100%, 3 threads 20
Kano: could be improved it seems..
Kano: well some of the tests scale better
Kano: michaellarabel: i am sure a pc with 512 mb wins all compare-pkbzip2 results ;)
Kano: michaellarabel: i would at least remove parallel bzip2 from the suites or use a MUCH smaller testfile to avoid ram full aborts
michaellarabel: Kano: Will use 256MB instead of 2GB
netsurf3: any of you guys here in the beta?
Kano: i guess that should run in most cases, but how about checking $? and ignore the value in that case
Kano: it is not 0
michaellarabel: Kano: There is already that support
Kano: michaellarabel: incorrect, it showed values
michaellarabel: I meant for pts-core, it's not implemented within that specific test profile
Kano: then activate it?
michaellarabel: add "echo $? > ~/test-exit-status" the line below that command, if you want to test it.
Kano: interesting, why not test that for all?
michaellarabel: I have added it to most test profiles. Some of the test programs are weird in that they are returning values other than zero when it properly runs
Kano: you also need that for all build profiles
michaellarabel: there is that support there too
Kano: but i dont think they are valid even if you check that
michaellarabel: echo $? > ~/install-exit-status works similarly
Kano: because some may have got extra devel packages installed which configure will use to activate extra options
Kano: and then somebody compiles more than somebody else
Kano: you dont use that for pre?
Kano: The test exited with a non-zero exit status. Test run failed.
Kano: would work yes
michaellarabel: It's not implemented in pre... In what case would something in pre fail?
Kano: like what you here
Kano: you create a file
michaellarabel: oh, running out of disk space? will that return a non-0 status?
Kano: when disk is full the file is not correct
Kano: usuallly it will
Kano: best test it, but i guess it does not show 0