Milyardo: michaellarabel: In order to familrize myself with PTS, I've decided that I was going to document and comment every single function in pts-core
Milyardo: I wrote a doxgen config file and I've started working on commenting pts-core/functions/pts-functions.php
Milyardo: I've uploaded the resulting documentation here http://acm.umflint.edu/~zpowers/doxy-pts-core/html/index.html
Milyardo: and the code I've commented on so far is here http://acm.umflint.edu/~zpowers/doxy-pts-core/html/pts-functions_8php.html
Milyardo: I'm a nub when I comes to git, so I don't know how I would commit the comments or generate patches
maligor: git diff?
Milyardo: doxygen can modify serveral hundred files at one
Milyardo: it'd be cumbersome to diff every one :P
Milyardo: win 13
raikazejin: holy hell redeeman you are in here
maligor: Milyardo, erhm.. git diff will diff against the working git
maligor: assuming you're working against head it'll generate a patch for head
Milyardo: :) thanks maligor
EvilGuru: I am looking at the VDPAU API and am wondering if it would be possible for ATI to provide an implementation of it on their newer cards (how nvidia specific is it as an interface?)
Milyardo: EvilGuru: I think you answered your own question, its just an interface
Milyardo: Of course, at the current moment, only ATi would know enough about the cards to implement VDPAU into any ATi driver
EvilGuru: OpenGL is just an interface, doesn't mean I'd want to try implementing it on a DSP :)
EvilGuru: It does seem to be gaining some momentum, however (I think SVN versions of mythtv provide support for it)
EvilGuru: http://http.download.nvidia.com/XFree86/vdpau/doxygen/html/group__api__winsys__x11.html -- maybe nvidia does intend for it to become something of a standard
Milyardo: <3 for doxygen
Milyardo: There are only pluses for nVidia to have VDPAU become a standard
maligor: the documentation drops won't have info on the UVD anyway
maligor: it's been stated a few times
EvilGuru: I am more than happy running binary blob drivers
maligor: I'm not
maligor: It leads to windowism
EvilGuru: I'd rather not; but I'm not going to loose any sleep over it
Milyardo: I'd rather not have them
maligor: Window's primary problem is crap drivers
maligor: they've started to enforce driver qualification in vista x64, but people don't like that because it stops them from installing third party drivers
Milyardo: PMP technologies are now pretty much ensuring now that neither company will ever drop propritary drivers for a fully open source one
maligor: PMP?
Milyardo: Protected Media Paths
EvilGuru: Intel has fully open source drivers for its video hardware although to say their crap is an understatement
EvilGuru: Pity as the X300/X3100 are actually quite reasonable, considering they are integrated chipsets
Milyardo: Whats crap about Intel drivers?
EvilGuru: Milyardo: Where do you want me to start?
EvilGuru: For the X3000 they advertise FBO support, (i.e., claim to support the extension) but don't actually implement it
EvilGuru: can easily bring down a system if you attempt to debug a GL application in GB
Milyardo: FBOs can only be used with UXA enabled IIRC
EvilGuru: *GDB
EvilGuru: For earlier cards there is no TEXTURE_RECTANGLE support and no claim of FBO support
EvilGuru: GLSL is nowhere to be found (even if the hardware does support GLSL)
EvilGuru: Just not suitable for any kind of serious OpenGL development or applications, even though the hardware is capable
raikazejin: i would like a decent card in my netbook for crying out loud but no
raikazejin: i wonder if they are required to cripple them for military purposes
EvilGuru: I doubt being able to play Quake on your netbook is a threat to national security
maligor: I expect intel just doesn't have a better chipset
maligor: for that purpose that is
EvilGuru: Anything faster would probably also have higher power requirements
Laverne: Hi everyone. I'm trying to get the test suite to compare my results with ones that are on global.pts.com, but if I run the "phoronix-test-suite benchmark anon-1216-28953-30592" for example, I always get "Not Recognized: anon-1216-28953-30592" back. I'm using the 1.6.0b1 Tydal release. Has anyone seen this problem before?
Milyardo: anon-1216-28953-30592 is not a test
Laverne: you're right about that :) what would the correct syntax be to compare results? the results page on the site says to run that
Milyardo: the benchmark command accepts a name for a test
Milyardo: for example if you want to benchmark open arena
Milyardo: phoronix-test-suite benchmark openarena
Milyardo: then pts will prompt you to save a named for the results after downloading open arena
Milyardo: After the test has been runm you will be prompted with the oppurtunity to compare the results on PTS Global
Laverne: so it is only possible to compare once the test is finished running? I have universe-cli benchmarked and saved the results locally, now I'd like to compare them to other results
Milyardo: use phoronix-test-suite upload
Milyardo: to upload the locally saved results
Milyardo: once they are on PTS global they can be compared
tschaka: according to e.g. http://www.phoronix.com/scan.php?page=article&item=linux_2628_benchmarks&num=6 it's possible to run the command Laverne mentioned, to compare ur local test with already uploaded.
tschaka: ah, ok ic
Laverne: ah, the command works if I allow external http within php.ini as specified in the install manual
Laverne: I do have to re-run benchmark universe-cli, but thats not a problem
redeeman: EvilGuru: intel's master branch should have glsl and stuff now
EvilGuru: redeeman: Might have to toy around with it
EvilGuru: What about FBO support, I read about it a while back possibly being added
redeeman: i believe opengl 2.1 support
EvilGuru: On the talk of FBO support, ATI need to work on theirs -- a lot of text-book OpenGL applications still fail
maligor: and their opengl 3.0 support segfaults
EvilGuru: wishes drivers wouldn't advertise support for extensions that they do not implement fully
redeeman: i didn't know ati had ogl3
maligor: redeeman, 9.1 does
redeeman: i see
EvilGuru: I am sure at least one of the fglrx devs has a OpenGL textbook, isn't that hard to run an example
maligor: well, it implements most of the required extensions
maligor: just trying to create a context using the opengl3 way causes a segfault
EvilGuru: As it makes developing GL applications very difficult, as we often have to check driver versions to decide if to use an extension or not
redeeman: i fucking hate slow cpu's
EvilGuru: While there is always the possibility that the examples are wrong (I am willing to bet they were tested on nvidia GPUs)
maligor: mm, phenom II 940's are already available
EvilGuru: When do ATI usually make their driver dumps (early, mid or late month)?
michaellarabel: EvilGuru: Mid
raikazejin: EvilGuru, if i can hook my netbook up to my Ew system on a ship and it can reliably handle maritime operations while maintaining that is a threat
EvilGuru: remembers the days when Sony PS2s had export regulations imposed on it due to potential use as a weapon
raikazejin: its why you dont see many EMP proof products in COTS
EvilGuru: I am sure even moderately component terrorists have a primitive enough understanding of electromagnetism to mitigate EMPs as a threat to their devices
raikazejin: EvilGuru, i mean in general
raikazejin: say NYC
raikazejin: 10 devices
raikazejin: i would guess that 97% of communications would down
raikazejin: silently
maligor: EvilGuru, yeah, you could use it to control ballistic missiles!
EvilGuru: Always wondered what to do with the old PS2 in the loft
EvilGuru: Ah the joys of fglrx: http://freddie.witherden.org/wz2100_2.jpg
EvilGuru: Strange, as it works on a 2400 with fglrx and on all my other systems
maligor: you're working on that?
EvilGuru: Yep
EvilGuru: http://freddie.witherden.org/wz2100.jpg is the semi-expected result
EvilGuru: (Apart from the transparent cliff edges)
Milyardo: That is a really high resolution to play that game with
EvilGuru: Wanted to see if I was fill rate limited
maligor: fill rate limited? on warzone?
EvilGuru: 1024x1024 textures do the job
EvilGuru: If one plays around with the drawing code it is possible to get an acceptable result at 3200x1200 (without perspective distortion)
maligor: interesting that tigerdirect and newegg don't even list phenom ii 940
EvilGuru: maligor: Dry launch?
maligor: what's a dry launch?
maligor: I mean they don't list them atall
maligor: you can get them here already
maligor: says they have 6 units in shop
EvilGuru: Dry launch is when the product has limited availability
maligor: ah, yes, finland first, americas next? :P
RobbieAB: Aren't tigerdirect and newegg budget online sellers?
maligor: they are?
maligor: I'm not entirely sure what a budget online seller is
maligor: I buy my parts in a online seller, they have 3 stores tho
maligor: Budget online selllers don't list products before they have them in their own warehouses?
redeeman: i imagine budget ones does
RobbieAB: maligor: possibly not, but they may have trouble getting the parts with a price while the items are still filtering into the system
maligor: what's a non-budget shop in the us?
RobbieAB: *shrug* Not a US person, so no clue.
EvilGuru: I think that both of them are reasonably large retailers
RobbieAB: The other obvious suggestion would be, if released this week, no one has yet had time to put them up
EvilGuru: Probably still hung over from new years
RobbieAB: Exactly
maligor: the shop I'm looking at has had them listed for a while now
maligor: would be interesting to see some reliable benches on them
buggs: michaellarabel, what is the sense in all the cpu bound benchmarks for linux ssd testing?
GNU\colossus: :)
maligor: buggs, what would be the point of benchmarking yet another ssd vs hdd?
maligor: I imagine the idea is to see which is better for netbook use
GNU\colossus: I imagine the idea is to assess whether the SSD in questions sucks (as all models we've seen so far below ~600US$ do) or not.
gentooer: i got my Ritek 32gb about 6 months ago for ~$100 and it definitely doesn't suck
gentooer: load times are much faster for everything
gentooer: hdparm reads about 120MB/s
GNU\colossus: reading usually is not a problem, yes
GNU\colossus: writing, otoh, is
GNU\colossus: (as the sqlite benchmark illustrates, iirc)
gentooer: well i've found it to be much faster than my previous 320gb western digital
gentooer: any easy way to benchmark writes?
gentooer: without installing a bunch of stuff..
GNU\colossus: bunzip this into a tarball on /dev/shm/ -> http://gentoo.inode.at/snapshots/portage-20090101.tar.bz2
GNU\colossus: then untar it onto your SSD
GNU\colossus: (running under `time`, of course)
GNU\colossus: flushing caches would also be an idea
GNU\colossus: before and afterwards
gentooer: how do i flush caches?
gentooer: just "sync"?
GNU\colossus: yes
GNU\colossus: sth like `sync; time (tar -xf /dev/shm/tarball.tar -C /your/ssd/mountpoint; sync)`
GNU\colossus: it's not 100% accurate
GNU\colossus: but sufficient, I think
gentooer: GNU\colossus, well it is a bit slower.. 3 seonds vs 20 seconds
gentooer: what would you expect on a "good" SSD?
GNU\colossus: so it's 7 times slower
gentooer: than RAM..
GNU\colossus: you're supposed to test against a hdd
gentooer: oh ok just a sec
GNU\colossus: the filesystem also matters a lot here of course
gentooer: i'm using reiser4 with lzo compression on both
gentooer: ok yeah i got 15 seconds on my 320gb hd
gentooer: it's slightly slower
GNU\colossus: it's probably linux' superior vfs cache that keeps your ssd from a posting worse result :)
GNU\colossus: ah or well maybe not
GNU\colossus: since we've synced
GNU\colossus: anyway, there are only a handful of SSDs that do not show that kind of performance regression
gentooer: well i've also turned on write caching and used the NOOP schedular. two things that aren't enabled by default
GNU\colossus: you could also try `time touch file{0..100000}`
GNU\colossus: on temp. directories on both volumes
GNU\colossus: (you might want to delete those files afterwards, too ;) and time that as well)
gentooer: creating files: SSD: 3.9sec, oldHD: 4.1 :)
PetoKraus: what's noop scheduler good for?
gentooer: all the other IO schedulars are optimized for disk hard drives, but they just decrease performance for SSD
PetoKraus: uh huh
gentooer: switching from the default to noop increased read speeds for me by about 25%
GNU\colossus: nice
GNU\colossus: seq. read isn't a problem for my disk setup
gentooer: raid?
GNU\colossus: yes
war9407: switching from the default to noop increased read speeds for me by about 25%
war9407: gentooer: what did you use to benchmark that
gentooer: war9407, just hdparm -t
gentooer: went from like 90MB/s to 120MB/s
war9407: not showing any difference for my x25-e
gentooer: really? what scheduler were you using before?
war9407: cfq vs noop
GNU\colossus: the intel's internal logic may try to cover up the lacking iosched that windows provides
GNU\colossus: (because it does not offer a choice there, and probably always works the way CFQ does for Linux)
Deanjo: I think windows does have a iosched options. That is what you set in it when you choose the machine role in windows settings IIRC.
cxo: Deanjo: belated merry christmas and happy new years
GNU\colossus: Deanjo: you mean "optimized for background tasks" and all that?
Deanjo: You too, (although I actually celebrate Ukrainian Christmas)
Deanjo: Ya
GNU\colossus: I see
GNU\colossus: well, who knows. we aren't allowed to look into it, right :)
Deanjo: Heh, well there was that win2k source code leak a few years back.............
Deanjo: gives up looking for details in the useless MSDN library
GNU\colossus: hehe
Deanjo: Ya, even though access is free it still costs too much :P
bkero: still has a copy of that source.
Deanjo: Ya, so do I........ somewhere.......
Deanjo: hands the spindles of "mystery dvd's" to bkero to sort out :P
bkero: lol
bkero: gb2jail
Deanjo: I think I could spend a week trying to sort those suckers out.
bkero: I do that every few years
bkero: It goes faster than you'd think