Phoronix IRC Log: 2008-06-27
redeeman: gah :|
redeeman: just got a mail from the shop i ordered the X1650 from, apparently they didn't have the card and just forgot to remove it from their shop, lol
vadi2: If anyone is tired of rickrolls: abp://subscribe/?location=http%3A%2F%2Frickrolldb.com%2Fricklist.txt&title=Rickroll%20Blacklist
vadi2: (paste into firefox)
Deanjo: heh, watching PTS run on a G15 keyboard is kinda cool
michaellarabel: Deanjo: Did you write a module or what
Deanjo: Working on it as we speak
Ivanovic: michaellarabel: saw my query yesterday?
michaellarabel: via the module interface in trondheim-12 or what
Ivanovic: the one about the openmoko freerunner being available now
michaellarabel: Ivanovic: Not sure, I'll need to scroll back.
Deanjo: using the guide from the latest GIT
Ivanovic: i posted it in here and in a query, too
michaellarabel: been very busy with 4850 and 4870, so I think I missed it
Deanjo: Just debating if it should recompile gmp and use a static or use the distro's gmp
Deanjo: Guess it kind of depends if you want to test the distro or the cpu
Deanjo: since distro's gmp are not completely optimized
michaellarabel: test the CPU
Deanjo: k then that requires recompiling of gmp
Kano: michaellarabel: did you think about the search feature for pts
michaellarabel: About the test comparisons not showing up in search results? I haven't had the time yet to look at that bug, as it should be working.
Kano: i guess when i seek for the kernel version then maybe not all are stored in the field you seek
Kano: as it combines several different ones sometimes
michaellarabel: does it work if you try to search for say a CPU from a comparison?
vadi2: quick poll, which one do you prefer: http://img511.imageshack.us/img511/7546/screenshot2la9.png or http://img150.imageshack.us/img150/9263/screenshot3ld2.png (pick out of those two)
redeeman: but deanjo, michaellarabel
redeeman: gmp is a more complex issue than that
redeeman: for example
redeeman: core2 needs a patch to gmp
redeeman: or it's miserably slow
redeeman: core2 is not good at executing generic x86_64 code
Deanjo: Ya I know redeeman
redeeman: so in reality i think you need more than 1 gmp test
redeeman: 1 for generic x86_64 (on 64bit), and one for core2 optimized x86_64
redeeman: but i guess it depends on what one really wants to test
Deanjo: I'm taking all of that into account, that's why it's not a small project
redeeman: because everyone needing gmp would obviously get a patched core2
redeeman: so it depends on if you want to test generic code, or optimal code
Deanjo: The problem with some of the patches as well is that GMP is really anal about their license and they don't like taking in patches
redeeman: no, it's not that
Deanjo: Well it's one of the problems
redeeman: it's that the gmp dude does NOT like unclean code doing hacks for a specific chip in a given architecture
Deanjo: Maybe that's why it should be tested against the distro's stock GMP
redeeman: that depends what you want to show
redeeman: or well..
Deanjo: ya that's what I was saying earlier
redeeman: with distros gmp you NEVER know what there might be
redeeman: so actually i would say you need to build gmp yourself either way
redeeman: just whether to apply patches or not, is the question
redeeman: for instance, the system gmp on my core2 has the patches, so obviously if i compared to a default gentoo, i would come out as a clear winner on core2
Deanjo: it would be nice if gmpbench was threaded as well
Deanjo: redeeman, do they have a svn or the like somewhere?
redeeman: Deanjo: i don't know
Deanjo: http://gmplib.org/devel/ <--- that's the core2 patch your using?
Deanjo: the one at the bottom?
Deanjo: heh, you think gmp benefits much from optimization? http://pastebin.com/m50ea773f
Deanjo: Stock = GMPbench result: 8950 Optimized = GMPbench result: 13579
jacques-work: I'm confused - what's this? http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543~126871,00.html
Deanjo: means AMD started fixing their openGL
jacques-work: is it just really old news?
jacques-work: or is there a new driver?
Deanjo: " Version 8.49.7 of the ATI Catalyst Driver is now available"
Deanjo: It does seem behind the current release of the driver
jacques-work: yeah :-\
jacques-work: I figured it might be old news because I didn't see anything on phoronix
Deanjo: The net's "tubes" must of been congested
michaellarabel: Deanjo: How's the G15 PTS coming along?
Deanjo: Oh, heh, there is a plugin for the G15 that comes with the daemon to display a terminal session on the g15's display
Deanjo: I just routed it to that
redeeman: http://ebblog.dk/ <-- nice error message
cxo: is there a /dev/null thing for folders?
redeeman: you mean so that you can move it into a /dev/foldernull ?
Ivanovic: cxo: the dev/null for folders is /dev/null
redeeman: Deanjo: http://www.geocities.jp/ct35z/gmp-core2-en.html <-- that's the best core2 patch
Ivanovic: cxo: everything is a file!
cxo: no its not
cxo: $ mkdir /dev/null/abc
cxo: mkdir: cannot create directory `/dev/null/abc': Not a directory
redeeman: that EXACTLY means it's a file
cxo: its not a folder
redeeman: Deanjo: the japanese gmp core2 patch is based upon the one mentioned on that gmp site
redeeman: Deanjo: allthough im not sure if Jason has integrated those changes back into itself, i did tell him about the japanese ones about a month ago
redeeman: Deanjo: on what box did you run that gmpbench?
Deanjo: Kernel: Linux 22.214.171.124-1.1-default CPU: AMD Phenom(tm) 9850 Quad-Core Processor@3013.638 MHz CPU: AMD Phenom(tm) 9850 Quad-Core Processor@3013.638 MHz CPU: AMD Phenom(tm) 9850 Quad-Core Processor@3013.638 MHz CPU: AMD Phenom(tm) 9850 Quad-Core Processor@3013.638 MHz Mem/Use: 4062832 kB/3104616 kB XInfo: Screen #0: @ 1600x1200 pixels (330x240 millimeters) /24 bit Uptime: 0:39
Deanjo: that one
redeeman: can you try run the gmp-chudnovsky again now with optimized gmp?
Deanjo: ya I can in a bit
maligor: what's gmp?
redeeman: last time you were alot slower than my core2, but this time ill bet you will be beating me
Deanjo: got some heavy compiling happening right now
redeeman: maligor: yes
redeeman: 100 million digits takes 238 seconds for me
redeeman: dislikes webshops saying they have a product, and then not having it
redeeman: now i must find another place to buy an R500
mha: redeeman: the laptop?
redeeman: no, i ordered an X1650
maligor: redeeman, the one I tend to use lists their own warehouse status and the source warehouse status in some cases
mha: ow. gfx
maligor: they have X1650s infact
redeeman: maligor: yeah this one said it had it at them, even i could pick it up, but apparently they had not bothered updating the status
redeeman: fortunately i think i found a store that has it
redeeman: (provided its not also false there)
maligor: what country in question?
Deanjo: redeeman, you here?
redeeman: Deanjo: now i am
Deanjo: Something weird here, I'm smoking every record here http://translate.google.ca/translate?hl=en&sl=ja&u=http://h2np.net/pi/pi_record.html&sa=X&oi=translate&resnum=2&ct=result&prev=/search%3Fq%3Dgmp-chudnovsky%2Bphenom%26hl%3Den%26client%3Dfirefox-a%26rls%3Dorg.mozilla:en-US:official%26hs%3DBsG%26sa%3DG
redeeman: i believe that
Deanjo: and here
redeeman: did you try just gmp-chudnovsky with 100 mil digits?
redeeman: amd seems to be leading in scientific stuff
Deanjo: but when it comes to running just the gmp-chudnovsky.c
Deanjo: bs time = 215.100
Deanjo: gcd time = 53.420
Deanjo: div time = 21.480
Deanjo: sqrt time = 12.450
Deanjo: mul time = 7.230
Deanjo: total time = 258.410
Deanjo: P size=145605885 digits (1.456059)
Deanjo: Q size=145605879 digits (1.456059)
redeeman: my core2 beats that
redeeman: 238 in total for me
Deanjo: and that's using the most aggresive flags I can possibly use
redeeman: you just have to make sure it builds the x86_64 asm
Deanjo: what the gmp-chud.c?
redeeman: no gmp itself
Deanjo: ya it's building the asm
Deanjo: what's your configure flags for your gmp build?
redeeman: i cant check my vdb right now
redeeman: but i most probably have -O2 -fomit-frame-pointer -march=core2 -ftree-vectorize
redeeman: or well.. that's what my gentoo is set to, whether the ebuild filters is another matter
Deanjo: k let me know when you can
redeeman: that's gonna be days
Deanjo: whenever :P
Kano: michaellarabel: did you try wine 1.1.0 yet?
michaellarabel: No Kano, not yet. At any given time I have over a dozen things on my short-term TODO list, so it may take a while.
Kano: well for kanotix i have a deb already
Kano: dont know if it will work with something else exept debian etch, but you can try
redeeman: michaellarabel: you should test R500 dri again :D to see if those problems you mentioned last time are gone
Kano: i thought you will do that soon ;)
redeeman: i will
redeeman: but it's been delayed to (hopefully) monday
redeeman: since the shop i bought from didn't actually have it anyway, lol, and now i had to order from another shop
redeeman: lol i only first now realized the "kano" in "kanotix" as being something you make
Kano: well i am sure kanotix is in most cases the first distro to give you latest wine or display drivers and in some cases even new kernels
redeeman: gentoo used to be extremely fast, but they seem to have slowed down
redeeman: ofcourse in 99% cases people just version bump the ebuilds and get it straight away
Kano: well i do not update kde or so
Deanjo: sadly, 3ds-max 9 still will not run on wine
cxo: haha 3dsmax on wine?
cxo: now thats really dreaming
cxo: there is a whole subsystem of twain drivers that need to be implemented for it
cxo: let alone the copy protection, and render driver
cxo: used to use 3dsmax in school
vadi2: if enough commercial places need to run it on linux, i'm sure it'll be ported :)
Deanjo: They have been asking for it for a long time
Deanjo: Maybe one day now that discreet owns Maya too
vadi2: what we really need to do is convince hollywood that somehow nvidia open-sourcing their drivers will help them lots, too.
Deanjo: As far as the copy-protection goes, the hardware config is derived off the MAC address
Deanjo: I'd doubt that would help hollywood
vadi2: that's the problem. home customers make up a surprisingly small amount of revenue
cxo: yes home customers buy their software from torrent sites
cxo: which works out a lot cheaper
Deanjo: I was referring to nvidia open-sourcing their drivers
Deanjo: not the copyprotection lol
vadi2: they made them so good 'cause they bought hollywood with that.
vadi2: thankfully consumers got a piece of the deal ^^
Deanjo: workstations rarely use bleeding edge distros
cxo: london bridge is falling down
cxo: falling down, falling down
cxo: my fair lady
cxo: goes to get another beer
cxo: *burp* workstation, is just a fancy name for "box that runs ssh sessions"
Deanjo: lol How do you figure that
cxo: cos i'm smart
Deanjo: Workstations are where the user works
Deanjo: hardly ssh boxes when it comes to animation
cxo: you dont render on your workstation
Deanjo: No but you manipulate the models on the work station
vadi2: they employ armies of enslaved penguins to perform the rendering
vadi2: 'cuz they're cheap and easy to modify
Deanjo: final rendering is slammed over to the render farm
cxo: we dont have all that advanced technology, we're still using paper and pencil, and then we flip the pages real fast, and it looks cool
Deanjo: pffft, we use sand and stick
Deanjo: adding "blur" effect requires a strong wind :P
vadi2: what about slow motion
Deanjo: we draw in really cold Jello for that
vadi2: and you know, macs can buy this rly special stick that makes it sooo easy to draw.
Deanjo: Those shots have to be done during the winter
Deanjo: But then there is always the problem of fighting off the polar bears taking those shots
vadi2: oh btw did anyone see the screencast of the upcoming open peach game?
Deanjo: Ya I saw it
vadi2: It looks quite promising
vadi2: I hope they'll make a dvd with the movie + game. that I'd get
Deanjo: hmmm that epic dude hasn't been back on the forums lol
Deanjo: but he's reading......
m-c: vadi link?
vadi2: m-c, subcsribe to http://freegamer.blogspot.com/
Deanjo: I love how he didn't even try to give a timeframe or anything really solid
m-c: Thanks, man!
m-c: Excellent - The free software game requires Flash Player to watch the promotional video
m-c: *golf clap*
vadi2: I hear they used some nvidia stuff in there too! *conspiracy*
Deanjo: nvidia took out Jimmy Hoffa too
Deanjo: And sabotaged Rieser
Deanjo: Framed him as well
Deanjo: and I'm pretty sure that they are responsible for global warming
Deanjo: Of course them opening up the drivers would cure all of that
Deanjo: But that won't happen since Linus holds 50%+1 shares in nvidia and is really out to make millions
cxo: the price of the 4870 is going up, cos the n00b retailers have realised it kicks the $600 nvidia card's ass
michaellarabel: Note to PTS users... Now that a thread is open in the forums, I guess it's now fine to let you guys know that Sun is working to bring a Java OpenGL harness to the Phoronix Test Suite.
m-c: what is a "harness" ? sounds dirty...
michaellarabel: well, a collection of tests.
cxo: test suite that uses opengl
cxo: its probably java3d, its been around for a while
m-c: hmm - googling opengl harness gave me a boat load of examples - :( sorry, bad joke
m-c: Great job developing your corporate partnerships Michael..
vadi2: now if he got paid for this, would've been even better eh ;)
cxo: but jre is an ugly dep
cxo: hopefully it will work with gcj
michaellarabel: cxo: JRE/OpenJDK is in the PTS External Dependencies for ease-of-use if using a supported distribution.
vadi2: Anyone know when the sourceforge nominations are ending?
michaellarabel: vadi2: Soon... The ceremony is next month during OSCOn
vadi2: There's supposed to be the waiting after nominations, right?
vadi2: Ah. "Nominations have now been closed. Come back soon to see the finalists!"
vadi2: dumb question, but a man page says "See BLANK LINES." with blank lines in bold. How do I go there?
vadi2: nvm, lots of scrolling did it
michaellarabel: Urban Terror test profile now in PTS git
vadi2: wow. I'm impressed. you can edit documentation in monodoc and upload it to the server, right inside the program.
starkmjolk: oki :)
starkmjolk: wrong place (:
vadi2: it fails at having a debugger frontend though :( back to fugly ddd..
cruiseoveride: hey michaellarabel
cruiseoveride: did you run MID on qemu?
michaellarabel: QEMU right now
cruiseoveride: with kvm image on the website?
cruiseoveride: what platform is it emulating?
cruiseoveride: nice, i can try it on my p2 thin client then..
cruiseoveride: "These two products gave me that old Radeon 9700 launch feel again, the AMD/ATI empire strikes back .. hard."
cruiseoveride: Now that author and i have the same blood
m-c: cxo What is the link to the image?
cxo: oh MID... 1 sec
cxo: m-c, http://cdimage.ubuntu.com/mobile/releases/hardy/
m-c: nice - thanks
m-c: appreciate it cxo
cxo: yup no problem