spartacus: Hi. Quick question re the doco for the R5xx chips... Is this available somewhere online at all?
boris: when should 3d support be released ?
bridgman: The current docs are up at http://www.x.org/docs/AMD/
bridgman: We have a few more docs ready to release but I'm trying to get them moved
bridgman: to an amd.com web site. The radeonhd devs already have the docs but
bridgman: if anyone else needs them over the holidays you can email me
bridgman: my user name at amd "dot" com. The additional docs are
bridgman: m76 (like 630 but not missing lvtma and i2c) and rs690
bridgman: I don't really want to post the new docs on x.org 'cause we pretty much
bridgman: killed the server last time we posted docs there ;(
bridgman: boris: first 3d stuff (info & sample code required for drm) should go out
udovdh: what is in the docs?
bridgman: in the first week of Jan, as soon as we re-open
udovdh: I mean registers
udovdh: good news!
bridgman: udovdh: basically what you need for initialization, display, modesetting
bridgman: what some people call a 2d driver but no 2d acceleration
udovdh: but the new stuff you mention for 2008
udovdh: that is good news
udovdh: i.e.: the info will continue
udovdh: thanks for getting this done
bridgman: absolutely. First priority was hiring people so I didn't have to do everything myself ;)
bridgman: Now that Alex has started we can get about 5x as much done ;)
bridgman: On the 3d side the top priority was figuring out exactly what we had to hold back
udovdh: any info in there that might speed up mpeg etc?
bridgman: ... so far it's looking pretty good
bridgman: re: mpeg, we'll be releasing idct/mc info for everything up to 5xx inclusive
bridgman: for 6xx it's harder because the UVD block handles DRM as well as decode and
udovdh: rv630 here...
udovdh: that was I was wondering...
bridgman: I don't have a good way to enable use of the decoder without also making it
udovdh: and uvd was a nogo so far
bridgman: too easy to get around the DRM on windows
udovdh: maybe something for the hardware devs to think about for the next chip design
udovdh: modular features
bridgman: That's already on the requirements list for future chips. Won't make the next
udovdh: thanks for updating us
bridgman: generation (too far along already) but the ones in early development should be
bridgman: more open-source-friendly
bridgman: I don't know how it will work out yet -- we combine stuff to reduce
bridgman: cost and improve performance so there's a price for making it too modular
bridgman: but I think decode vs drm should be OK to split out
bridgman: The one thing folks won't like is that we really need to do 3d before video
bridgman: since the foundation is common, and from 5xx up we use shaders to get higher
bridgman: quality video processing and that needs the 3d base as a first step
bridgman: I was kinda surprised how much interest there was in video -- last time we polled
bridgman: users for the closed source driver we found AIGLX was about 5x as popular as
bridgman: improving video support. Turned out video was at least as high a priority, at least
bridgman: for the folks I have been talking to.
udovdh: depends on the type of users I guess?
udovdh: for now the driver is Ok for me
udovdh: but it uses quite some CPU when doing video intensive stuff
udovdh: so if that becomes better the card is very isable
bridgman: Yep. The sense I got from talking to the open source devs was that (with the
bridgman: possible exception of laptops) it was mostly users of older cards who were interested
bridgman: in accelerated decoding since they also tended to have slower CPUs
bridgman: I'm not actually seeing that so far but not sure how much of the pushback
udovdh: that is a good reason yes
bridgman: on UVD is hard requirement and how much is just outrage if we hold anything back
udovdh: but even on this AMD X2 BE2350 all the CPU you can spare is a bonus
udovdh: maybe *part* of it can be `opened`?
udovdh: or is that not an option?
bridgman: We're going to see if there's a way to open part, but right now we're stuck
bridgman: at initialization -- don't have a way to give you enough info to get the block up
bridgman: and running without putting DRM at risk. For now I'm just going to keep saying "no UVD"
udovdh: I understand. Thanks again for doing the work.
bridgman: but probably some time 2Q08 we'll look at it more and *might* get lucky.
udovdh: I think I can wait that long
bridgman: I have to stress that right now we don't have a solution so you need to
bridgman: assume less than 50/50 chance of opening UVD. The good news for you
dvandyk: bridgman: will the information on 3d be sufficient to write custom shader programs? I'm interested from the HPC point of view
bridgman: is that the IDCT block is outside UVD on 6xx so at least
bridgman: IDCT should be able to be accelerated
bridgman: dvandyk: there is a separate team working on opening up the shader cores
udovdh: ok, any part is welcome!
dvandyk: bridgman: will that be available together with the other stuff in the first week of january?
bridgman: for HPC work; that should come out some time in 1Q08 but I don't have an exact
dvandyk: thank you for that information
bridgman: schedule. The first release might rely on fglrx, not sure.
bridgman: I think there will be a few different approaches floating around in 1Q08. We'll have
bridgman: the first set of open source dev info but there probably won't be enough time
ajax: bridgman: actually _you_ didn't kill the server. phoronix linking directly into gitweb killed it.
ajax: but i'd happily set up a torrent ;)
bridgman: for the community to write tools; in parallel we'll be releasing some ready to go tools
bridgman: ajax: you're probably right. "Slashdotted" is just a lot easier to pronounce than
bridgman: "Phoronicized" or whatever ;)
bridgman: Good point about torrent, hadn't thought about that. I'm on dial-up so video
bridgman: downloading is not big in my world ;)
udovdh: send me the torrent and I'll support seeding for a few weeks
bridgman: OK, will look into it -- have to run now, just realized the time
dvandyk: bridgman: merry christmas
udovdh: ok thanks and bye, bridgman
bl4val: where do i get the radeonhd 1.1.0 drivers?
drac: from the link in /topic
udovdh-afk: and they work well!
bl4val: i can't find them
udovdh-afk: First http://wiki.x.org/wiki/radeonhd
udovdh-afk: then http://wiki.x.org/wiki/radeonhd#head-f79351b4e2b19fad40529ce297ac2d2a1e90354c
udovdh-afk: and don't quit on me.
udovdh-afk: like you did.
drac: hate when they do that
udovdh-afk: this is a slow channel. respect that :-)
tgilber1: is it ok to ask a question?
tgilber1: here goes
tgilber1: read-edid does not work on the ubuntu x86_64 platform
tgilber1: hence, the /etc/gdm/failsafeXServer script fails because it calls for the get-edid function
tgilber1: I am trying to use the ddcprobe command to get edid
tgilber1: when I use the ddcprobe, I get edid: 1 3
tgilber1: Does the 1 and 3 correspond to the video and audio functions of my 2400HD card
tgilber1: what does the get-edid command return for results
tgilber1: appreciate any help with this issue
tgilber1_: sorry got bumped off
tgilber1_: any thoughts on the ddcprobe and edid question?
udovdh-afk: does rhd_modprobe work?
udovdh-afk: gotta go btw...
Plouj: can someone give me some valid reasons not to buy a 2600XT for use with Linux?
Plouj: and what are the "M76 / RS690" mentioned here: http://www.phoronix.com/scan.php?page=news_item&px=NjI2MQ ?
werdz: Plouj: if you want any sort of 3D, you're currently forced to use fglrx, which is generally a train wreck of a driver
Sir_Lewk: particularly bad with the latest incarnation :S