12:40 karolherbst: mupuf: saw my ping in regards to shipping GPUs
12:40 mupuf: karolherbst: nope... my server ran out of storage and I lost some messages
12:41 karolherbst: ahh
12:42 karolherbst: I was asking about you shipping some nv4x gpus to me which have the nv44 or nv30 3d class, as I have like 3 nv4x gpus, but all use the nv40 gpu class, but I was able to hit the issue regardless after some tweaking, but I think the general question is still valid :D
12:42 karolherbst: I have no fermi GPUs e.g.
12:45 mupuf: well, I definitely have a truck-load of GPUs
12:46 mupuf: I guess we'll need to review the ones you find of value, and we can ship that
12:46 karolherbst: yeah.. I will come up with a list, would be cool if you could do the same :D
12:47 mupuf: I pretty much have all the teslas and fermis
12:47 mupuf: Kepler, Pascal, and Turing are more spotty
12:47 karolherbst: yeah, but I have a bunch of those
12:47 mupuf: and I have some nv4B/4
12:47 mupuf: 43
12:47 karolherbst: in a meeting atm, so I will just get back with a list :D
12:48 mupuf: ack
20:30 felco: I'm solid guys =]
23:24 glennk: karolherbst, fwiw can confirm nv43 renders gnome a lot better with the nv30 index buffer fix
23:24 karolherbst: yay
23:25 glennk: also glxgears is down 40% due to the recent display list index buffer optimization
23:25 karolherbst: yeah...
23:25 karolherbst: we might want to make that optional
23:25 glennk: kind of a worst case test as its only clear + display lists
23:26 karolherbst: except for nv40 class GPUs there is no benefit of using index buffers
23:26 karolherbst: but... this being a regression on nv43 is also kind of... crappy
23:26 karolherbst: or does it matter more for more complex applications?
23:27 glennk: mileage will vary
23:32 karolherbst: annoying
23:36 glennk: huh, GL_ARB_color_buffer_float isn't advertised on nv43, the blob has it
23:37 karolherbst: glennk: send patches :p
23:38 karolherbst: imirkin_ might know what's missing
23:40 karolherbst: okay.. slowly tracking down this MT issue and it's very annoying indeed
23:47 imirkin_: karolherbst: yeah, support for float RT's is missing
23:47 imirkin_: other than that, everything's there.
23:47 imirkin_: ;)
23:47 imirkin_: float RT's have extra-special restrictions iirc
23:47 imirkin_: wasn't worth the hassle
23:50 imirkin_: among other things, i think no MSAA?
23:50 imirkin_: i don't remember precisely
23:50 imirkin_: would have to look through some notes
23:54 glennk: no blending i think was one of the restrictions
23:55 glennk: for 32 bit, 16 bit floats worked
23:55 imirkin_: yeah, that soudns right