12:58RSpliet: imirkin: Do you know where this bit of smart intelligent cleverness is that makes the IMM param the second one of the lot for nv50? is that legalise? I can't find it :p
12:59RSpliet: but it seems to be done even before a pass of optimisation
12:59RSpliet: despite MAD not having IMM support defined in the initProps at all (because... that silly rule where SDST == SSRC2)
13:03RSpliet: calim: ...mind playing imirkin for me while he's gone? is there some bit of "magic" I can rely on for doing that?
13:32calim: RSpliet: TargetNV50::insnCanLoad
13:33calim: I know, it's not completely obvious why using an immediate would be classified as loading
13:34RSpliet: calim: I don't know the compiler quite well enough to judge that
14:07calim: oh, right, I wanted to say that some optimization pass automatically does it if insnCanLoad reports true
14:18RSpliet: calim: you mean "folding" the imm into MAD? yeah, I know, but it's only allowed if SDST == SSRC2; which isn't clear until after RA
14:19RSpliet: or are you just talking about swapping the sources to get IMM into SRC1, in which case: thanks, that's precisely what I was curious about :)
14:24RSpliet: oh right, brilliant, that pass does just that
14:26RSpliet: now I just need to find a way to trick RA into believing SRC2 == SDST is the preferred reg assignment whenever liveness allows
14:26RSpliet: surely I can have some fun with that!
15:04barry_UTD: how could I find my nvidia pci-id?
15:06barry_UTD: 02:00.0 VGA compatible controller: NVIDIA Corporation GF108 [GeForce GT 630] (rev a1)
15:06barry_UTD: 02:00.1 Audio device: NVIDIA Corporation GF108 High Definition Audio Controller (rev a1)
15:06barry_UTD: which one is that then?
15:07tobijk_: you got 1 guess :P
15:07barry_UTD: tobijk_: 02:00.0?
15:09barry_UTD: tobijk_: so you mean for vbois dump I have to go to /sys/bus/pci/devices/0000:02:00.0/rom
15:09tobijk_: i'd guess so
15:10tobijk_: alternatively /sys/class/drm/cardX
15:10tobijk_: but i'm not sure