00:32 imirkin: anholt: do you know offhand which chips have been tested with the NIR backend and what that testing was?
00:33 imirkin: i'm esp concerned about nir's support for nv50 compute
00:34 imirkin: some stuff has to be done carefully, e.g. ensuring that address regs are used for shared mem indexing, etc
00:36 imirkin: anholt: i was going to do testing on the nir-to-tgsi variant ... should i be testing the nir backend instead?
00:37 imirkin: with glsl-to-tgsi going away, i'm probably going to bow out of caring about nouveau completely, so i'm also happy to go with the flow.
00:37 imirkin: [still some display things i want to try to figure out, but that's completely separate]
00:48 karolherbst: imirkin: I could throw CL at nv50
00:53 imirkin: karolherbst: that might be too much, but that'd be fine. i'm more looking for ES 3.1
00:54 imirkin: if it's pre-nva3, you'll have to force-enable ES 3.1, but most of the compute stuff should work
00:54 karolherbst: ahh right, we do compute there as well
00:54 imirkin: some features won't, but i can point out which ones aren't there
00:54 karolherbst: yeah. I can test on the nv50+ GPUs I have over the weekend
00:54 karolherbst: just run CTS+piglit whatever tests with TGSI and with NIR and see where differences are or something?
00:55 imirkin: karolherbst: i've been primarily testing CTS + dEQP, not so much piglit -- piglit is more geared to desktop, which won't work there
00:55 imirkin: (for compute)
00:55 imirkin: the deqp-runner configs are checked in for gt215, feel free to use those as a refernece
00:57 karolherbst: okay
04:45 anholt: imirkin: I've done piglit, GLES, and GL CTS on gm206b and nv92 with both the NIR and NTT-TGSI paths, and NIR's the clear winner on tests.
12:29 imirkin: anholt: ok. i'd really like to see a dEQP-GLES31 test on either nva3+ or on the nv92 you have with ES 3.1 force-enabled. some things will fail on there, but that's fine - features are just missing.
19:52 anholt: imirkin: ok, can do.
22:09 anholt: maybe not. the test box is not recognizing its rootfs.
23:10 imirkin: anholt: doh!
23:11 imirkin: anholt: i do have one plugged in, and it appears that it's friday. perhaps i'll have some time over the weekend
23:11 imirkin: last weekend was lost to vacation.
23:15 imirkin: anholt: not sure if you saw my messages, you seem to have timed out
23:16 imirkin: anholt: but it's possible i'll be able to do some test runs over the weekend