02:28 albamart: i am not entirely sure, some pdf's not read yet, mainly keplers one i had queued up, they had some optimizations in hw, i did not look too close
02:29 albamart: it is based of abondening full scoreboard in favor of something more efficient or thinner
02:32 albamart: which was power in itself keeping allready power usage somewhat lower, but it can probably yet be lowered
02:45 albamart: http://meseec.ce.rit.edu/722-projects/spring2015/3-2.pdf
02:51 albamart: but probably it's something others know allready, since the card works on nouveau, this arch review claims, that sched info opcodes also fill in some hw dependency tracking block
03:01 albamart: no extensive info on it to non-devs, but probably it is not quite sched latency opcode, but rather something used for determining the natural order of execution, so compiler reorders the instructions allegedly
03:18 albamart: it is wiser tactics in fact indeed, so when compiler queues things that have lower latency one first bank conflict arbitration then gives probably preemptive rights to the first instruction, while the longer latency one is parked to wait
03:22 albamart: copying the case switches based of nvidia blob gave them in correct order to make it work naturally, cause that operand collector would detect persistent thread read and writeback conflicts in hw
03:23 albamart: hence 25percentage of power save is right away actual, but of course this marginal compared to "real code" :D
03:39 albamart: hence yeah NVIDIA implements axel davy's style of scheduling, which is proper for ILP opportunities, and also , hehe, well what i name my style of scheduling or davids -- which is also ILP but also runahead i-buffer fetch decode with added regrouped warps from idling stuff in exec stage
03:41 albamart: the stuff what i talked about and have been doing for some while is to implement tarjans style of scheduling in sw -- which is largely composed of different ideas from web by me on AMD radeon too (tarjan though did it differently in hw instead, but mentioned that can be done in sw without specifying how)
03:43 albamart: and as wise reader has understood, compared to axel's code , it has fewer complexity -- but is much more efficient
11:28 john_cephalopoda: Hey
12:19 pmoreau: marmistrz: pong
14:24 mastermart: so make no mistake about idiots(1) which is the manual for clueless monkeys like imirkin and mwk, we have another dane on osdev who belongs to that group too
14:26 mastermart: muhahaa, i used his own words, such a monkey managed to say that i do not make sense
14:32 mastermart: generally world is full of scientists but this arrogant and clueless combination from mit who parks in the middle of the road here with his shitty patches, is just an average -- i am surry to insult monkeys , but that is the way it is
14:33 mastermart: i just try to give a lecture about it, they say work made human from a monkey
17:01 jayhost: This channel gets the strangest trolls
17:06 imirkin: just the one.
18:20 anEpiov: does linux got a way to check gfx ram usange such as 'free'?
18:20 imirkin: generically, no
18:20 anEpiov: wow, after 30 years ...
18:35 Asu: jayhost: damn, i think we had the same troll on #osdev for a while
18:57 Tom^: Asu: hes been in here from time to time, for years. :/
18:58 Asu: literal years? i'm sorry for you :D
18:58 imirkin: many many many years
18:58 imirkin: maybe 10? not sure.
18:58 imirkin: at least like 5 or so.
19:00 imirkin: pain in our collective asses...
19:11 jayhost: Must be the same guy I saw years ago who was acting like someone had screwed him over, assuming he's schizophrenic
19:13 anEpiov: imirkin: you're not talking about me right? I've only used nouveu for 2 years.
19:14 imirkin: jayhost: yes, i think he is
19:14 imirkin: anEpiov: no, not about you. although you've definitely gotten to the point where i seriously weigh the pros and cons of responding to you.
19:15 anEpiov: :'(
19:15 imirkin: but you don't post walls of semi-sensical text.
19:18 anEpiov: imirkin: I am just a nouveau user and report when things go awry to improve it.
19:18 anEpiov: imirkin: as a matter fact, some time ago I when all the trouble of uploading debuging info and all that.
19:18 imirkin: anEpiov: yeah, that's what i kinda figured. when is why i do tend to respond to your more serious questions.
19:19 imirkin: anEpiov: however it feels like your starting assumption is that we, the group of volunteers who occasionally have time to put towards nouveau, owe you a perfectly working driver
19:20 anEpiov: oh, about the 'free' utility? I didn't mean nouveu specific but linux wide.
19:20 imirkin: in general, from all your various comments.