ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
mclasen has quit [Ping timeout: 480 seconds]
dviola has joined #dri-devel
dviola has quit []
<graphitemaster>
So you know how I'm always like hay, lets fix GPU preemption on Linux and I was like WebGPU pretty serious stuff, imagine locking up a desktop with a simple website. Well macOS has the same issues and the air's have a software power button so you can't even power cycle them ... https://twitter.com/DasSurma/status/1495096911333842946
<graphitemaster>
So now you just got to wait for the batteries to die when a website decides to spawn a bunch of compute work or ... do something dumb like an infinite loop
<HdkR>
Neat
<bnieuwenhuizen>
disconnect the battery? Or is that not possible on macbooks either?
<HdkR>
You'd have to open the shell and pull the cable, which isn't really something a normie would want to do :P
<graphitemaster>
GPU going to become one of the greatest attack surfaces in computing since ActiveX for IE once WebGPU is supported everywhere.
<HdkR>
Oh, and I think they use Pentalobe bits on their macbooks now, so the likelyhood of having the bit driver to remove those is a bit unlikely
<FLHerne>
my Samsung is almost like that, but after the second time I discovered it has a paperclip-through-hole 'reset button' on the underside
<bnieuwenhuizen>
graphitemaster: can't you have infinite loops in webgl already?
<FLHerne>
no idea what's wrong with the usual hard-button-after-10-seconds thing
<bnieuwenhuizen>
FLHerne: needs the firmware for the timer probably. What are you going to do when the firmware got stuck?
<FLHerne>
See, if I was designing it there'd just be an RC timer circuit attached to the button
<FLHerne>
but that would add 1p to the hardware BoM, and this is why They don't let me design consumer electronics :-(
sdutt has joined #dri-devel
<graphitemaster>
bnieuwenhuizen, believe it or not, long running graphics work loads are just killed with a watchdog thing, still not ideal
<graphitemaster>
compute workloads just run on mac until completion
<graphitemaster>
Sort of like Linux actually, where I observed the same behavior
Haaninjo has quit [Quit: Ex-Chat]
JohnnyonFlame has quit [Read error: Connection reset by peer]
cworth has quit [Ping timeout: 480 seconds]
pcercuei has quit [Quit: dodo]
cworth has joined #dri-devel
<robclark>
"according to math" I don't think battery would last for 72+hrs with gpu max'd ;-)
co1umbarius has joined #dri-devel
columbarius has quit [Ping timeout: 480 seconds]
LexSfX has quit []
LexSfX has joined #dri-devel
LexSfX has quit []
LexSfX has joined #dri-devel
camus has joined #dri-devel
pendingchaos_ has joined #dri-devel
sdutt has quit [Ping timeout: 480 seconds]
pendingchaos has quit [Ping timeout: 480 seconds]
mhenning has quit []
Company has quit [Quit: Leaving]
<imirkin>
dcbaker: what's the simplest way to tack -gdwarf-4 to all c/c++ compilation?
<imirkin>
(looks like my toolchain has suddenly started using dwarf5, despite me not being able to think of what changed ... and the gdb i have on the target machine only does dwarf4)
<dcbaker>
No
<dcbaker>
I think gcc and clang both changed to dwarf 5 recently by default
<dcbaker>
*np
<imirkin>
yeah, but i didn't think i updated my cross-gcc. but clearly i've done _something_. whatever =/
<imirkin>
and updating gdb on the arm board has failed in tears
<imirkin>
all good now though :)
<HdkR>
It's fun when you hit a feature in dwarf5 that gdb or lldb doesn't support
<HdkR>
Since it doesn't support all features of dwarf5 yet :|
mbrost has joined #dri-devel
<imirkin>
HdkR: well, it just says "version 5 aaaaaaah, no line numbers for you"
<HdkR>
I like the one where it's "I don't understand this symbol, assert for you."
<maxzor>
i'm just inquiring about .debug_addr... but there is the continent of dwarf vendor extensions which I'm afraid might be even in a worse state
adjtm has quit [Quit: Leaving]
linearcannon has joined #dri-devel
pcercuei has joined #dri-devel
luckyxxl has joined #dri-devel
Company has joined #dri-devel
Haaninjo has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
rasterman has joined #dri-devel
tango_ is now known as Guest10
tango_ has joined #dri-devel
Guest10 has quit [Ping timeout: 480 seconds]
flacks has quit [Quit: Quitter]
flacks has joined #dri-devel
rasterman has quit [Quit: Gettin' stinky!]
luckyxxl has quit [Quit: bye]
cafuffu has joined #dri-devel
cafuffu has quit [Read error: No route to host]
cafuffu has joined #dri-devel
gouchi has quit [Remote host closed the connection]
libv_ has joined #dri-devel
libv has quit [Ping timeout: 480 seconds]
adjtm has joined #dri-devel
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
cafuffu has quit [Read error: No route to host]
cafuffu has joined #dri-devel
adjtm is now known as Guest17
adjtm has joined #dri-devel
cafuffu has quit [Ping timeout: 480 seconds]
Guest17 has quit [Ping timeout: 480 seconds]
kts has quit [Quit: Konversation terminated!]
pcercuei has quit [Quit: brb]
pcercuei has joined #dri-devel
<graphitemaster>
I'm curious about something. Since wave programming has become part of the programming model, what contingency plans are in place for compat in the future?
<graphitemaster>
Like, all those games out there assuming NV = 32, AMD = 64
<graphitemaster>
Are they just going to break terribly on future hardware that has like 256 waves or some shit
<graphitemaster>
Like drivers are going to have to emulate the behavior of a specific wave count on given hardware just to avoid breaking shaders that make assumptions of the subgroup size
<graphitemaster>
I wonder if anyone has asked this question when they decided to make this implementation detail part of the programming model
<graphitemaster>
Only time will tell but this decision could've been a mistake :|
kts has joined #dri-devel
<maxzor>
HSA Platform System Architecture Spec: "Wavefront size: Every call convention of every kernel agent of an implementation must have a
<maxzor>
wavefront size that is a power of 2 in the range from 1 to 256 inclusive"
<FLHerne>
Surely anything > 64 would be terribly inefficient for game workloads?