chewitt has quit [Read error: Connection reset by peer]
raster has joined #panfrost
chewitt has joined #panfrost
chewitt has quit [Quit: Zzz..]
chewitt has joined #panfrost
stikonas has quit [Remote host closed the connection]
raster has quit [Read error: Connection reset by peer]
<HdkR>
bwarp
afaerber has joined #panfrost
<hanetzer>
prawb
marcodiego has joined #panfrost
chewitt has quit [Read error: Connection reset by peer]
<hanetzer>
hmmm.... tracepoint
Kwiboo has joined #panfrost
chewitt has joined #panfrost
chewitt has quit [Read error: Connection reset by peer]
chewitt has joined #panfrost
stikonas has joined #panfrost
chewitt has quit [Read error: Connection reset by peer]
<hanetzer>
robher:hey, think you could gimme a copy of those logs again?
<HdkR>
Lyude: Setting up super basic clause emitting just to get it out of the way so I can do further testing
<HdkR>
One fma or one add per bundle/clause thing :P
<Lyude>
HdkR: just dummy clauses or?
<HdkR>
actual clauses, just not packed so I don't have to deal with scheduling
<HdkR>
Super inefficient but eh
<Lyude>
it it works!
<Lyude>
*if it
<HdkR>
That's what I'm hoping for :P
<HdkR>
Disassembler is happy at least. Just need to encode the register ports correctly
<Lyude>
sorry I haven't had the time to get to ra stuff yet
<Lyude>
(will this include ra?)
<HdkR>
No problem. It'll be the most basic RA in the world, won't even be using T0/T1
<Lyude>
(also brb, rebooting server)
Lyude has quit [Quit: WeeChat 2.2]
Lyude has joined #panfrost
<HdkR>
Port encoding working. Now to make some tea and do some dead simple RA
<HdkR>
FMA 3 source I skipped since I don't want to deal with one of the sources having to be the destination :P
<HdkR>
er, 4 source even
* HdkR
watches .NET runtime optimization service steal a few cores
<HdkR>
I hate Windows so much
<hanetzer>
yep. sucks
<HdkR>
Hoping I won't have to deal with it soon
<HdkR>
Basic RA shamelessly stolen from the midgard compiler and ...working?
<hanetzer>
heh
<hanetzer>
so, I'm thinking to do a bit of TRACE_EVENT sprinkling in the panfrost kernel module
<HdkR>
What does TRACE_EVENT do? Profiling?
<hanetzer>
if you check out the dwc3 kernel driver (which the rk3399 uses) you'll get an idea of it. basically you can do a lot of cool stuff with it, including wrapping readl/writel calls
<hanetzer>
those'll show up in a /sys/kernel/debug/tracing path if you flip certain .config and module parameters, plus runtime setting via same sysfs interface
<HdkR>
I mean. It's specifically trace instrumentation of kernel functions right? :P
<hanetzer>
yeah
<hanetzer>
pretty extensible as well, apparently. the dwc3 driver uses it for readl/writel tracing, as well as usb specific stuffs
<hanetzer>
eg, urbs, control packets, etc
<HdkR>
Always nice to have. can show problem areas when you're spamming functions too heavily
<HdkR>
Is it number of times the event is hit or does it also include time within that scope?
<hanetzer>
as in, timestamps? it does include those, and other stuff. can do custom printf strings to print exacly what data you want for the most part :)
<HdkR>
well, duration
<HdkR>
enter at n time stamp, leave at n+m timestamp
<HdkR>
I'm used to number of events and duration. Allows you to see the most important bits depending on what you care about at any given time :D
<HdkR>
Next step! Trying to get kmscube or es2tri shaders emitting sane code