ChanServ changed the topic of #panfrost to: Panfrost - FLOSS Mali Midgard & Bifrost - Logs https://oftc.irclog.whitequark.org/panfrost - <macc24> i have been here before it was popular
<alyssa>
dschuermann: Vvvvery roughly priority listed, not strictly though
<alyssa>
blog is number 1 though
<alyssa>
might push down scoreboarding...
<alyssa>
boolean rework is number 2 since it fixes a piglit, mumble
<alyssa>
between Valhall new dev and Bifrost maintainence, eh?
<alyssa>
Instruction selection on Valhall is more or less the same as Bifrost (roughly Valhall = Bifrost with the scheduler moved to hardware)
<alyssa>
...which means the vectorization stuff is all 100% needed on Valhall too, which makes it easy for me to bin with whichever arch is higher priority at the moment
moa has quit []
<alyssa>
HdkR: that was 90% of deqp-gles2 for whatever it's worth
<alyssa>
broken gles2 shouldn't make anyone ridiculously happy just yet ;)
<alyssa>
dschuermann: What *is* the status of vectorization stuff?
<alyssa>
If all you're blocked on is positive shader-db results on Mali and a Reviewed-by tag, I can up that priority to "as soon as I'm done with this darn blog post"
<HdkR>
alyssa: Of course people are greedy but even ES2 will make people happy :P
<alyssa>
HdkR: The last 10% of any test suite takes 90% of the work tbf
bluebugs has joined #panfrost
<HdkR>
Of course. I know that well
jambalaya has quit [Remote host closed the connection]
jambalaya has joined #panfrost
nlhowell has joined #panfrost
vstehle has quit [Ping timeout: 480 seconds]
<alyssa>
Sigh :)
JulianGro has quit [Remote host closed the connection]
camus has joined #panfrost
camus1 has quit [Ping timeout: 480 seconds]
<chewitt>
[ ] Show T628 some love. <= new list item ;)
<chewitt>
I'm tracking the Baikal Electronics folks but there's no eureka moment yet on XU4
hyrc has joined #panfrost
hyrc_ has quit [Ping timeout: 480 seconds]
camus1 has joined #panfrost
camus has quit [Ping timeout: 480 seconds]
Danct12 has quit [Ping timeout: 480 seconds]
vstehle has joined #panfrost
camus has joined #panfrost
camus1 has quit [Remote host closed the connection]
Danct12 has joined #panfrost
cyrozap-ZNC has quit []
cyrozap has joined #panfrost
MajorBiscuit has joined #panfrost
<dschuermann>
alyssa: you wrote "Patch looks good now". is that an r-b or did you want to check it again?
<dschuermann>
because then you could go ahead and test if it gives positive results to change the callback. if it regresses, you might want to test !12468
<dschuermann>
(and optionally !14480)
MajorBiscuit has quit [Quit: WeeChat 3.3]
MajorBiscuit has joined #panfrost
camus1 has joined #panfrost
camus has quit [Ping timeout: 480 seconds]
camus has joined #panfrost
camus1 has quit [Ping timeout: 480 seconds]
dianshi has joined #panfrost
MajorBiscuit has quit [Ping timeout: 480 seconds]
MajorBiscuit has joined #panfrost
Major_Biscuit has joined #panfrost
MajorBiscuit has quit [Ping timeout: 480 seconds]
nlhowell has quit [Ping timeout: 480 seconds]
nlhowell has joined #panfrost
nlhowell is now known as Guest478
nlhowell has joined #panfrost
Guest478 has quit [Ping timeout: 480 seconds]
camus1 has joined #panfrost
camus has quit [Ping timeout: 480 seconds]
<alyssa>
no new list items, thanks.
<alyssa>
dschuermann: That was a "the code looks right but I'm holding off on r-b until I can actually test it myself"
<dschuermann>
kk
<alyssa>
and a "if this is happy in shader-db and CI on panfrost, you have my r-b"
nlhowell has quit [Quit: WeeChat 3.1]
nlhowell has joined #panfrost
camus has joined #panfrost
camus1 has quit [Read error: Connection reset by peer]
Major_Biscuit has quit [Ping timeout: 480 seconds]
camus has quit [Remote host closed the connection]
camus has joined #panfrost
Major_Biscuit has joined #panfrost
<alyssa>
Apparently reviewing my Twitter feed is a fertile source of blog post material..
<macc24>
if it's not mediatek that'd be "fortunately"
<macc24>
OH
<macc24>
oh no
<macc24>
i just got it
<urja>
?
<alyssa>
reverse-engineering checklist
<alyssa>
- linkage, especially with gl_pointsize and mediump and things, I think I'm missing some details still
<alyssa>
- textures, look harder since I think I botched something
<macc24>
alyssa: FYI i don't have anything valhall... yet.
<alyssa>
1. gl_PointSize is /not/ included in the output/input size calculations, whether points are drawn or not
<alyssa>
psiz array format is only set when drawing points
<alyssa>
culling is disabled when drawing points
<alyssa>
depth bias is disabled when drawing points
<alyssa>
unk 26/27/28 has more nonsense but I'm not sure if that's spurious or not
<alyssa>
unk 26 = 0x10001
<alyssa>
unk 27 = 0x5F
<alyssa>
unk 28 = 0x180028
<alyssa>
18,28 were the interpolation counts
anholt has quit [Remote host closed the connection]
<alyssa>
5f00010001 is a tagged pointer into the tiler heap, I think
anholt has joined #panfrost
<alyssa>
not sure if this is just a timing difference with the hardware, there
<alyssa>
Apparently, no support for FP32 point size -- DDK inserts a conversion if necessary
<alyssa>
(Or if supported, it's slower than the extra ALU. that one is likely.)
<alyssa>
vary stores should have offset:0 to start
<alyssa>
all stores in position shader should be offset 0, the mechanism I documented seems correct
JulianGro has joined #panfrost
<alyssa>
2. textures seem fine but I just grabbed a more nuanced dump in the hopes of finding something I missed the first 3 times around
<alyssa>
gah I have too many devices that tether to one another
<alyssa>
work laptop, personal laptop that boots the mac mini, mac mini that boots the valhall chromebook, the valhall phone, the personal phone to tether when internet isn't available ...
<macc24>
networking ;)
<urja>
lmao be careful not to create a bootstrap circle
<alyssa>
urja: chromebook booting mac booting chromebook is already pretty bad
<alyssa>
two different chromebooks, but
<macc24>
alyssa: port that valhall machine to cadmium and install mainline linux kernel straight into emmc ;)
<alyssa>
I could install straight to the emmc trivially, I just don't want to because I want to keep chromeos
<alyssa>
and modify kernel easily
<macc24>
well, how you boot it currently?
<alyssa>
external usb disk
<alyssa>
with kernel written by the mac
<alyssa>
and serial cable connected to the mac
<macc24>
disk?
<urja>
are you moving the usb stick or have you optimized that to the point of using an USB KVM or such? :P
<alyssa>
urja: It's an external disk with a female usb-c port
<alyssa>
so I just move the cable
<macc24>
urja: even i don't do that xD
<alyssa>
still not great but meh I'm making do with the parts I have sitting in my desk from previous projects :-p
<macc24>
for a long long time i used a literally cheapest 32gb pendrive i could find
<anholt>
you don't just use a servo cable and netbooting? so much nicer than fussing with disks
<alyssa>
anholt: I was told this requires fussing with depthcharge
<macc24>
^ it does iirc
<alyssa>
also, I just got the suzyq less than a week ago, I'm still new :)
<anholt>
you can find a custom depthcharge that always netboots, but ^N at the right time should always work I think.
<macc24>
alyssa: let's swap projects, i'll do panfrost for a while and you'll do cadmium now xDD
* macc24
yeets all her 4 chromebooks at alyssa
nlhowell is now known as Guest511
nlhowell has joined #panfrost
Guest511 has quit [Ping timeout: 480 seconds]
<alyssa>
with a bunch of hacks, some texture tests are passing
<alyssa>
apparently missing a few bits in the texture instructions (whoops), and screwed up the surface descriptor
<alyssa>
and yes ok I see how to do points now
camus has quit [Remote host closed the connection]
camus has joined #panfrost
camus1 has joined #panfrost
camus has quit [Read error: Connection reset by peer]