<ente>
I'm sure all your code is great and reads like a poem
<ente>
which is the intent of code
<ente>
good night!
<shadeslayer>
I'd rather not have my code read like poems, that would open it up to subjective interpretation
<icecream95>
shadeslayer: Compilers do that anyway, if you use too many optimisation flags...
<alyssa>
shadeslayer: UB
<alyssa>
wait icecream95 beat me to it :P
<shadeslayer>
hahaha
<shadeslayer>
I'm a bit pissy this evening because the BO unmapping code doesn't go through a kernel ioctl, meaning I have to figure out a different way to trigger BO mapping and unmapping ftrace events :C
<alyssa>
i write compilers don't look at me :C
stikonas has quit [Remote host closed the connection]
<shadeslayer>
yeah, and, the profiling tools from nVidia and Apple make me jealous, Linux can do prettier graphs than Apple!
<icecream95>
Tip for saving registers in assembly: You can use pc to store a couple of bits (duplicate your routine as needed)
<HdkR>
Alternatively, use the tagged bits that live in the upper bits of the address
<HdkR>
Gives you 4,8, or 16 bits depending on hardware configuration :P
<chewitt>
you'd hope that the exynos_defconfig enabled everything :)
<chewitt>
time for a diff-ing evening..
<robmur01>
yuck, looks like it's an MFD mess, so it might just be a case of getting modules in the right places in the right order
<urjaman>
without knowing anything about the XU4 my plan would be 1. find the regulator compatible strings in the device tree, 2. find which drivers provide them (grep for them in source), 3. make sure they're built-in in the config
<urjaman>
(if they already are, then something more funky going on ..)
<robmur01>
then you have to recurse into all the I2C, GPIO, etc. drivers that the regulator itself depends on ;)
<urjaman>
ah true
buzzmarshall has joined #panfrost
<robmur01>
sometimes if you're really lucky you can end up with all the drivers loading in an order such that nothing actually kicks the rest of the deferred probe list
<robmur01>
then you have to modprobe something random from userspace to un-stick it
adjtm has quit [Remote host closed the connection]
adjtm has joined #panfrost
adjtm has quit [Remote host closed the connection]
adjtm has joined #panfrost
<chewitt>
VDD_G3DS is the regulator
<chewitt>
which resolves to max77686, which is enabled in defconfig
<chewitt>
hmm
<chewitt>
break out the funk
<robmur01>
does /sys/kernel/debug/devices_deferred show anything besides the GPU waiting?
<robmur01>
weird... another thing to try is just giving it a manual kick in case the dependencies did happen to show up in the meantime - something like "echo 11800000.gpu > /sys/bus/platform/drivers/panfrost/bind"
<chewitt>
urgh.. I have so much [ 2229.479392] mmc_host mmc0: Bus speed (slot 0) = 50000000Hz (slot req 400000Hz, actual 396825HZ div = 63) garbage in dmesg it's hard to see anything
<chewitt>
need to find a way to disable thaat
<robmur01>
hmm, are you on an older kernel version? I have a vague memory of dwmmc being really spammy with clock messages on rockchip boards for a while, but that was probably a year ago
<chewitt>
5.7-rc3
<robmur01>
OK, probably not the same thing then
<alyssa>
:q
<robmur01>
dG:w!q
<robmur01>
aww, the universe is all still here :(
paulk-leonov has quit [Excess Flood]
nerdboy has joined #panfrost
nerdboy has quit [Ping timeout: 246 seconds]
ente has joined #panfrost
adjtm_ has joined #panfrost
adjtm has quit [Ping timeout: 265 seconds]
<HdkR>
"GPU – Arm “Natt” GPU" Don't mess with me with your naming Rockchip
<alyssa>
HdkR: Natt = G31 iiirc?
<HdkR>
cnx-soft coverage is a bit weird with it. `Arm “Natt” GPU (i.e. NOT Mali-G31 MP2),`