stikonas has quit [Remote host closed the connection]
<HdkR>
Lyude: So there are four major components now. NIR optimization(using default optimizations atm); NIR to machine IR emission; Instruction scheduling; Instruction emission. Scheduling and emission are living in their own files. NIR optimization and NIR->MIR are in the "main" compiler file.
<HdkR>
Will clean up more and push tonight or tomorrow if I'm not happy with it by tonight :P
<Lyude>
yesss
<HdkR>
I'm not happy with the MIR that was generated via copy and pasting. Going to fix it first
<robertfoss_>
Lyude: good to know!
<robertfoss_>
robher: thanks :)
_whitelogger has joined #panfrost
Elpaulo has joined #panfrost
_whitelogger has joined #panfrost
ph5 has joined #panfrost
<cwabbott>
HdkR: I can take a look at it this weekend maybe
<cwabbott>
I did have some thoughts about how I would do things
<cwabbott>
I'd have the pre-clause IR, which consists of a list of instructions that can be either ADD, FMA, or both (decomposed into FMA and ADD after clause bundling)
<cwabbott>
the both category is for stuff like swapping two registers (needed for SSA-based register allocation)
<cwabbott>
and also, a few things where FMA and ADD have to be scheduled together, like the 64-bit addition stuff
<cwabbott>
altogether it would do 1) pre-RA scheduling 2) SSA-based allocation 3) post-RA scheduling/clause bundling
<cwabbott>
and then the analysis for clause dependencies and data write barriers
stikonas has joined #panfrost
raster has joined #panfrost
<cwabbott>
Lyude: btw, I already figured out the 64-bit register field format... turns out that their SPIR-V compiler, even though it doesn't claim float64/int64 support, will happily accept them and produce... something
<cwabbott>
not clear how much of that something would run on a real device
<cwabbott>
I pushed to mali-isa-docs and the spd
Elpaulo has quit [Quit: Elpaulo]
Elpaulo has joined #panfrost
stikonas has quit [Remote host closed the connection]
raster has quit [Remote host closed the connection]
NeuroScr has quit [Read error: Network is unreachable]
NeuroScr has joined #panfrost
marcodiego has joined #panfrost
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
cwabbott has quit [Client Quit]
cwabbott has joined #panfrost
cwabbott has quit [Quit: cwabbott]
cwabbott has joined #panfrost
cwabbott has quit [Client Quit]
cwabbott has joined #panfrost
belgin has joined #panfrost
belgin has quit [Quit: ERC (IRC client for Emacs 24.5.1)]
<Lyude>
cwabbott: cool!
adjtm_ has joined #panfrost
adjtm has quit [Ping timeout: 244 seconds]
<Lyude>
time to setup all of my systems today! :)
<HdkR>
woo
<Lyude>
lvrp16: poke; is there a guide somewhere on how to setup this rk3399 with it's case?
<davidlt>
expect of course inside, inside it's plastic
<davidlt>
just the outside shell is magnesium alloy
<davidlt>
bu that's typical
<hanetzer>
but yeah. ath9k_htc can have open firmware built for it
* davidlt
off
<HdkR>
Neat
<hanetzer>
yep
<HdkR>
ath9 is a couple generations old right?
<hanetzer>
more or less. but then again, folks stick with whatever old intel cpu is in those thinkpads over freedom so old wifi equip isn' much diff :)
<HdkR>
Oh, wikipedia actually has an article on wifi open source
<hanetzer>
half a mind to steal one of the pcie wifi adapters inside one of my meraki mr24s and using them elsewhere :)
<HdkR>
Oh, apparently there is an open firmware for QCA USB wifi chips written by QC. Neat