alanc has quit [Remote host closed the connection]
alanc has joined #dri-devel
ycs has quit [Remote host closed the connection]
ycs has joined #dri-devel
The_Company has joined #dri-devel
Company has quit [Ping timeout: 480 seconds]
ycs has quit [Remote host closed the connection]
ycs has joined #dri-devel
ycs has quit [Ping timeout: 480 seconds]
ycs has joined #dri-devel
ycs has quit [Ping timeout: 480 seconds]
feaneron has quit [Ping timeout: 480 seconds]
Daanct12 has joined #dri-devel
feaneron has joined #dri-devel
Leftmost has quit [Read error: Connection reset by peer]
Leftmost has joined #dri-devel
nerdopolis has quit [Ping timeout: 480 seconds]
Mangix_ has quit []
Mangix has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
vert4564557[m] has joined #dri-devel
The_Company has quit []
feaneron has quit [Ping timeout: 480 seconds]
sguddati has joined #dri-devel
kts has joined #dri-devel
Duke`` has joined #dri-devel
fab has joined #dri-devel
fab has quit []
kzd has quit [Ping timeout: 480 seconds]
mattst88 has quit [Ping timeout: 480 seconds]
bonzini has joined #dri-devel
DemiMarie__ is now known as DemiMarie
kts has quit [Quit: Konversation terminated!]
Duke`` has quit [Ping timeout: 480 seconds]
bonzini has quit [Ping timeout: 480 seconds]
Shibe has quit []
glennk has joined #dri-devel
sally has joined #dri-devel
kts has joined #dri-devel
sally has quit []
sally has joined #dri-devel
sally has quit [Remote host closed the connection]
sally has joined #dri-devel
dolphin has joined #dri-devel
sally has quit []
sally has joined #dri-devel
sally has quit []
sally has joined #dri-devel
sally has quit []
sally has joined #dri-devel
kts has quit [Quit: Konversation terminated!]
sally has quit [Remote host closed the connection]
sally has joined #dri-devel
mriesch has quit [Remote host closed the connection]
mriesch has joined #dri-devel
sima has joined #dri-devel
jsa1 has joined #dri-devel
coldfeet has joined #dri-devel
rasterman has joined #dri-devel
kts has joined #dri-devel
kts has quit []
tzimmermann has joined #dri-devel
sally has quit []
sally has joined #dri-devel
vliaskovitis has joined #dri-devel
fab_ has joined #dri-devel
fab_ is now known as Guest25816
chewitt has joined #dri-devel
sghuge has quit [Remote host closed the connection]
sghuge has joined #dri-devel
Guest25816 has quit []
fab__ has joined #dri-devel
sally has quit []
sally has joined #dri-devel
fab__ is now known as Guest25818
sally has quit []
sally has joined #dri-devel
phasta has joined #dri-devel
coldfeet has quit [Quit: Lost terminal]
sarnex has quit []
sarnex has joined #dri-devel
sguddati has quit [Ping timeout: 480 seconds]
alane has quit []
alane has joined #dri-devel
lynxeye has joined #dri-devel
sguddati has joined #dri-devel
kts has joined #dri-devel
apinheiro has joined #dri-devel
jkrzyszt has joined #dri-devel
alarumbe has joined #dri-devel
kts has quit [Quit: Konversation terminated!]
sguddati has quit [Ping timeout: 480 seconds]
sguddati has joined #dri-devel
Mangix_ has joined #dri-devel
Mangix has quit [Remote host closed the connection]
alane has quit [Remote host closed the connection]
zsoltiv____ has joined #dri-devel
lsntvt has joined #dri-devel
zsoltiv___ has quit [Ping timeout: 480 seconds]
alane has joined #dri-devel
marcf has quit [Ping timeout: 480 seconds]
marcf has joined #dri-devel
sguddati has quit [Ping timeout: 480 seconds]
yshui has joined #dri-devel
yshui_ has quit [Ping timeout: 480 seconds]
lsntvt_ has joined #dri-devel
lsntvt has quit [Read error: Connection reset by peer]
sguddati has joined #dri-devel
SquareWinter68 has joined #dri-devel
xroumegue has quit [Ping timeout: 480 seconds]
xroumegue has joined #dri-devel
OftenTimeConsuming has joined #dri-devel
Guest25818 has quit []
ity has joined #dri-devel
sguddati has quit [Ping timeout: 480 seconds]
feaneron has joined #dri-devel
fab has joined #dri-devel
guludo has joined #dri-devel
coldfeet has joined #dri-devel
coldfeet has quit []
pepp_ has quit []
pepp has joined #dri-devel
cmichael has joined #dri-devel
dsimic is now known as Guest25837
dsimic has joined #dri-devel
Guest25837 has quit [Ping timeout: 480 seconds]
<zmike>
eric_engestrom: when is next branchpoint? I didn't see in docs
Daanct12 has quit [Quit: WeeChat 4.7.1]
OftenTimeConsuming has quit [Remote host closed the connection]
nerdopolis has joined #dri-devel
OftenTimeConsuming has joined #dri-devel
davispuh has joined #dri-devel
Company has joined #dri-devel
mattst88 has joined #dri-devel
dolphin has quit [Quit: Leaving]
rasterman has quit [Quit: Gettin' stinky!]
<simon-perretta-img>
The demote_to_helper_invocation spec states that stores should be suppressed, but doesn't specify how that applies to atomics; is it safe enough to assume it follows the gl_HelperInvocation spec (atomic op return values are undefined), or should atomic ops essentially turn into memory loads?
<glehmann>
undefined helper return values should be fine
<glehmann>
amd drivers implement it like that at least
<alyssa>
simon-perretta-img: undefined, but I don't have a spec citation
<simon-perretta-img>
Cool cool, thanks both!
<alyssa>
(undefined value, not undefined behaviour, although that's actually an interesting question.)
<alyssa>
anyway uh we have a nir pass for this
<alyssa>
predicate_helper_writes or something like that
<alyssa>
should work for demote too if you do the right magic sequence
lumidify has quit [Ping timeout: 480 seconds]
<alyssa>
or maybe im thinking of terminate
<alyssa>
idk, just copy whatever i did in my powervr compiler, this works there
kzd has joined #dri-devel
<simon-perretta-img>
To be honest I was considering handling it in the backend instead, replacing the store addresses in demoted/helper invocations and null buffer descriptors with a dummy buffer, just saves on a bit of control flow
<alyssa>
it's allowed
bolson has joined #dri-devel
<alyssa>
on AGX, assuming the r/e cycle count table is accurate, the if-endif isn't anymore expensive than the pile of bcsel's
<alyssa>
so I didn't bother
<simon-perretta-img>
Fair
haaninjo has joined #dri-devel
<alyssa>
Faith and I have talked about adding some sort of predicated load/store stuff to NIR
<alyssa>
I'm not sure what that would look like yet
<alyssa>
but it seems like "@store_if(value, address, condition 1-bit bool)" would be useful to a bunch of backends
<alyssa>
and would let nir_opt_peephole_select be a lot more effective
<alyssa>
the trap here is that IDK the right level of generality
<alyssa>
do we want to be able to predicate any intrinsic? probably not
<glehmann>
I wouldn't worry about it during bring up, stores/atomics in FS are not that common
<alyssa>
do we want to add 50 new intrinsics that are just predicated variants of existing ones? probably not
<alyssa>
do we want to add a new source to 50 existing intrinsics? probably not
<alyssa>
I think NAK has a backend version of peephole select using hw predication (?)
<simon-perretta-img>
Makes sense yeah, that'd be useful for us too but I see the generality issue
<alyssa>
but doing it in NIR pre-RA would bring benefits for scheduling (bigger basic blocks & less control flow)
yshui_ has joined #dri-devel
<alyssa>
but like... atomics are uncommon & heavyweight anyway
<alyssa>
so if we just did load_global/store_global that would cover probably 80% of compute shader use cases in practice
<alyssa>
(and then replacing a pile of bcsel with hw predication is probably doable.)
<alyssa>
(doable in backend post-RA i mean)
yshui has quit [Ping timeout: 480 seconds]
<simon-perretta-img>
That sounds like it'd be a good first step, yeah; beyond that, the other sorts of ops I can think of that could be useful to use predication on are basically already in that form (demote/terminate_if, etc.)
kts has joined #dri-devel
<alyssa>
right
<simon-perretta-img>
I feel like going even more general with predication could make sense as part of e.g. something like a big control-flow overhaul to introduce constructs beyond the single while {} loop form perhaps
<simon-perretta-img>
But I suspect that's not in the cards anytime soon :P
<alyssa>
I don't want to think about the implications of that
<simon-perretta-img>
Understandable haha
epoch101 has joined #dri-devel
<glehmann>
loops in NIR have to be simple, and most backends can optimize them well enough at the moment to whatever the hw provides
RSpliet has quit [Quit: Bye bye man, bye bye]
RSpliet has joined #dri-devel
<simon-perretta-img>
Oh no for sure; I was deliberately choosing a relatively unlikely scenario by using that as the example
epoch101 has quit []
epoch101 has joined #dri-devel
<simon-perretta-img>
I have seen a paper before that modelled predicated execution using a psi instruction, but don't think it really covered ops like stores where no defs are being produced
epoch101 has quit []
guludo has quit [Quit: WeeChat 4.7.0]
epoch101 has joined #dri-devel
guludo has joined #dri-devel
kts has quit [Quit: Konversation terminated!]
epoch101 has quit []
epoch101 has joined #dri-devel
<alyssa>
glehmann: switch statements are where NIR is being held back
<glehmann>
sure, but I don't think they would gain amd much
<glehmann>
unless the condition is uniform, then we can use a jump table
<glehmann>
I guess NAK could always use a jump table...
<glehmann>
but I think there's also low hanging fruit for switches in vtn