ChanServ changed the topic of #dri-devel to: <ajax> nothing involved with X should ever be unable to find a bar
<karolherbst>
ehhh... I might have to drop down to rust 2018 :(
<karolherbst>
or I simply use rustup... mhh
bluetail276 has joined #dri-devel
columbarius has joined #dri-devel
co1umbarius has quit [Ping timeout: 480 seconds]
Daanct12 has joined #dri-devel
flto_ has quit []
flto has joined #dri-devel
nielsdg has joined #dri-devel
<daniels>
use rustup
nchery is now known as Guest30
nchery has joined #dri-devel
Guest30 has quit [Ping timeout: 480 seconds]
kusma has joined #dri-devel
nchery has quit [Quit: Leaving]
nchery has joined #dri-devel
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #dri-devel
nchery is now known as Guest36
nchery has joined #dri-devel
Guest36 has quit [Ping timeout: 480 seconds]
camus has joined #dri-devel
camus has quit []
camus has joined #dri-devel
camus has quit [Remote host closed the connection]
nchery has quit [Ping timeout: 480 seconds]
camus has joined #dri-devel
nchery has joined #dri-devel
Mis012[m] has joined #dri-devel
Peuc_ has joined #dri-devel
Peuc has quit [Ping timeout: 480 seconds]
ella-0_ has joined #dri-devel
ella-0 has quit [Read error: Connection reset by peer]
nchery has quit [Ping timeout: 480 seconds]
vsyrjala has quit [Ping timeout: 480 seconds]
vsyrjala has joined #dri-devel
kts has joined #dri-devel
sdutt has quit [Ping timeout: 480 seconds]
alanc has quit [Remote host closed the connection]
alanc has joined #dri-devel
kts has quit [Ping timeout: 480 seconds]
kunal_1072002[m] has joined #dri-devel
chema has joined #dri-devel
KunalAgarwal[m]1 has joined #dri-devel
zzoon2627thholiday[m] has joined #dri-devel
Strit[m] has joined #dri-devel
chivay has joined #dri-devel
egalli has joined #dri-devel
srslypascal has quit [Ping timeout: 480 seconds]
Daaanct12 has joined #dri-devel
Daanct12 has quit [Read error: Connection reset by peer]
<mareko>
It's kinda sad that calling a GL function takes 3 jumps to get into Mesa. The call enters a function that has just a jump due to PLT in the dynamic linker. After the jump, we end up in glvnd, which reads the dispatch table and jumps to the target function. Then we end up in Mesa, which reads its dispatch table and jumps to the function the app called.
<mareko>
so: callq + jmpq + jmpq + jmpq
<airlied>
mareko: i think ajax has some ideas about collapsing those
<airlied>
but yeah mesa when glvnd is around could lose a layer
Daanct12 has joined #dri-devel
nchery has joined #dri-devel
Daaanct12 has quit [Read error: Connection reset by peer]
YaLTeR[m] has joined #dri-devel
jenatali has joined #dri-devel
cheako has joined #dri-devel
Company has quit [Quit: Leaving]
Daanct12 has quit [Read error: Connection reset by peer]
Daanct12 has joined #dri-devel
sdutt has joined #dri-devel
kts has joined #dri-devel
kts has quit [Quit: Konversation terminated!]
Sumera[m] has joined #dri-devel
halfline[m] has joined #dri-devel
kts has joined #dri-devel
mairacanal[m] has joined #dri-devel
zamundaaa[m] has joined #dri-devel
MajorBiscuit has joined #dri-devel
Daanct12 has quit [Remote host closed the connection]
frieder has quit [Remote host closed the connection]
frieder has joined #dri-devel
frieder_ has joined #dri-devel
frieder has quit [Read error: Connection reset by peer]
YuGiOhJCJ has quit [Quit: YuGiOhJCJ]
bluetail276 has quit []
bluetail has joined #dri-devel
<bluetail>
mareko that is intriguing. Thanks for sharing
<HdkR>
I was also unaware of such a large perf hit due to glvnd, good to know.
rasterman has joined #dri-devel
pjakobsson has joined #dri-devel
mvlad has joined #dri-devel
<mareko>
the first jump is due to the dynamic linker, it shouldn't be there for GetProcAddress functions, the second jump is due to glvnd, the third jump is due to glapi in Mesa
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #dri-devel
lynxeye has joined #dri-devel
cengiz_io has joined #dri-devel
Daanct12 has quit [Remote host closed the connection]
Daanct12 has joined #dri-devel
Jeremy_Rand_Talos has quit [Remote host closed the connection]
JohnnyonFlame has quit [Ping timeout: 480 seconds]
Jeremy_Rand_Talos has quit [Remote host closed the connection]
kts has quit [Quit: Konversation terminated!]
kts has joined #dri-devel
Guest665 is now known as shadeslayer
thellstrom has quit [Ping timeout: 480 seconds]
Guest66 is now known as go4godvin
morphis has quit []
Haaninjo has joined #dri-devel
JohnnyonFlame has joined #dri-devel
epoll has quit [Ping timeout: 480 seconds]
epoll has joined #dri-devel
halves has joined #dri-devel
<Ristovski>
mareko: Are the actual perf counters not hooked up to GL_AMD_performance_monitor for SI on amdgpu? None of them show up (apart from the GPIN_000-0004 ones)
thellstrom has joined #dri-devel
thellstrom1 has joined #dri-devel
thellstrom has quit [Ping timeout: 480 seconds]
kts has quit [Ping timeout: 480 seconds]
kts has joined #dri-devel
thellstrom1 has quit [Ping timeout: 480 seconds]
shadeslayer has quit [Quit: ZNC 1.8.2+deb1+bionic2 - https://znc.in]
alatiera5 has joined #dri-devel
alatiera is now known as Guest89
alatiera5 is now known as alatiera
tomba has joined #dri-devel
Guest89 has quit [Ping timeout: 480 seconds]
shadeslayer has joined #dri-devel
shadeslayer is now known as Guest90
Guest90 is now known as shadeslayer
khfeng has joined #dri-devel
bylaws has joined #dri-devel
dafna33[m] has joined #dri-devel
yshui` has joined #dri-devel
kts has quit [Ping timeout: 480 seconds]
Mershl[m] has joined #dri-devel
pendingchaos_ is now known as pendingchaos
fxkamd has joined #dri-devel
JohnnyonFlame has quit [Ping timeout: 480 seconds]
bmodem has joined #dri-devel
frieder_ has quit [Ping timeout: 480 seconds]
kts has joined #dri-devel
frieder_ has joined #dri-devel
sdutt has joined #dri-devel
kts has quit [Ping timeout: 480 seconds]
heftig has joined #dri-devel
lemonzest has joined #dri-devel
mclasen has joined #dri-devel
kallisti5[m] has joined #dri-devel
sdutt has quit [Read error: Connection reset by peer]
<kisak>
I wasn't able to focus on merge requests landing in mesa over the last week or so. Are there any game-fixing commits that folks would like to see fast tracked into my 22.1.6 PPA build? My list to evaluate currently has !14208 (splitgate) !18037 (general/device select) !17556 (multiple games, radeonsi) and the functional bits of !17859 (Intel D16 follow up)
bmodem has quit []
cleverca22[m] has joined #dri-devel
<mareko>
Ristovski: not for gfx6
<karolherbst>
dcbaker: meson.get_compiler('rust').version() prints 0.1.62. Is this intentional or?
FireBurn has quit [Read error: Connection reset by peer]
frankbinns has joined #dri-devel
FireBurn has joined #dri-devel
abhinav__2 is now known as abhinav__
frankbinns1 has joined #dri-devel
frankbinns has quit [Ping timeout: 480 seconds]
daniel has joined #dri-devel
idr has joined #dri-devel
daniel is now known as Guest110
kts has quit [Ping timeout: 480 seconds]
Guest110 has left #dri-devel [#dri-devel]
<danvet>
robclark, I think we should at least have a wider discussion and maybe improve the docs a bit around caching and cache coherency
<danvet>
there was a parallel thread with könig on this topic, and he promised to create some doc patches summarizing what we discussed
<danvet>
also I guess some more implementations might be good, or are we ok with fragmentation here?
<danvet>
robclark, on the implementation I wonder whether we should lift this up into an exporter flag or something
<danvet>
and make it invariant
<danvet>
maybe even have a check that the vma is set up correctly after dma_buf_mmap finishes
<robclark>
so, I think there will always be some corner cases where mapping to vm guest won't work, in particular anything that needs the cpu begin/end ioctls.. I think we just have to live with that
FireBurn has quit [Remote host closed the connection]
<robclark>
but yeah, I could get behind moving it to exporter flag
<danvet>
robclark, well you could fwd the begin/end ioctls
<robclark>
perf won't be so hot
<danvet>
also, if your userspace doesn't, then I think "figuring out whether avoiding begin/end is ok" needs to be part of your uapi contract
<robclark>
I think you'd be better off not direct mapping in those corner cases
<danvet>
yeah :-)
<danvet>
robclark, also, what should we do with ttm backed drivers where the pgprot can change at runtime?
MajorBiscuit has joined #dri-devel
<danvet>
or do we require that dma_buf_mmap must have the same pgprot everywhere?
<danvet>
which I don't think works, because device wc is different from system memory wc on armv8 afaik
FireBurn has joined #dri-devel
<robclark>
re: things where it is not ok to skip beg/end ioctls, I think map-info ioctl could just fail in that case? The ioctl is basically returning "what sort of coherent access is possible"
unrelentingtech has joined #dri-devel
<danvet>
robclark, also, can't you fish this out through kvm with an mmu notifier and follow_pte()?
<danvet>
that would solve all the above problems except for the "do I need begin/end?"
<robclark>
re: ttm, maybe ttm drivers need some flag that exported dmabufs need to support consistent pgprot
<robclark>
ie. not all buffers need to support this, and maybe it is ok for something to be suboptimal
<danvet>
robclark, yeah but if you want it to fail when begin/end is needed, that needs to be documented
<robclark>
fair
<danvet>
and probably enforced, i.e. if begin/end is filled out maybe just fail this in the dma-buf.c code?
<robclark>
hmm, doing it based on ops maybe means having multiple sets of ops unnecessarily
<danvet>
robclark, on follow_pte(): I'm bringing this up because I screamed at that code in kvm a year ago
<danvet>
so there's already something there
<danvet>
robclark, yeah but less room for mistakes
<robclark>
I don't think mmu notif is enough
<robclark>
since it would have to propagate up to the guest kernel
<robclark>
to change the S2 tables
feaneron has joined #dri-devel
Major_Biscuit has quit [Ping timeout: 480 seconds]
Haaninjo has joined #dri-devel
x512[m] has joined #dri-devel
<karolherbst>
.... I need a newer rustc than what debian provides... seems like I really have to figure out how to use rustup then
<karolherbst>
which I suspect is something we want to do anyway
<karolherbst>
daniels: soo.. if I have to modify PATH, how'd you do that for CI?
<karolherbst>
I install rustup and the rust toolchain I want to use when building the debian container, but apparently PATH isn't updated although the rustup installer is doing that
<karolherbst>
I suspect the shell doens't pick it up
MattGPU_ has joined #dri-devel
MattGPU has quit [Read error: Connection reset by peer]
MattGPU_ has left #dri-devel [#dri-devel]
stuart has joined #dri-devel
enick_403 is now known as pmoreau
italove has joined #dri-devel
<Ristovski>
mareko: Is there any work being done for GFX6 perf counters? does radeonsi not have them as well?
rasterman has quit [Quit: Gettin' stinky!]
colemickens has joined #dri-devel
<daniels>
karolherbst: export $PATH
<karolherbst>
simply in the build script?
<daniels>
if rustup just appends to ~/.profile or something, that doesn’t get sourced
<daniels>
yeah
<karolherbst>
mhh.. I've added it like that ccache stuff atm
<karolherbst>
ohh wait, no. I added it to meson/build.sh actually
<Frogging101>
That assertion failure with DMC4 just magically went away and I have no idea why. It was happening consistently and now it's gone
<karolherbst>
okay.... I hope this time it compiles.. :D
<karolherbst>
"Pipelines 177" uhhh
danvet has quit [Ping timeout: 480 seconds]
<Frogging101>
It's the assertion that ensures that an image view on an image without VK_IMAGE_CREATE_MUTABLE_FORMAT_BIT has the same format.
JohnnyonFlame has quit [Ping timeout: 480 seconds]
paulk has quit [Quit: WeeChat 3.0]
paulk has joined #dri-devel
tleydxdy has joined #dri-devel
DavidHeidelberg[m] has joined #dri-devel
arisu has joined #dri-devel
<karolherbst>
dcbaker: huh.. something is still bonkers with bindgen and dependencies...
viciouss[m] has joined #dri-devel
<karolherbst>
ohh wait.. we can't declare dependencies mhhhh
<dcbaker>
karolherbst: yeah, I have patches for that, but they need this "I rewrote half of our internal IR because it was bonkers" series to land first...
<karolherbst>
:(
<karolherbst>
is there any other way to add dependencies to an existing target in meson?
<karolherbst>
ehh wait...
<dcbaker>
no, by design you can't mutate a target once it exists
<karolherbst>
ehhh... I was wrong, it's not a bindgen problem...
<karolherbst>
my wrapper doesn't depend on nir...
<dcbaker>
oh good, my monday has already been full of "you changed something and exposed a big, annoying bug that's existed for a long time, but is probably also your fault because you wrote the original code that's broken too" :/
<karolherbst>
oops
ralf1307[theythem][m] has joined #dri-devel
ybogdano has quit [Read error: Connection reset by peer]
<karolherbst>
heh... I also want to depend on idep_nir I think :D
kts has quit [Ping timeout: 480 seconds]
ybogdano has joined #dri-devel
hasebastian[m] has joined #dri-devel
JohnnyonFlame has joined #dri-devel
Andy[m] has joined #dri-devel
mclasen has quit [Ping timeout: 480 seconds]
rasterman has joined #dri-devel
YuGiOhJCJ has quit [Remote host closed the connection]
rsripada has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
MajorBiscuit has quit [Ping timeout: 480 seconds]
mclasen has joined #dri-devel
pcercuei has quit [Quit: dodo]
rasterman has quit [Quit: Gettin' stinky!]
heat_ has quit [Read error: No route to host]
heat has joined #dri-devel
mclasen has quit []
mclasen has joined #dri-devel
<karolherbst>
... now the llvmpipe-piglit-cl test picks up rusticl instead of clover :D
dliviu has quit [Ping timeout: 480 seconds]
YuGiOhJCJ has quit [Ping timeout: 480 seconds]
pi2 has joined #dri-devel
YuGiOhJCJ has joined #dri-devel
dliviu has joined #dri-devel
mclasen has quit [Ping timeout: 480 seconds]
khfeng has joined #dri-devel
JohnnyonFlame has quit [Ping timeout: 480 seconds]
JohnnyonFlame has joined #dri-devel
heat has quit [Read error: Connection reset by peer]
heat has joined #dri-devel
Haaninjo has quit [Quit: Ex-Chat]
nchery is now known as Guest145
nchery has joined #dri-devel
nchery has quit [Remote host closed the connection]