ChanServ changed the topic of #lima to: Development channel for open source lima driver for ARM Mali4** GPUs - Kernel has landed in mainline, userspace driver is part of mesa - Logs at https://people.freedesktop.org/~cbrill/dri-log/index.php?channel=lima and https://freenode.irclog.whitequark.org/lima - Contact ARM for binary driver support!
Da_Coynul has quit [Ping timeout: 245 seconds]
<bshah|matrix> At one point I had done a lima dump as well
<bshah|matrix> Although best do it latest now
megi has quit [Ping timeout: 245 seconds]
kaspter has joined #lima
camus has joined #lima
kaspter has quit [Ping timeout: 258 seconds]
camus is now known as kaspter
jrmuizel has quit [Remote host closed the connection]
kaspter has quit [Ping timeout: 245 seconds]
kaspter has joined #lima
Elpaulo has joined #lima
jrmuizel has joined #lima
minecrell has quit [Quit: Ping timeout (120 seconds)]
minecrell has joined #lima
jrmuizel has quit [Remote host closed the connection]
<bshah|matrix> Btw I'll provide you with logs and testing if blending function fails this weekend, today I'm traveling out for KDE Akademy
<anarsoul> OK
<anarsoul> have a safe trip :)
<MoeIcenowy> retested on Lima on Android
<MoeIcenowy> seems that still largely broken
dddddd has quit [Remote host closed the connection]
guillaume_g has joined #lima
bhoj has quit [Write error: Connection reset by peer]
griffinp has quit [Read error: Connection reset by peer]
griffinp has joined #lima
bhoj has joined #lima
guillaume_g has quit [Quit: Konversation terminated!]
guillaume_g has joined #lima
guillaume_g has left #lima [#lima]
Elpaulo has quit [Read error: Connection reset by peer]
Elpaulo has joined #lima
mripard has quit [Ping timeout: 276 seconds]
mripard_ has joined #lima
mripard_ is now known as mripard
mardestan has joined #lima
nerdboy has quit [Ping timeout: 246 seconds]
nerdboy has joined #lima
nerdboy has quit [Ping timeout: 244 seconds]
Barada has joined #lima
nerdboy has joined #lima
Barada has quit [Client Quit]
nerdboy has quit [Ping timeout: 268 seconds]
mardestan has quit [Ping timeout: 268 seconds]
megi has joined #lima
Wizzup has quit [Ping timeout: 245 seconds]
Wizzup has joined #lima
camus has joined #lima
camus has quit [Client Quit]
kaspter has quit [Ping timeout: 246 seconds]
mardestan has joined #lima
<mardestan> My memory is vague how to map the shader object into some buffer object in gl2.1 or es2, the intention is:
<mardestan> access the machine instructions from vram, write a shader with mali offline compiler to checksum the stuff, integrate the pass to the compiler (done on gpu this time to chekcsum the instructions)
<mardestan> later write the shader in mali offline assembler to run the checksums. integrate that one to the toolkit as well.
<mardestan> anarsoul: branches and flow control is not needed with this method though, overall gpu branching is pretty numb and pointless stuff.
Wizzup has quit [Ping timeout: 245 seconds]
<mardestan> such thing could be done on binary drivers too, the method, i try to hack this together for IOS, LINUX and other platforms.
Wizzup has joined #lima
<mardestan> i can not remember how to trigger in the binary blobs in es2 level, well the checksum shader should be fairly easy
mardestan has quit [Quit: Leaving]
megi has quit [Ping timeout: 258 seconds]
jrmuizel has joined #lima
jrmuizel has quit [Remote host closed the connection]
jrmuizel has joined #lima
jrmuizel has quit [Remote host closed the connection]
jrmuizel has joined #lima
dddddd has joined #lima
yuq825 has joined #lima
jrmuizel has quit [Remote host closed the connection]
afaerber has quit [Quit: Leaving]
jrmuizel has joined #lima
megi has joined #lima
afaerber has joined #lima
adjtm_ has quit [Ping timeout: 245 seconds]
adjtm_ has joined #lima
yuq825 has quit [Quit: Leaving.]
kaspter has joined #lima
minecrell has quit [Read error: Connection reset by peer]
minecrell has joined #lima
kaspter has quit [Remote host closed the connection]
drod has joined #lima
kaspter has joined #lima
nerdboy has joined #lima
nerdboy has quit [Changing host]
nerdboy has joined #lima
jernej has quit [Quit: Free ZNC ~ Powered by LunarBNC: https://LunarBNC.net]
jernej has joined #lima
drod has quit [Ping timeout: 245 seconds]
kaspter has quit [Ping timeout: 264 seconds]
drod has joined #lima
megi has quit [Quit: WeeChat 2.5]
adjtm_ has quit [Ping timeout: 268 seconds]
megi has joined #lima
afaerber has quit [Quit: Leaving]
adjtm_ has joined #lima
afaerber has joined #lima
jrmuizel has quit [Remote host closed the connection]
drod has quit [Remote host closed the connection]
Da_Coynul has joined #lima
<anarsoul> enunes: so we can spill only vec4 registers?
<enunes> on my experiments we can spill any size, as long as we don't have to swizzle
<enunes> so we can spill with vec1 or even vec2 alignment, but it must be .x or .xy
<anarsoul> that's crazy.
<anarsoul> I can't believe they added this limitation
<enunes> I can't convince the offline compiler to show me some other way
<enunes> loading is ok as we load to ^uniform and can swizzle then in the reading instruction
<enunes> for store, it seems to be the least bad way as any other way would require creating a separate instruction, while this way it can be easily merged
<anarsoul> can we just store whole reg as is and unswizzle it when loading?
<anarsoul> should work fine for SSAs
<anarsoul> but not for regs though
<enunes> I think it's pretty much what is happening now, there is no is_head limitation for the read
<anarsoul> it doesn't really matter if it's read or write, it's a single reg (or ssa) and you set is_head then it can only be at the beginning of register
<enunes> yes, it seems like the best solution so far
<anarsoul> are there any regressions in piglit?
<enunes> I have tested this with piglit, piglit forced spilling 1,2,3 times
<enunes> also with modified cost calculation to make it exercise vec1, vec2, vec3 with forced spilling
<enunes> no regressions so far
<anarsoul> ideas is still broken?
<enunes> yes, since I am unable to cause a failure by introducing spilling to any test with any vec size, I believe the problem with ideas is not spilling
<enunes> ideas resolves regalloc even with the is_head solution
<enunes> (and your other MRs applied)
<anarsoul> :(
<anarsoul> that's rather big shader and I hoped I wouldn't have to manually walk it through...
<enunes> well, we can go just trying to fix other issues in smaller piglit tests, one of them might be what is in ideas
<anarsoul> yeah, likely
<anarsoul> you can use my r-b and merge this MR
<enunes> thanks for the reviewing and feedback
<enunes> we should start merging your other ones too
<anarsoul> thanks for working on spilling
<anarsoul> I still need r-b for NIR part
<enunes> anholt reviewed the phis to scalar one but it's not clear to me if he also meant it for the scalar fcsel as the commits are in the same MR
<anarsoul> probably not, since I mentioned that it depends on another MR
<anarsoul> let me ask him