07:08 hungerinworld: it's that wrap around of 64bit value would either remove or add highest bit if you try to subtract the msb, one can controll this through the compiler flags, it depends on whether one starts by smaller or higher value or inverse. dmabuf is fifo mode head and tail CPU validated delegation also. but it's also possible to remove all but highest power on wrap around in this case highest
07:08 hungerinworld: power is removed from both the value and inverse. And it has to be added back to the bigger outcome , hence it's maxvalue - both outcomes added together. to access the bigger value or bigger inverse later. and if subtracting initial from that and accessing two inverse register files the result can be offered without the offered skeleton on all hardware accessible in the world. so 1
07:08 hungerinworld: provides zero output and above provides the inverse output register file is authored or pinned so. without wraparound it's register file has to have different polarity.
07:31 mripard: lumag: thanks for reviewing the vs drm driver
09:33 lumag: mripard, :-)
09:35 mripard: lumag: it's been stuck without much review since forever so it's greatly appreciated :)
11:11 generalrob: I already looked at the p2p mode engine that dickhead offered on #nouveau, he appears to be correct that Nvidia employee patch is better faster for such purpose as well as securer, so I agree with airlied there, dmabuf should never been landed in such form for safety, the other engine is securer and more performant.
11:17 mripard: lumag: if you have the chance (and time), it would be great if you could review https://lore.kernel.org/r/20240521-kms-hdmi-connector-state-v14-0-51950db4fedb@kernel.org too :)
11:47 generalrob: I have such terminology on conversion I use pseudo register file for both directions except the low bits upto 7th power in bigger to smaller transition like encoding so the true register file access has same polarity however the pseudo register file has different polarity in the procedures it depends whether you operate on wraparound or not, but overall we can congratulate you and me , we
11:47 generalrob: have everything now. I am doing super fine myself too.
11:50 generalrob: it's because 513-513 and 511-513 would both trigger 0 as result without wraparound so it's clear that upon pseudo register file access described 1 needs to present 513 on no wraparound and zero on wraparound.
11:51 generalrob: the actual format is different but logics behind exactly the same just I did not fiddle with higher powers for simplicity here.
11:59 zmike: dwfreed:
12:00 generalrob: so big int to small int has three pseudo register file routines whose results get added together on wraparound, but only two on no wraparound. illustrated as said 1023-513 is 510 511-513 is 2 510 + 2 +513 is 1023 so three values appended.
12:00 generalrob: zmike: you go suck your cock in your own corner filthy trash
12:03 generalrob: on no wraparound it's 1023-513 is 510 +0 is still 510 so it does 510+513 is 1023
12:04 lumag: mripard, the problem with that series is that generic bits look fine to me.
12:05 lumag: I'll give it a thorough look
12:05 mripard: lumag: "looks fine to me" is kind of the same thing than Reviewed-by though :)
12:07 generalrob: as 513-513 is 0 and 510-513 is zero the value 0 becomes 513 after adding back, so they are at different polarity in fact.
12:08 generalrob: although there's no true sw register file wraparound and not one have slightly different bottom encoder routines.
13:17 stsquad: does the Mesa for the backend of VirtIO need the virtio drivers enabled for gallium and vulkan or does it just need the host gpu drivers?
13:20 stsquad: also does the graphics loader honour LD_LIBRARY_PATH when loading the GPU userspace (as its not linked with the main binary)
13:21 alice: LIBGL_DRIVERS_PATH exists for drivers
13:27 unclegori: the routine is internally different only, it's functional public handler is the same, nothing needs to change not even initializations differences apply until the third triplet works on zero it just adds zero, what I am saying my code does not care if it's wraparound or not at all if zmike flies so high I doubt but that's the way it is, just that the dma API is more secure for exploits and
13:27 unclegori: more performant the one that airlied showed there is a possibility to improve it to configure it to compute mode however and tbh. I am by then done with my compiler too while you argue how to make dma compute engine functional through engines global configurator, until then the encoder decoder needs to be done on some other device, but dma is most preferred, cause these routines always run
13:27 unclegori: the most often.
13:28 stsquad: ahh my LD_LIBRARY_PATH of /home/alex/lsrc/qemu.git/builds/extra.libs/install/lib also needs /home/alex/lsrc/qemu.git/builds/extra.libs/install/lib/x86_64-linux-gnu/ I think
13:28 stsquad: alice and LIBGL_DRIVERS_PATH should point to /home/alex/lsrc/qemu.git/builds/extra.libs/install/lib/x86_64-linux-gnu/dri/
13:29 alice: think so
13:29 stsquad:wonders if he can get the library triplet easily for his shell env setup
14:32 zmike: kusma: you are first on the agenda today if you want to call in
14:33 kusma: When is the meeting? I'm busy the next hour ish, but can manage later
14:36 zmike: it's in 25 minutes
14:40 kusma: Ok, won't make it
16:07 zmike: kusma: we approved anyway
16:07 zmike: it was quick.
16:09 kusma: Ait, thanks
16:11 zmike: (now you're back to cts waiting)
16:11 zmike: but also I think you probably need to rebase your cts cl because a huge cleanup thing landed last week
17:38 todotomorrow: you are the foolest fecalists in computing history, it's fucking disgusting fuckfarter shit hoses, how is that zmike trash qualified to abuse me ooh you end up dead garbages and very soon, I ain't gonna pay for your farts and sausage show, as dead as your other allies. Ridiculous people I say fecalists as much as those Estonian abortion leftovers. total tyranny you do both. "nice"
17:38 todotomorrow: experience was this sexual abuse humiliating comments in my vacation, you gonna get treated like Mussolini, where just all my allies come and smash your leftover crew.
17:57 DavidHeidelberg: btw. are we ready to switch from -O0 to -O2 in the CI? https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/29246
18:13 iive: DavidHeidelberg, iirc gcc has -Og for optimizations that doesn't hinder debugging.
18:15 DavidHeidelberg: iive: I know, but meson.. said no https://github.com/mesonbuild/meson/issues/2940 (at least for now)
18:15 iive: heh
18:16 DavidHeidelberg: anyway, currently it's the -O0 option which isn't used much as sometimes (read with LTO/IPO) leads to compiler bugs
18:34 DemiMarie: Is it reasonable for a driver to take arbitrary pages (provided by userspace) and do a FOLL_LONGTERM pin on them?
18:41 stsquad: hmm still can't get virglrenderer to load the locally built libs - does it rely on mesa to do the loading of drivers for it?
19:05 capuchinoboy: in some way dma buf is fun, and it was my original idea, since there is a path of clock domain cross checking for self-timed clocks, but after they aimed my heart with mobile phones used as frequency generator and grew bacteria after poisoning me with staphylococcus I realized such code is wrong to be supported, so simply said I was wrong with my first idea, I later shifted towards
19:05 capuchinoboy: safety and security, cause the code that I do is much harder than clockless mode, but to me still simple, In other words I encouraged you to do this code but I was wrong.. I wish I never said the things about earlier prototype of my research, and that's what airlied says now or maybe said all the time I dunno, but I was clearly shortsighted, people do so bad things in the world where I
19:05 capuchinoboy: did not account with that.
19:27 marex: hey, am I right in assuming https://gitlab.freedesktop.org/mesa/mesa/-/tree/24.0?ref_type=heads (mesa.git, branch 24.0) should contain mesa-24.0.8 tag ?
19:27 marex: currently the tag is only in https://gitlab.freedesktop.org/mesa/mesa/-/tree/staging/24.0?ref_type=heads (mesa.git, branch staging/24.0)
19:27 airlied: eric_engestrom: ^
19:28 marex: airlied: thanks
19:28 marex: might just be a missed push
20:03 gigglestoday: so reddit has today's what's your random genetic win? feed question. For those curious enough in my case, I am a born nuclear catastrophe worker, I have only wins such are all marked by science and the system is Keller's antigen antibody system they lean on. so that's why docs who gave birth to retards and similar others violate me, crack up my joints and do not get ever responsible, I
20:03 gigglestoday: am a superman still at my organs, and that's what you whitenessed by attacking them with waves and poisoning, my organs did not give up to lasers, bacteria, waves and I have perfect pressure indicators in ideal range, they flux hard, but if 20 digits is off, I know I am being attacked again, hence I knew about this method, they normally get away with such killoffs cause no traces are
20:03 gigglestoday: left when reverse frequency is delivered, but my organs were way too strong, and I knew all about that and the method too that I will survive.
20:43 kisak: mattst88: hey, you wouldn't happen to know why my mesa-9999.ebuild is fighting me so hard with `Run-time dependency llvmspirvlib found: NO (tried pkgconfig and cmake)` and llvm 15 / dev-util/spirv-llvm-translator-15.0.0-r1? How is it supposed to pick up /usr/lib/llvm/15/lib64/pkgconfig/LLVMSPIRVLib.pc under normal conditions?
20:45 stsquad: kisak does pkg-config --cflags llvmspirvlib work on your command line?
20:46 kisak: nope >_<
20:47 alice: do you have PKG_CONFIG_PATH set to /usr/lib/llvm/15/lib64/pkgconfig/
20:49 mattst88: the .pc file is named LLVMSPIRVLib.c -- all caps
20:49 mattst88: .pc*
20:49 mattst88: this work for me: PKG_CONFIG_PATH=/usr/lib/llvm/17/lib64/pkgconfig pkg-config --cflags LLVMSPIRVLib
20:50 mattst88: the ebuild sets PKG_CONFIG_PATH according to the selected llvm/spirv-llvm-translator version
20:50 kisak: PKG_CONFIG_PATH="/usr/lib/llvm/15/lib64/pkgconfig/" pkg-config --cflags LLVMSPIRVLib works
20:51 kisak: okay, so that means I have historical relics fighting me, thanks. I can look that direction.
20:51 mattst88: the ebuild sets PKG_CONFIG_PATH iff USE="opencl llvm" -- are those set for you?
20:51 kisak: USE=opencl was off. Not needed before.
20:54 mattst88: what was causing spirv-llvm-translator to be needed then?
20:55 kisak: I expect intel-clc
20:55 mattst88: wondering if I've missed a configuration
20:56 kisak: since I'm older than the norm, I'm still forced to try funky stuff.
20:56 mattst88: damn, that looks like a missing dependency in the ebuild
20:57 kisak: I wouldn't immediately jump to a bad mainline config. You're forcing intel-clc as a separate package, and that separate package doesn't support llvm slot 15.
21:01 mattst88: ah yeah, I don't see llvmspirvlib requested by meson when I build with USE=-opencl
21:03 DavidHeidelberg:wonders why would anyone build mesa with OpenCL :P
21:04 kisak: This part isn't optional anymore
21:08 kisak: uh ... I think I out of luck here. It's looking for 32 bit libLLVMSPIRVLib and gentoo's libLLVMSPIRVLib packaging isn't multiarch. This is going to be an order of magnitude more pain on the Ubuntu side.
21:09 kisak: This shouldn't matter for normal Gentoo users, since the single arch intel-clc standalone binary takes care of it?
21:10 karolherbst: might be finally time to clean up this llvm packaging mess :P
21:13 mattst88: kisak: right, for intel_clc a single binary works for 64-bit and 32-bit multilib
21:13 mattst88: for opencl support, I've chosen to just not offer it on 32-bit x86 anymore
21:14 mattst88: so in both cases, only a native ABI spirv-llvm-translator is needed
21:14 karolherbst: what's the problem with providing 32 bit builds anyway?
21:16 mattst88: IIRC it meant I would need to make the spirv-llvm-translator package multilib in Gentoo, and I thought that was just a waste of effort
21:16 karolherbst: but why would that be in any way difficult?
21:16 mattst88: can't remember if there was more to it
21:16 karolherbst: there is a tool thing it installs, but you could disable it for 32 bit
21:16 karolherbst: or non native rather
21:19 mattst88: some discussion of why we just disabled 32-bit support here: https://github.com/gentoo/gentoo/pull/28209
21:19 mattst88: looks like https://github.com/gentoo/gentoo/pull/28209#issuecomment-1316946367
21:19 mattst88: which I'd forgotten about
21:19 mattst88: and then tried to fix in https://github.com/KhronosGroup/SPIRV-LLVM-Translator/pull/2555
21:22 yesisleepwell: gigglestoday> and by the way why you spam my phone with the whore terrorists prays, I do not care what she prays at, she was humiliating me for 2.5 years and cheated with around ten people at my sight tongue kissing and one even fucked her pregnant and those were as big of retards as she is. and btw I of course use different technique for super execution engine than what I described to
21:22 yesisleepwell: airlied, but
21:22 yesisleepwell: <gigglestoday> you are so numb people you do not even understand anything when delivered on silver plate, the distance can be arbitary per PC for the compiler as long as you eliminate it through eliminating the constants involved, so it's very simple to code it, instead of distance in the data access case you mark operands to use it for executions , yes those can be assymetric to
21:22 yesisleepwell: account with
21:22 yesisleepwell: <gigglestoday> dependencies.
21:22 yesisleepwell: <gigglestoday> I do no longer care what she does, she extorted money from me by telling others how I raped her, which was not the case either i.e it wasn't a rape, she is a whore.
21:23 yesisleepwell: <gigglestoday> why she got what she deserved was that all the relations she invited herself started them to get fucked and complained to so called raper, that wormbitch is a monster, after some British trash allegedly was ordered to assault me I heard those monsters on gulf celebrating and bullying, they got finally treated to hospital .
21:23 yesisleepwell: <gigglestoday> handled by force.
21:23 yesisleepwell: And also I check their accounts cause if they are again near my father's hotel with this Gloria they will be put to wheelchair, Glorias so called brothers too. I do not care what they plan and threat I sleep well I know they are done as soon as they approach my territory. I do not care of your or theirs gangsters nor gods you are shit to me. Don't bother with your clickbate whore
21:23 yesisleepwell: threats reading universe godly messages or whatever AI it is.
21:23 DemiMarie: dwfreed
21:23 yesisleepwell: gigglestoday> and by the way why you spam my phone with the whore terrorists prays, I do not care what she prays at, she was humiliating me for 2.5 years and cheated with around ten people at my sight tongue kissing and one even fucked her pregnant and those were as big of retards as she is. and btw I of course use different technique for super execution engine than what I described to
21:23 yesisleepwell: airlied, but
21:23 yesisleepwell: <gigglestoday> you are so numb people you do not even understand anything when delivered on silver plate, the distance can be arbitary per PC for the compiler as long as you eliminate it through eliminating the constants involved, so it's very simple to code it, instead of distance in the data access case you mark operands to use it for executions , yes those can be assymetric to
21:24 yesisleepwell: account with
21:24 yesisleepwell: <gigglestoday> dependencies.
21:24 yesisleepwell: <gigglestoday> I do no longer care what she does, she extorted money from me by telling others how I raped her, which was not the case either i.e it wasn't a rape, she is a whore.
21:24 yesisleepwell: <gigglestoday> why she got what she deserved was that all the relations she invited herself started them to get fucked and complained to so called raper, that wormbitch is a monster, after some British trash allegedly was ordered to assault me I heard those monsters on gulf celebrating and bullying, they got finally treated to hospital .
21:24 yesisleepwell: <gigglestoday> handled by force.
21:24 yesisleepwell: And also I check their accounts cause if they are again near my father's hotel with this Gloria they will be put to wheelchair, Glorias so called brothers too. I do not care what they plan and threat I sleep well I know they are done as soon as they approach my territory. I do not care of your or theirs gangsters nor gods you are shit to me. Don't bother with your clickbate whore
21:24 yesisleepwell: threats reading universe godly messages or whatever AI it is.
21:27 mattst88: ffs. he's really active today
21:30 Sachiel: you should have kept your promise not to be a suicidal terrorist
21:31 dwfreed: O.o
21:31 DemiMarie: I think some gaming platforms have resorted to banning computers, rather than humans.
21:32 DemiMarie: The reason being that a hardware ban is much easier to enforce, at least if one has a fat client.
21:32 Sachiel: irc doesn't give you that information
21:32 dwfreed: ^
21:32 zmike: what if
21:32 zmike: and hear me out
21:32 zmike: what if next XDC is in estonia
21:33 alice: 💪🇪🇪
21:36 kisak: mattst88: this would be a little less evil if the mesa build actually listened to LLVM_MAX_SLOT, then I could have intel_clc built against llvm 17 and the rest of the test build picking up llvm 15.
21:37 DemiMarie: zmike: why estonia?
21:37 kisak: (use standalone intel_clc and tell gentoo to build mesa with llvm 15)
21:37 DemiMarie: dwfreed: this is *especially* true if there is a TPM (or other secure hardware) involved
21:37 dwfreed: yeah, there's no way we can do that
21:37 DemiMarie: indeed
21:38 dwfreed: nor would I really want access to that information
21:38 dwfreed: (GDPR would be a huge problem)
21:38 DemiMarie: I think there are fancy cryptographic solutions to that problem, such as what Privacy Pass uses.
21:39 mattst88: kisak: why do you think the mesa ebuild doesn't listen to LLVM_MAX_SLOT?
21:39 zmike: DemiMarie: https://www.youtube.com/watch?v=vuBWbpTJRqk
21:39 DemiMarie: I agree that it is impractical, though.
21:40 kisak: mattst88: it historically hasn't meson does whatever it wants. set older llvm as max and have llvm 17, 17 is used. Maybe that's changed.
21:41 kisak: my test ebuild has some ancient bits, so this could have been solved a long time ago.
21:41 karolherbst: you probably have to tell meson what llvm-config file to use
21:42 mattst88: I think that's how the llvm.eclass is supposed to function, but I'm not an expert there
21:46 DemiMarie: Could I get a fact check on some information about how GPU memory mappings have special requirements? I'm giving a talk to people who are not GPU experts and I want to make sure that my information is accurate. Thanks in advance for the feedback.
21:53 iive: I thought that you have to be registered, so you can talk in the channel.
21:53 dwfreed: iive: joss registers
21:54 iive: maybe network admins could do something about him?
21:54 dwfreed: we are
21:54 dwfreed: (I'm a network admin)
21:57 iive: dwfreed, how is he evading detection? I assume he is using vpn, public proxies, new email every time?
21:57 dwfreed: there are a number of things he's doing, yes
21:58 airlied: he should just go sell his services to spammer companies :-P
22:00 mattst88: dwfreed: would it help if I collected a list of his connections?
22:00 dwfreed: mattst88: nah, I've got a list already, doesn't help much because they generally aren't used more than once
22:01 mattst88: yeah, looking for "ChanServ has kicked ..." is 99% him
22:02 mattst88: he's been doing this stuff for more than 10 years across two IRC networks.
22:02 mattst88: it's pretty incredible.
22:18 Calandracas: fwif void has rusticl and llvm-spirv packages for 32 bit archs (armv{6,7}l, i686)