#
✕ nick:lina
#asahi-gpu · 2023-03-01 11:48
linaOther than that I haven't been able to repro any pink glitches, or any other weirdness, with everything I've thrown at it so far ^^
#asahi-gpu · 2023-03-01 11:34
linaShell Shockers has the same problem ^^
#asahi-gpu · 2023-03-01 11:01
linaThis is not a regression, it's always been like that apparently...
#asahi-gpu · 2023-03-01 10:57
linaalyssa: On that branch + latest firefox, quakejs renders without faults but all geometry is jittery, as if vertices are being computed in fp16.
#asahi-gpu · 2023-03-01 10:34
lina256FPS in terrain!
#asahi-gpu · 2023-03-01 10:31
linaThere was both some fence breakage in Mesa and a kernel bug (or rather a kernel I-didn't-know-about-a-firmware-corner-case). The latter was causing the hangs, so please uprev the kernel too ^^
#asahi-gpu · 2023-03-01 10:30
linaalyssa: MR with the stuff (I think) is already upstreamable is !21620, and asahilina:agx/next should be your rebased next ^^
#asahi-gpu · 2023-02-27 08:05
linaTry 256 for now until that all goes away please ^^
#asahi-gpu · 2023-02-27 08:05
linaIt's... 20? No wonder that doesn't work, we can queue up to 64 commands in parallel or something like that...
#asahi-gpu · 2023-02-27 08:01
linaThe mailbox thing... just increase that constant, I don't think there is anything I can do there until marcan gets rid of it entirely...
#asahi-gpu · 2023-02-27 08:00
linaI'm thinking it's WSI, like we used not to install fences in buffers at all and now we do and somehow one isn't getting signaled.
#asahi-gpu · 2023-02-27 08:00
linaalyssa: I got the glmark2 hang on KDE/Xwayland too, there's definitely something there ^^;;
#asahi-gpu · 2023-02-26 15:16
linaBut at least xonotic now regularly reaches the default 250fps cap!
#asahi-gpu · 2023-02-26 15:15
linaI still get some weird hangs though... I think something is still broken, probably in WSI ^^;;;
#asahi-gpu · 2023-02-26 07:29
linaRelease build GBM score: 7252 ^^
#asahi-gpu · 2023-02-26 07:21
lina^^
#asahi-gpu · 2023-02-26 07:21
linaThat's on GBM, a bit less on Wayland but the output looks legit... I don't think it's fast because it's broken...
#asahi-gpu · 2023-02-26 07:18
linaNow hitting over 9000 FPS on some tests...
#asahi-gpu · 2023-02-26 07:17
linaWait and that's a debug build
#asahi-gpu · 2023-02-26 07:16
linaWooooah I don't know if this is me removing the sync in flush_resource or something you did, but glmark2 just jumped up a ton. Like, from 4500 to 6600 (!!)
#asahi-gpu · 2023-02-26 07:10
linaI also added some nice debugging stuff along the way and more asserts ^^
#asahi-gpu · 2023-02-26 07:09
linaAlso the ail stuff for multisampling compression is wrong, I threw in a random guess but that needs tests. That fixed WebGL Aquarium.
#asahi-gpu · 2023-02-26 07:09
linaalyssa: Fixed the Firefox issue, it turns out we need to make resources shareable in flush_resource... so now it has to blit if not already shareable.
#asahi-gpu · 2023-02-26 02:28
linaYay! ^^
#asahi-gpu · 2023-02-26 02:26
linaRight...
#asahi-gpu · 2023-02-26 02:25
linaOof...
#asahi-gpu · 2023-02-26 02:25
linaSo do, like, other pure explicit sync drivers not exist? Because that is the only usage of this API anywhere in mesa...
#asahi-gpu · 2023-02-26 02:24
linaOkay, now where do I plug this in for us...
#asahi-gpu · 2023-02-26 02:24
linaalyssa: Good news: it's in common WSI code! Bad news: it's in common *Vulkan* WSI code. ^^;;
#asahi-gpu · 2023-02-26 02:22
linaI get the feeling what we're doing now is *sufficient* for correct WSI but may not be what is actually expected of us...
#asahi-gpu · 2023-02-26 02:21
linaFingers crossed it's a magic cap we set and mesa does it for us ^^
#asahi-gpu · 2023-02-26 02:21
linaLet me look for the DMA-BUF stuff...
#asahi-gpu · 2023-02-26 02:19
linaBut there is the whole attaching fences to dma-bufs story that I haven't looked at at all... I don't know how the mesa hooks for that work, if it isn't all taken care of by gallium already based on the existing fence support...
#asahi-gpu · 2023-02-26 02:19
linaAs for the WSI, my understanding is that that ctx->syncobj is supposed to just track the last submitted batch, so that when WSI wants a barrier on "all past work" we just clone that.
#asahi-gpu · 2023-02-26 02:17
linaI still don't know exactly what caches the firmware flushes, when, and how to control any of that. We still have at least 3 mystery bits of freedom in the submission commands.
#asahi-gpu · 2023-02-26 02:17
linaOf course this is all cache issues aside. If we run into cache issues between back to back submissions themselves (like the CPU->GPU one I ran into) we probably need to set more bits in that VDM Barrier command...
#asahi-gpu · 2023-02-26 02:16
linaThis is the "currently vertex and fragment are serialized" issue (if you remove those barriers, vertex for batch+1 can run concurrent with fragment for batch)
#asahi-gpu · 2023-02-26 02:15
linawe don't
#asahi-gpu · 2023-02-26 02:15
linaOn the other hand, we could elide that when we know we do *not* have such a dependency, which I was thinking is the case whenever we submit batches back to back (e.g. whenever we flush_all and there is more than one batch to flush, I think we can avoid that barrier for all but the first, since they're guaranteed not to have any dependencies, otherwise they would have been flushed already), but right now
#asahi-gpu · 2023-02-26 02:14
linaalyssa: All batches are submitted with BARRIER_RENDER | BARRIER_COMPUTE, which tells the backend code to insert a fence on the last render and compute commands submitted (if any - the kernel will elide that if it already got a completion notification and it knows the GPU queue is idle, since at that point it doesn't even have event IDs attached so there is nothing to fence on)
#asahi-gpu · 2023-02-22 17:28
linaThe VDM barrier is userspace, firmware structs are kernel
#asahi-gpu · 2023-02-22 17:27
linaAt least none of the pending unknowns I have helped
#asahi-gpu · 2023-02-22 17:27
linaYeah, that's what that VDM barrier seems to accomplish. I don't know if there is another way, like a field in the firmware structs...
#asahi-gpu · 2023-02-22 17:26
linaI think the flushing is handled by the firmware, I already ran into that with compute and it seems to be related to a sequence number in some firmware structures.
#asahi-gpu · 2023-02-22 16:25
lina^^
#asahi-gpu · 2023-02-22 16:25
linaMESA: warning: [Batch 1] Render Complete: TVB 196608/ 9306112 bytes (0 ovf) G | vtx 0.000005 frag 0.000015
#asahi-gpu · 2023-02-22 16:25
linaMESA: warning: [Batch 4] Render Complete: TVB 196608/ 1048576 bytes (0 ovf) | vtx 0.000006 frag 0.000035
#asahi-gpu · 2023-02-22 16:25
linaMESA: warning: [Batch 0] Render Complete: TVB 196608/ 1048576 bytes (0 ovf) | vtx 0.000005 frag 0.000018
#asahi-gpu · 2023-02-22 16:25
linaMESA: warning: [Batch 2] Render Complete: TVB 3080192/ 1048576 bytes (3 ovf) O | vtx 0.001174 frag 0.000807
#asahi-gpu · 2023-02-22 16:25
linaMESA: warning: [Batch 1] Render Complete: TVB 196608/ 1048576 bytes (0 ovf) | vtx 0.000006 frag 0.000011

plus ›