#
✕ nick:lina
#asahi-gpu · 2022-10-27 03:01
linaHuh.
#asahi-gpu · 2022-10-27 03:00
linaI see the stencil aux buffers too, so this was good to name those fields...
#asahi-gpu · 2022-10-27 02:59
linaWhat the heck. Well, that's a new one.
#asahi-gpu · 2022-10-27 02:59
linadepth_flags = 0x5000000000c4154
#asahi-gpu · 2022-10-27 02:56
linaopens up WebGL Aquarium
#asahi-gpu · 2022-10-27 02:55
linaSome random webGL demo?
#asahi-gpu · 2022-10-27 02:55
linaWhat's a good macOS app to exercise the depth buffer? I just did boot->safari->youtube and there isn't a single render with depth...
#asahi-gpu · 2022-10-27 02:54
linaFB compression isn't about size, it's about bandwidth, so the buffers are actually slightly larger (extra meta)
#asahi-gpu · 2022-10-27 02:54
linaYou always need more space any time you use compression, by definition
#asahi-gpu · 2022-10-27 02:54
linaYes, exactly
#asahi-gpu · 2022-10-27 02:53
linaThis is pretty typical framebuffer compression stuff
#asahi-gpu · 2022-10-27 02:53
linaThe metadata buffer has one byte per 8x4 pixel block in the main buffer
#asahi-gpu · 2022-10-27 02:53
linaThe main buffer is the same, it just has compressed tiles and the metadata buffer tells it how they were compressed.
#asahi-gpu · 2022-10-27 02:52
linaWhich is needed for compression
#asahi-gpu · 2022-10-27 02:52
linaThey aren't separate, the extra buffer is the *metadata* buffer
#asahi-gpu · 2022-10-27 02:50
linaBut then why is *store* being forced on, but not *load*?
#asahi-gpu · 2022-10-27 02:48
linaFirmware bug would be kind of weird... especially since it's a shared codebase
#asahi-gpu · 2022-10-27 02:47
linaMaybe this is an outright bug? G13X always does partial depth stores compressed?
#asahi-gpu · 2022-10-27 02:46
linaThat's... exactly what you would expect of compressed data loaded as uncompressed
#asahi-gpu · 2022-10-27 02:46
linaActually this fits the image I'm seeing... I get lots of mini-tiles with stuff in the top left.
#asahi-gpu · 2022-10-27 02:45
linaYeah, but not having 0x44 seems to break the overflow, like it stores compressed but loads uncompressed.
#asahi-gpu · 2022-10-27 02:44
linaphire: Partially, even!
#asahi-gpu · 2022-10-27 02:44
linaIn old G13G traces (that I now found - there are few of them)
#asahi-gpu · 2022-10-27 02:44
linaI see a 1:1 correlation between 0x44 and the accel/meta buffer existing
#asahi-gpu · 2022-10-27 02:43
linaOK, so this is related to compression... but then why is G13D enabling compression even without them, partially, then breaking?
#asahi-gpu · 2022-10-27 02:42
linaNo wait they do
#asahi-gpu · 2022-10-27 02:42
lina*And* those logs don't have the compression meta buffer on G13G
#asahi-gpu · 2022-10-27 02:42
linaDon't think so w
#asahi-gpu · 2022-10-27 02:41
linaAlso I found logs with 0x80044 that were definitely taken on an M1...
#asahi-gpu · 2022-10-27 02:41
linaIt just breaks
#asahi-gpu · 2022-10-27 02:41
linaAnd yet I still get it using the compression buffer without that
#asahi-gpu · 2022-10-27 02:39
linaInteresting...
#asahi-gpu · 2022-10-27 02:39
linaOh, I have logs with those flags in the past... but I don't recall ever taking logs on an M1 Ultra before?
#asahi-gpu · 2022-10-27 02:38
lina(Still storing to the accel buffer either way though)
#asahi-gpu · 2022-10-27 02:38
linaLooks like it needs 0x44 ORed in
#asahi-gpu · 2022-10-27 02:37
linaSetting zls_control differently fixes the bunny on TVB overflows
#asahi-gpu · 2022-10-27 02:37
linaI just got there w
#asahi-gpu · 2022-10-27 02:33
linaI just don't see them in any HV logs I have lying around
#asahi-gpu · 2022-10-27 02:33
linaI also need to check again on the M1 Mini, since I didn't actually look there
#asahi-gpu · 2022-10-27 02:33
linaYeah, they show up in the command buffers
#asahi-gpu · 2022-10-27 02:30
linaNeed to clean it up
#asahi-gpu · 2022-10-27 02:30
linaNot yet
#asahi-gpu · 2022-10-27 02:29
linaAlso I just realized my bunny is very, very broken with TVB overflows...
#asahi-gpu · 2022-10-27 02:21
linaAnd at least a couple general bitmask like things with different values
#asahi-gpu · 2022-10-27 02:20
linaLet me look at what changed...
#asahi-gpu · 2022-10-27 02:20
linaEither global or per batch
#asahi-gpu · 2022-10-27 02:20
linaIt could be a config change
#asahi-gpu · 2022-10-27 02:20
linaSo maybe intermediate Z flushes on Ultra using compression is a new G13X thing
#asahi-gpu · 2022-10-27 02:20
linaBut I still haven't seen this on M1, only Ultra
#asahi-gpu · 2022-10-27 02:19
linaYou just fault for a few frames iff the geometry is complex enough

plus ›