#asahi-gpu · 2022-10-28 14:05
_jannau_they started selling square greyish samples for $19 a piece last year#asahi-gpu · 2022-10-28 14:05
svenit's their interconnect which connects most (all?) thing inside the SoC#asahi-gpu · 2022-10-28 14:02
alyssathere a luxury clothing line I didn't hear about?#asahi-gpu · 2022-10-28 14:02
alyssa"Apple Fabric"#asahi-gpu · 2022-10-28 05:46
phirethough, I can't actually remember what level we called "job"#asahi-gpu · 2022-10-28 05:43
phirethe mask is surely per job, so you can just put the alpha blended verts in a new job#asahi-gpu · 2022-10-28 05:41
phirehmm, does it really need a full flush to switch vertex modes?#asahi-gpu · 2022-10-28 05:41
phiredepending on the exact impelementation, you might have to allocate some of the tile buffer for a copy of the depth buffer#asahi-gpu · 2022-10-28 05:40
linaAt the fragment level though, right? Not the vertex level...#asahi-gpu · 2022-10-28 05:40
phirealpha blending already puts the TBDR renderer into a very sub-optimal mode#asahi-gpu · 2022-10-28 05:40
linaAnyway, I should get ready for the stream (and have a sandwich or something) ^^#asahi-gpu · 2022-10-28 05:39
phirethey already do#asahi-gpu · 2022-10-28 05:39
linaYeah, but if they have to do some kind of flush around alpha draws...#asahi-gpu · 2022-10-28 05:38
phirebecause alpha blending is somewhat expensive everywhere#asahi-gpu · 2022-10-28 05:38
phirelike more than 95% of your geometery in a typical game will be opaque#asahi-gpu · 2022-10-28 05:38
linawwwww#asahi-gpu · 2022-10-28 05:38
phirewhy not make it a bit more expensive?#asahi-gpu · 2022-10-28 05:37
phireyeah, and apple's docs go on and on about how alpha blending is expensive#asahi-gpu · 2022-10-28 05:37
linaThat seems really wasteful...#asahi-gpu · 2022-10-28 05:37
phirewhich would have a side-effect that it has to fall back to single-cluster vertex processing when doing alpha blending#asahi-gpu · 2022-10-28 05:36
phireactually, maybe they cheat and their depth sorting implementation also sorts by primitive ID#asahi-gpu · 2022-10-28 05:36
lina*that#asahi-gpu · 2022-10-28 05:36
linaIt's possible they have some sort of pass to count the total number of primitives first, then split based on tat?#asahi-gpu · 2022-10-28 05:35
linaFor a single draw op it's trivial, just split the primitives into equal 8ths. More interesting is what happens with multiple draw ops.#asahi-gpu · 2022-10-28 05:34
linaAlso, they might have primitive IDs in the TVB data anyway, I think they need that later during rasterization? So maybe it doesn't matter...#asahi-gpu · 2022-10-28 05:34
linaAnd then if you can do that, chaining just works#asahi-gpu · 2022-10-28 05:34
phireand much larger diffrences if blending or image load/store are enabled#asahi-gpu · 2022-10-28 05:34
linaProbably in draw order...#asahi-gpu · 2022-10-28 05:34
linaSo that tells me the first cluster processed all the clear ops and every tile, and then the other clusters progressively processed additional slices of geometry for busy tiles#asahi-gpu · 2022-10-28 05:33
linaYeah, so the question is whether they do something at the dispatch level to make sure every cluster gets an ordered slice of the overall workload. I think they do, because the first tilemap is ~complete, and they get sparser as you step through the buffer.#asahi-gpu · 2022-10-28 05:33
phirethere are very slight graphical diffrences (were depths are equal) if you render triangles out of API order#asahi-gpu · 2022-10-28 05:32
phireexcept you have to respect ordering, because stupid reasons#asahi-gpu · 2022-10-28 05:32
linaSo it's very likely they just trivially chain those#asahi-gpu · 2022-10-28 05:31
linaAnd since that next level is allocated out of the heap, it has to have chain pointers#asahi-gpu · 2022-10-28 05:31
linaThe top level tile array just has 5-byte memory pointers to the initial position of every tile, and then there's at least one intermediate meta/pointer array structure before the raw data (it's been a while since I dumped these buffers...)#asahi-gpu · 2022-10-28 05:31
phireit's a lot less waseful when you put it that way#asahi-gpu · 2022-10-28 05:30
phireoh, point. we are just merging the indices#asahi-gpu · 2022-10-28 05:30
linaIt's just chaining together slices for each tile#asahi-gpu · 2022-10-28 05:30
linaThe actual vertex data pages are definitely not merged, that'd be wasteful#asahi-gpu · 2022-10-28 05:30
linaSo when I say "merging" it's just writing out some pointers#asahi-gpu · 2022-10-28 05:29
linaAnd then the rest of the top level pointers are probably chained together at the next level, or something like that#asahi-gpu · 2022-10-28 05:29
linaThe top level tile array is just 5 bytes per tile, a memcpy of that seems wasteful but is trivial#asahi-gpu · 2022-10-28 05:29
linaThe structures are trees, so they can probably just chain pointers#asahi-gpu · 2022-10-28 05:26
phireand reading and merging them in a single pass#asahi-gpu · 2022-10-28 05:25
phireHang on, is there any reason not to just implement reading from N TVBs (one per cluster) duing the 3d invocation?#asahi-gpu · 2022-10-28 05:22
phirebut merging also has better scaling?#asahi-gpu · 2022-10-28 05:22
phirebut merging just seems wasteful#asahi-gpu · 2022-10-28 05:22
phireyes, there would be scaling issues#asahi-gpu · 2022-10-28 05:22
linaBut between clusters, don't think so#asahi-gpu · 2022-10-28 05:21
linaI mean I'm sure they have that within a cluster