2021-01-11
00:04aratuk has joined #asahi-gpu
00:09aratuk has quit [Ping timeout: 265 seconds]
00:30bloomOh, and by properly splitting out unaligned/aligned portions, the M1 gets the job done in 0.33s while the RK3399 is at ~1.58s
00:31bloomWhat's intriguing there is that the optimization helps only 4% on the Rockchip but 23% on the Apple.
00:32bloomThe optimization reducing ALU while keeping memory access constant. So it seems not only is the ALU on the Mac all around faster, the memory is _so_ much better.
00:32davidrysk[m]This was making the rounds the other day, but these tests are too elementary to really provide any interesting info
00:32bloomI don't think that should come as a surprise to anyone, memory performance is critical to both CPU and GPU performance, and if you want to kick ass at both for an SoC, optimizing the hell out of the memory access is a 2 for 1 deal..
00:34davidrysk[m]I wish we knew more about Apple's (CPU) microarchitecture. From what I can tell, they're still using the scheduler model that was intended for Cyclone (A7) in LLVM, and that chip had much less instruction level parallelism available.
00:36bloomIf this is about your NEON routine, yes, you win, I'm over it, we both know you're better at writing tiling routines than me :P
00:39icecream95bloom: IIRC My NEON tiling routine was "only" about 3x faster
01:01icecream95 has quit [Ping timeout: 265 seconds]
01:22icecream95 has joined #asahi-gpu
02:04klaus has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
02:24krbtgthuh, i thought other companies had different strategies, but it's been a while since i looked into it
02:37davidrysk[m]https://developer.amd.com/wordpress/media/2012/10/gdc2008_ribble_maurice_TileBasedGpus.pdf
02:37davidrysk[m]found this with a quick google :)
03:50ransom has joined #asahi-gpu
03:52Tokamak has joined #asahi-gpu
04:05aratuk has joined #asahi-gpu
04:11aratuk has quit [Ping timeout: 264 seconds]
05:32ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
05:36Tokamak has quit [Remote host closed the connection]
05:44Tokamak has joined #asahi-gpu
06:40ransom has joined #asahi-gpu
06:41q3k|m has quit [Remote host closed the connection]
06:57ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
07:08Tokamak has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
07:40klaus_ has joined #asahi-gpu
08:07aratuk has joined #asahi-gpu
08:12aratuk has quit [Ping timeout: 264 seconds]
09:13icecream95 has quit [Ping timeout: 240 seconds]
09:46marcanPSA: we are adjusting the RE policy and banning userspace GPU stack disassembly, since it is unlikely that approach will be necessary. at the same time, this channel is no longer allowed for any kind of disasm/binary RE
09:46marcan changed the topic of #asahi-gpu to: Asahi Linux: porting Linux to Apple Silicon macs | GPU / 3D graphics stack black-box RE and development (NO binary reversing) | Keep things on topic | GitHub: https://alx.sh/g | Wiki: https://alx.sh/w | Logs: https://alx.sh/l/asahi-gpu
09:47marcanwe are not banning GPU kernel driver disassembly (but see https://alx.sh/re for the policy), but that discussion will be confined to #asahi-re
09:49marcanthe discussion here should focus on black-box RE of the userspace stack (what bloom is doing), and on hardware details and kernel driver development discussion (strictly *results* i.e. documentation resulting from kernel driver RE are allowed only, absolutely no discussion of the actual kext code)
11:03DarthCloud has quit [Ping timeout: 240 seconds]
11:05DarthCloud has joined #asahi-gpu
12:09aratuk has joined #asahi-gpu
12:13aratuk has quit [Ping timeout: 265 seconds]
13:40mogery has joined #asahi-gpu
14:00aratuk has joined #asahi-gpu
14:05aratuk has quit [Ping timeout: 272 seconds]
14:07q3k|m has joined #asahi-gpu
14:13robher has joined #asahi-gpu
15:06sumoon[m] has joined #asahi-gpu
15:19klaus_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
15:30HeN has quit [Quit: Connection closed for inactivity]
15:34Tokamak has joined #asahi-gpu
15:58klaus_ has joined #asahi-gpu
16:01aratuk has joined #asahi-gpu
16:04davidrysk[m]bloom: Is there any documentation on retrieving compiled and assembled object code of a shader?
16:06aratuk has quit [Ping timeout: 264 seconds]
16:10ransom has joined #asahi-gpu
16:26ransom_ has joined #asahi-gpu
16:29ransom has quit [Ping timeout: 272 seconds]
16:55ransom has joined #asahi-gpu
16:57modwizcode has joined #asahi-gpu
16:57ransom_ has quit [Ping timeout: 256 seconds]
17:03mogeryy has joined #asahi-gpu
17:04mogery has quit [Ping timeout: 260 seconds]
17:30bloom(there is some code I need to push to help with that, but my local setup is still a giant mess of shell scripts split acorross two machines)
17:33davidrysk[m]bloom: "Does that require dual licensing the other files too?" APSL sections 1.5 and 4 seem to say no
17:50HeN has joined #asahi-gpu
18:02aratuk has joined #asahi-gpu
18:06aratuk has quit [Ping timeout: 256 seconds]
18:15tiagom has joined #asahi-gpu
18:39tiagom has quit [Quit: tiagom]
18:41tiagom has joined #asahi-gpu
20:38icecream95 has joined #asahi-gpu
21:12mogeryy has quit [Quit: Leaving]
21:20ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
21:20wolf511[m] has joined #asahi-gpu
21:22DarthCloud has quit [Ping timeout: 240 seconds]
21:28DarthCloud has joined #asahi-gpu
22:03aratuk has joined #asahi-gpu
22:07aratuk has quit [Ping timeout: 240 seconds]
22:41tiagom has quit [Quit: tiagom]
23:26ransom has joined #asahi-gpu
23:39ransom has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]