#
…

2023-08-23

00:10cylm__ has quit [Ping timeout: 480 seconds]
00:17alyssa1. correct
00:18alyssa2. uhhhh not 100% sure off hand. I would expect alignment is needed
00:18alyssaagx_nir_lower_sysvals aligns 64-bit to 64-bit boundaries so go with that
00:52mxlcjckmsmz has joined #asahi-gpu
00:52mxlcjckmsmz has quit []
02:02crcvxc has quit [Read error: Connection reset by peer]
02:03user982492 has joined #asahi-gpu
02:07Z750 has quit [Quit: Ping timeout (120 seconds)]
02:08Z750 has joined #asahi-gpu
02:30ourdumbfuture has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
02:33ourdumbfuture has joined #asahi-gpu
02:50user982492 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
02:59user982492 has joined #asahi-gpu
03:11user982492 has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
04:12possiblemeatball has quit [Quit: Quit]
06:16crabbedhaloablut has joined #asahi-gpu
06:53mkurz has quit [Ping timeout: 480 seconds]
07:17seb4nihel has quit [Quit: ZNC 1.8.2 - https://znc.in]
07:18seb4nihel has joined #asahi-gpu
07:45rpirea has joined #asahi-gpu
07:45rpirea has joined #asahi-gpu
08:15flatz has joined #asahi-gpu
08:36rpirea has quit [Remote host closed the connection]
08:37rpirea has joined #asahi-gpu
09:06bisko has joined #asahi-gpu
09:15crcvxc has joined #asahi-gpu
09:25eric_engestromjannau: I think you can close https://github.com/KhronosGroup/VK-GL-CTS/pull/405, I think it might take a while for the khronos maintainer to get around to that
10:04flatz has quit [Remote host closed the connection]
10:11ATiltedTree has quit [Remote host closed the connection]
10:11ATiltedTree has joined #asahi-gpu
10:17cylm__ has joined #asahi-gpu
10:18nepeat has joined #asahi-gpu
10:20flatz has joined #asahi-gpu
10:28flatz has quit [Ping timeout: 480 seconds]
10:30flatz has joined #asahi-gpu
11:15as400 has quit [Remote host closed the connection]
11:16as400 has joined #asahi-gpu
11:18as400 has quit [Remote host closed the connection]
11:18mkurz has joined #asahi-gpu
11:19as400 has joined #asahi-gpu
11:21rpirea has quit []
11:24flatz has quit []
12:37bisko has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
13:21crcvxc has quit [Read error: Connection reset by peer]
13:23lena6 has joined #asahi-gpu
13:24amarioguy has joined #asahi-gpu
14:21iaguis has joined #asahi-gpu
14:34joek has joined #asahi-gpu
14:46crabbedhaloablut has quit [Remote host closed the connection]
14:47crabbedhaloablut has joined #asahi-gpu
14:59arkalon has joined #asahi-gpu
15:02iaguis has quit [Quit: leaving]
15:04iaguis has joined #asahi-gpu
15:12possiblemeatball has joined #asahi-gpu
15:15Misthios1 has joined #asahi-gpu
15:15Misthios1 has quit []
15:16Misthios1 has joined #asahi-gpu
15:18Misthios has quit [Quit: Misthios]
15:19Misthios1 has quit []
15:19Misthios1 has joined #asahi-gpu
15:20Misthios1 is now known as Misthios
15:51possiblemeatball has quit [Quit: Quit]
15:54possiblemeatball has joined #asahi-gpu
15:56joek has quit [Ping timeout: 480 seconds]
16:29cr1901 has quit [Remote host closed the connection]
16:32cr1901 has joined #asahi-gpu
16:40bisko has joined #asahi-gpu
17:14iaguis has quit [Quit: leaving]
17:32i509vcbI believe I've done lowering for base_vertex and num_workgroups correctly. But I need to have these values be set before the main shader executes somehow? I'd imagine I need to do something when recording the command to load the uniform registers with the expected values?
17:33i509vcb(how do I go from the parameters passed into vkCmdDispatch to putting those in the uniform registers I have reserved for that sysval)
17:35i509vcbfor indirect, probably the same thing but I just tell where to load the values from?
17:36i509vcb(well I mean what buffer to read from for indirect)
18:09alyssai509vcb: agx_usc_uniform
18:16alyssaalong with each shader, the driver specifies a list of GPU addresses to be preloaded into uniform register ranges
18:17alyssagrep for agx_usc_uniform to see how that works
18:33alyssaAHAHAHAH
18:33alyssaalyssa you just need to add a load_texture_handle_agx intrinsic
18:33alyssathen the special cases disappear
18:34alyssagood ok i like this.
18:51flom84 has joined #asahi-gpu
19:30billak has joined #asahi-gpu
19:37flom84 has quit [Remote host closed the connection]
19:38c10l2 has joined #asahi-gpu
19:40flom84 has joined #asahi-gpu
19:40c10l has quit [Ping timeout: 480 seconds]
19:43crabbedhaloablut has quit []
19:50billak has quit [Quit: Konversation terminated!]
19:53flom84 has quit [Ping timeout: 480 seconds]
19:58cr1901_ has joined #asahi-gpu
20:02nst has quit [Ping timeout: 480 seconds]
20:03cr1901__ has joined #asahi-gpu
20:06cr1901 has quit [Ping timeout: 480 seconds]
20:11cr1901_ has quit [Ping timeout: 480 seconds]
20:11nst has joined #asahi-gpu
20:16i509vcbis there a way I can dump the state of uniform registers?
20:21alyssanot really
20:21alyssacould hack decode.c
20:21alyssa----------------
20:21alyssaI just discovered that, due to a horrendously small sampler heap, it is impossible to implement standards conformant bindless samplers on M1
20:21alyssaThat means DX12 is impossible on this hardware
20:21alyssaI am really peaved about this. What a stupid corner to cut.
20:22alyssaMakes the "game porting toolkit" announcement even more obnoxious when I now *know* that samplers are necessarily broken in there.
20:22i509vcbI guess I'll have to guess where I went wrong since the dEQP says I have the wrong values
20:22cr1901__ is now known as cr1901
20:27i509vcbsomething a bit simpler, dumping nir?
20:28alyssaAGX_MESA_DEBUG=shaders
20:28alyssaif you need to see NIR before/after every pass, NIR_DEBUG=print but it's very verbose
20:28alyssaASAHI_MESA_DEBUG=trace AGX_MESA_DEBUG=shaders is my go-to
20:35possiblemeatball has quit [Quit: Quit]
20:38c10l2 has quit []
20:38c10l has joined #asahi-gpu
20:39i509vcbThe shader looks fine, I can see the load from the right uniform registers and the store into the buffer where the dEQP test compares the result
20:39i509vcbI guess blame my code for uploading the uniforms?
20:50i509vcbFound my problem, forgot to copy a struct...
20:50i509vcbWell dEQP-VK.compute.builtin_var.num_work_groups passes
20:50i509vcbalyssa: thank you for the help so far
20:53i509vcbnope premature victory it passes some of the tests...
20:56possiblemeatball has joined #asahi-gpu
20:57i509vcbOk now it passes
21:25cr1901_ has joined #asahi-gpu
21:25yeetbiscuit has joined #asahi-gpu
21:26yeetbiscuit has quit []
21:27cr1901 has quit [Read error: Connection reset by peer]
21:27cr1901_ has quit [Read error: Connection reset by peer]
21:28i509vcbHmm I seem to have found a problem with the vk_meta_copy stuff
21:28i509vcbit clobbers the currently bound pipeline
21:28cr1901 has joined #asahi-gpu
21:29i509vcbI guess all the stuff I built to make the sysval stuff expansible in the future is going to sit unsued for a bit...
21:29i509vcbs/unsued/unused
21:30i509vcbat least one is done
21:31alyssaoh, yeah
21:31alyssawe need to do meta save/restore
21:31alyssasee how NVK does it
21:31alyssai509vcb: specifically, see nvk_meta_begin/nvk_meta_end
21:56cr1901 has quit [Quit: Leaving]
21:57crcvxc has joined #asahi-gpu
21:58nela has quit [Quit: bye!]
22:00nela has joined #asahi-gpu
22:04cr1901 has joined #asahi-gpu
22:05ella-0i509vcb: yeah I have those save/restore stuff locally since it was causing some issues with occlusion query tests. will push tomorrow.
22:13cylm__ has quit [Ping timeout: 480 seconds]
22:34i509vcbAnd I realize I did not consider indirect yet in my MR
22:34i509vcbI'll do that later
22:47i509vcb(now I see why the gallium driver doesn't lower every intrinsic the same way)
22:58i509vcbs/intrinsic/sysval
23:02possiblemeatball has quit [Quit: Quit]
23:04possiblemeatball has joined #asahi-gpu
23:10alyssai509vcb: :)
…