#asahi-gpu · 2024-01-01 19:31
i509vcbhmm okay thanks for the info#asahi-gpu · 2024-01-01 05:45
i509vcbAlthough I assume there is a reason why vk-meta is currently using a fragment shader for that?#asahi-gpu · 2024-01-01 05:43
i509vcbUnless I've missed something, we'd need to write in the worst case 8192 layers of a 2d copy for each layer of the 3d image#asahi-gpu · 2024-01-01 05:42
i509vcbUpon further thought for vk-meta-copy to properly do 3d copies without doing something horrible for performance like submitting 8192 draw calls for a full 8192x8192x8192 copy for an image, compute shaders with imageLoad and imageStore might be best?#asahi-gpu · 2024-01-01 05:35
i509vcbI think this might hint at some bug with the gpu recovery?#asahi-gpu · 2024-01-01 05:34
i509vcbI wonder if this info would be helpful for figuring out how to prevent the unrecoverable timeouts I get. I noticed with deqp-runner if I set the maximum number of jobs to 1 or 2 (although I think the unrecoverable timeouts will occur with less frequency) I don't get unrecoverable timeouts. By default deqp-runner will use all my cores (8 on the M2 air)#asahi-gpu · 2023-12-31 19:27
i509vcbUNORM and SNORM are pretty much SFLOAT but with a fixed range#asahi-gpu · 2023-12-31 19:25
i509vcbI don't get any 2d_to_2d failures if I disable those formats, but I get a feeling there is a more precise issue at hand here#asahi-gpu · 2023-12-31 19:24
i509vcbIs AGX_TEXTURE_TYPE_FLOAT actually UNORM/SNORM?#asahi-gpu · 2023-12-31 06:35
i509vcbdEQP-VK.api.copy_and_blit.core.image_to_image.all_formats.color.2d_to_2d.r16_sint.r16_sfloat.* seems especially suspicious#asahi-gpu · 2023-12-31 06:35
i509vcbhttps://gist.github.com/i509VCB/2dba6f60e8e73ccde116cb965de3739d#asahi-gpu · 2023-12-31 06:35
i509vcbI noticed a pattern that no formats with a source sfloat numeric type (except for the exception of packed formats being the destination and depth/stencil) fail#asahi-gpu · 2023-12-31 06:18
i509vcbI guess it seems valid to be Float then?#asahi-gpu · 2023-12-31 06:18
i509vcbHmm pipe_format says SFLOAT is FLOAT except for stencil formats it seems#asahi-gpu · 2023-12-31 06:16
i509vcbI grepped out 2d to 2d tests and noticed sfloat is the vast majority of the failures#asahi-gpu · 2023-12-31 06:15
i509vcbI noticed that Texture Type in cmdbuf.xml does not list SFLOAT, I assume this means the hardware can't do anything with an SFLOAT numerical type?#asahi-gpu · 2023-11-26 20:12
i509vcb(Ella did give me the vk-meta code)#asahi-gpu · 2023-11-26 20:08
i509vcbIf I disable all the compressed formats then its Pass: 8170, Fail: 708, Crash: 2, Flake: 1 for the copy_and_blit.image_to_image tests#asahi-gpu · 2023-11-26 20:05
i509vcbwhen a meta shader is built agx_nir_lower_tilebuffer seems to have problems when emitting code with a compressed format (like PIPE_FORMAT_ETC2_R11_SNORM for where agx_tilebuffer_physical_format returns PIPE_FORMAT_NONE). I'm guessing that these types of formats need some sort of emulation for reads and writes?#asahi-gpu · 2023-11-24 02:51
i509vcbSeems like 2 ways to do the same thing?#asahi-gpu · 2023-11-24 02:51
i509vcbAlso there seems to be a texture size intrinsic and texoo_txs?#asahi-gpu · 2023-11-24 02:50
i509vcbI guess while you are here, I'm struggling to understand how I'd change a binded access (if that is even the right term) to a bindless for the bindless image stuff the compiler wants #asahi-gpu · 2023-11-24 02:48
i509vcbYeah I mentioned it yesterday to ella and she said she has patched #asahi-gpu · 2023-11-23 17:18
i509vcbodd... Commenting out some stuff in agxv while trying to do the proper bindless image lowering I can crash my entire session when running vkcube#asahi-gpu · 2023-11-21 23:20
i509vcbabout 75% of the draw tests in the dEQP that are supported crash in vk_meta on that one assert#asahi-gpu · 2023-11-21 23:19
i509vcbregarding the comment in vk_meta, what specifically was broken? https://gitlab.freedesktop.org/asahi/mesa/-/blob/agxv/main/src/vulkan/runtime/vk_meta_copy.c?ref_type=heads#L406#asahi-gpu · 2023-11-21 03:40
i509vcbVK_QUEUE_GLOBAL_PRIORITY_REALTIME_KHR does exist which is technically higher than all values, which could make sense if everything else is shifted down by 1#asahi-gpu · 2023-11-21 03:39
i509vcbGallium and agxv currently use priority = 1#asahi-gpu · 2023-11-21 03:38
i509vcbFor drm_asahi_queue_create::priority: 0 = EGL_CONTEXT_PRIORITY_HIGH_IMG/VK_QUEUE_GLOBAL_PRIORITY_HIGH_KHR, 1 = EGL_CONTEXT_PRIORITY_MEDIUM_IMG/VK_QUEUE_GLOBAL_PRIORITY_MEDIUM_KHR, 2 = EGL_CONTEXT_PRIORITY_LOW_IMG/VK_QUEUE_GLOBAL_PRIORITY_LOW_KHR, 3 = ????#asahi-gpu · 2023-11-21 02:49
i509vcblina: timelines appear to work: https://gitlab.freedesktop.org/asahi/mesa/-/merge_requests/219#asahi-gpu · 2023-11-12 06:21
i509vcbsorry for the annoyance#asahi-gpu · 2023-11-12 06:21
i509vcbHmm more discussion, it seems to be a misunderstanding of EGL#asahi-gpu · 2023-11-12 00:12
i509vcbI have no idea what semantics that should have#asahi-gpu · 2023-11-12 00:12
i509vcbWould be great if I could find documentation for drmSyncobjWait somewhere #asahi-gpu · 2023-11-11 22:45
i509vcbHmm this wasn't noticed because it seems to be a gap in the piglit tests#asahi-gpu · 2023-11-11 22:41
i509vcbWhen I create an EGLSync and immediately check if it is signalled, this appears to be causing EGL to say a Sync that has never been used is already signalled#asahi-gpu · 2023-11-11 22:38
i509vcbThis check seems to be wrong? https://gitlab.freedesktop.org/asahi/mesa/-/blob/main/src/gallium/drivers/asahi/agx_fence.c#L56#asahi-gpu · 2023-11-09 20:44
i509vcb(the second part of this is that for us Wayland compositor folks, we have no real way to communicate that info since we can't use VK or GL WSI)#asahi-gpu · 2023-11-09 20:43
i509vcbI've ended up in a discussion in the smithay matrix around EGL_KHR_partial_update, is agx in theory capable of using the damage regions to eliminate drawing places on screen the application said didn't get changed?#asahi-gpu · 2023-10-31 00:41
i509vcbWell sure it won't ever exist if we get nuclear armageddon in the next week lol#asahi-gpu · 2023-10-31 00:40
i509vcbSo uhh M3 now exists#asahi-gpu · 2023-10-24 21:06
i509vcbNVK appears to treat it like that as well#asahi-gpu · 2023-10-24 21:05
i509vcbI'd assume NULL_HANDLE would just unbind the handle#asahi-gpu · 2023-10-24 21:05
i509vcbvkCmdBindIndexBuffer2KHR is allowed to have VK_NULL_HANDLE for the buffer argument if I am to believe from the "buffer must be a valid VkBuffer handle" statement?#asahi-gpu · 2023-10-21 01:59
i509vcbCan't be layout, MTLDrawIndexedPrimitivesIndirectArguments is the same layout as VkDrawIndexedIndirectCommand#asahi-gpu · 2023-10-20 23:09
i509vcbThe only other thing I could think of for my CmdDrawIndirectIndexed implementation being wrong is that Apple expects a different layout for the indirect buffer in IndirectIndexed than what Vulkan and GL provide?#asahi-gpu · 2023-10-20 19:06
i509vcbI'm suspecting the command stream having two VDM_STATEs might be causing my problems?#asahi-gpu · 2023-10-20 19:06
i509vcbella-0: For agxv should only one VDM_STATE exist per axv_emit?#asahi-gpu · 2023-10-17 19:46
i509vcbVUID-vkCmdDrawIndirect-drawCount-00476 pretty much#asahi-gpu · 2023-10-17 19:45
i509vcbit seems to work with something equvilant to VkDrawIndirectCommand but the spec allows setting a stride as long it's a multiple of 4