04:01HdkR: jekstrand: Looks like upgrading to kernel 5.6.11 didn't solve my alacritty hanging problem
04:01HdkR: May 9 20:59:43 XPS13 kernel: [14185.745270] Asynchronous wait on fence 0000:00:02.0:compton:a6 timed out (hint:intel_atomic_commit_ready+0x0/0x54 [i915])
04:06imirkin: do you have a second gpu in there?
04:07HdkR: nope, only an Ice Lake GPU
10:55mike-south: I'm considering a driver for a custom device (Zynq AXI to Arm device) that need to read complete packets in multiple of 1024 bytes. What would be the proper way to enforce reads in multiple of 1024 bytes from user space?
12:26EdB: It seems that if i want to add print support on Clover for AMD gpu, I need to add a global buffer. However, I suspect that binding a new bufer that is not explicitly referenced will disturb other buffer usage.
12:30EdB: Anyone willing to help me understanding how buffer are managed ?
12:37EdB: on llvm, there is a pass that look for a __printf_alloc, that is expected to return a buffer adr. For the moment, I don't know how to retrieve it and how to bind the buffer
14:59pmoreau: EdB: Nice! I saw the notification about your MR, and will take a look at it later today.
15:56EdB: pmoreau: :) I'll be happy
17:41pmoreau: EdB: Help appreciated on this MR if you have any inputs: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/4979
17:43EdB: ah nice
17:47EdB: pmoreau: why Optional image formats is not started ? If no wrong, lots of stuff is here, at least on the cl runtime side. It is true that the driver side is missing
17:48pmoreau: None of those optional formats were listed in format.cpp
17:49pmoreau: So even the platform side is missing AFAICT
17:58EdB: Read from, write to, copy rectangular regions in progress <-- I didn't test it, but the code seems to be fine
17:58EdB: however, there is no piglit test for it.
17:59pmoreau: Could try the CTS for it; I think they have some tests for ti.
18:03EdB: the code used is the same as normal read, it should be fine :)
18:03EdB: clSetMemObjectDestructorCallback seems to be done for me (again, not tested)
18:05EdB: pmoreau: when you set, in progress, it's because you didn't test it ?
18:06pmoreau: Right; I added a comment on the MR about it.
18:07EdB: ok, some of them seems to be working well for me
18:09pmoreau: Okay; I am currently only running against nv50, so support is still much in progress there and I can’t check that all the features work yet.
18:11EdB: I'm full AMD here
18:11EdB: most it work just fine :)
18:11EdB: most of
18:12pmoreau: Cool! :-)
18:13pmoreau: I’m working on making the CTS OpenCL 1.0 and 1.1 compatible, but not fully there yet.
18:14EdB: do you have a list of failling test ?
18:15pmoreau: Do you think I should mark the rectangular regions and clSetMemObjectDestructorCallback as DONE then?
18:16pmoreau: Not off-hand; computeinfo is now passing. I have fails on basic, but I think it is still trying to test for some >1.1 features.
18:16EdB: I can try a CTS run for them. I usually only do partial test
18:17EdB: I was thing of creating a automatique list for piglit, starting from what tstellar already did
18:18pmoreau: That sounds good, and a CTS run would be very nice!
18:18EdB: also the CTS now seems to have .csv file to group test
18:19pmoreau: Sadly not for 1.0 and 1.1, yet
18:19pmoreau: https://github.com/pierremoreau/OpenCL-CTS/commits/update_basic is my current WIP for updating the CTS to support 1.0 and 1.1
18:20EdB: I use 2.2, since I do unique test run :)
18:23EdB: test_api mem_object_destructor_callback PASS
18:24EdB: mh :)
18:25EdB: Can't test buffer read : ring 2 stalled for more than 1254025msec
18:25EdB: I mess up with my GPU earlier, it need a reboot :D
18:28pmoreau: Messing with GPUs, how so unexpected by people in this channel. :-D
19:07EdB: pmoreau: test_buffers buffer_read_int
19:07EdB: Testing with cl_mem_flags src: CL_MEM_USE_HOST_PTR
19:07EdB: int1 test failed
19:08EdB: je crois que le host_ptr m'avais déja posé problème lors d'un autre test
19:08EdB: que je n'avai spa pu reporduire avec un test simple d'ailleurs :/
19:10EdB: oups sorry
19:10pmoreau: :-/ Might need to look into that, but HOST_PTR creating issues kinda rings a bell
19:10pmoreau: Almost done commenting on your MR btw
19:11EdB: Saddly CTS is not verbose regarding error
19:12pmoreau: Yeah, like they usually don’t print compile logs when program fails to compile.
19:14EdB: nice, however I think I'll re do arg info to no longer copy them on general arg list iterated over what is into m.syms
19:15EdB: s/iterated/and iterate
19:36EdB: pmoreau: thanks for your comments
19:48pmoreau: EdB: No problems! Thanks for looking into that.
19:50EdB: Hehe, it' been a long time since I develop for Clover, I missed it :)
23:31karolherbst: pmoreau: ehhh, you are on my branch, right?