#
#asahi-gpu · 2024-08-08 14:37
alyssaand that off-by-1 propagates to the totally wrong final rsult
#asahi-gpu · 2024-08-08 14:36
alyssaagx has an off-by-1 compared to my cpu
#asahi-gpu · 2024-08-08 14:36
alyssaint((1.0 / float(b)) * (4294966784.0))
#asahi-gpu · 2024-08-08 14:36
alyssathe problem is the value of `rcp` at the start
#asahi-gpu · 2024-08-08 14:36
alyssakarolherbst: ok, this is interesting
#asahi-gpu · 2024-08-08 14:31
alyssapoking at idiv
#asahi-gpu · 2024-08-08 14:26
karolherbsthuh.. this blit bug is weird...
#asahi-gpu · 2024-08-08 14:16
alyssabuilds rusticl
#asahi-gpu · 2024-08-08 13:36
alyssarebuilds piglit
#asahi-gpu · 2024-08-08 13:30
karolherbsthow rude
#asahi-gpu · 2024-08-08 13:30
alyssaexcuse me
#asahi-gpu · 2024-08-08 13:30
alyssa Could NOT find PythonNumpy (missing: PythonNumpy_STATUS) (Required is at
#asahi-gpu · 2024-08-08 13:28
alyssaah
#asahi-gpu · 2024-08-08 13:27
karolherbstand then just bin/cl-program-tester
#asahi-gpu · 2024-08-08 13:26
karolherbstneeds -DPIGLIT_BUILD_CL_TESTS=ON
#asahi-gpu · 2024-08-08 13:26
alyssa.. where is that in piglit?
#asahi-gpu · 2024-08-08 13:25
karolherbstthe piglit one helps, because you won't have to run it for a minute until hitting it :D
#asahi-gpu · 2024-08-08 13:25
alyssauhhh ok
#asahi-gpu · 2024-08-08 13:24
karolherbstin the CL CTS it's integer_ops/test_integer_ops uint_math int_math
#asahi-gpu · 2024-08-08 13:24
karolherbstwith piglits cl-program-tester: https://gist.githubusercontent.com/karolherbst/f49e8f5091887d60bd88da58f54d5b7e/raw/c55ef2d681c3eb6481a0645cf7beb71556181157/test.cl
#asahi-gpu · 2024-08-08 13:23
alyssaand how do I reproduce it?
#asahi-gpu · 2024-08-08 13:23
alyssawhat test?
#asahi-gpu · 2024-08-08 13:17
karolherbstat least reverting 65e431e61a3bace7e50c11d699880ae860f76133 doesn't help.. but maybe the other changes might? But I'm not in the mood of doing reverts across the big nir rename...
#asahi-gpu · 2024-08-08 13:13
karolherbstthat's gonna be pain to debug
#asahi-gpu · 2024-08-08 13:12
karolherbstI've asked austriancoder if it also happens on etnaviv...
#asahi-gpu · 2024-08-08 13:12
alyssauhoh
#asahi-gpu · 2024-08-08 13:09
karolherbstprobably something wrong in nir_lower_idiv
#asahi-gpu · 2024-08-08 13:07
karolherbstoh no.. it's a real bug..
#asahi-gpu · 2024-08-08 11:57
karolherbstsomething with block/grid sizes
#asahi-gpu · 2024-08-08 11:57
karolherbstokay.. I think the int bug above is a rusticl one
#asahi-gpu · 2024-08-08 10:15
karolherbstahh.. and 1Darray
#asahi-gpu · 2024-08-08 10:15
karolherbstbut mhh.. only fails for 1D and 2D
#asahi-gpu · 2024-08-08 10:15
karolherbstimage format that means
#asahi-gpu · 2024-08-08 10:14
karolherbstalyssa: okay, your block/blit stuff works for fp32, but not for fp16
#asahi-gpu · 2024-08-08 09:24
karolherbst"Testing "/" cl_int Verification failed at element 1749980: 0x2ba97a73 / 0xfa8d9df5 = 0xfffffff8, got 0xffffffef" noooo
#asahi-gpu · 2024-08-07 22:41
karolherbstnot sure if he is open for backporting, because uhh.. ABI breakage or something.. but we'll see
#asahi-gpu · 2024-08-07 22:41
alyssai like this plan :3
#asahi-gpu · 2024-08-07 22:41
alyssayeah, if we can get CTS submitted this month, and get spirv link backported in september, we can release the whole thing at XDC no problem
#asahi-gpu · 2024-08-07 22:40
alyssaawesome
#asahi-gpu · 2024-08-07 22:40
alyssahahahah
#asahi-gpu · 2024-08-07 22:40
karolherbstlol
#asahi-gpu · 2024-08-07 22:40
karolherbstI think I can make it work :D
#asahi-gpu · 2024-08-07 22:40
karolherbstairlied is the maintainer 🙃
#asahi-gpu · 2024-08-07 22:39
alyssa(if not we have to copr it and ughh)
#asahi-gpu · 2024-08-07 22:39
karolherbstlet's see...
#asahi-gpu · 2024-08-07 22:39
karolherbstmhh maybe?
#asahi-gpu · 2024-08-07 22:39
alyssacan you get your fix backported into f40 proper?
#asahi-gpu · 2024-08-07 22:39
alyssawe need spirv link bumped in fedora, not sure who to poke about that
#asahi-gpu · 2024-08-07 22:38
karolherbstahh and the spirv-link stuff
#asahi-gpu · 2024-08-07 22:38
karolherbsta lot of upstream rusticl cherry-picks

plus ›