Friday, 2022-09-16

lkcltests passed, all pushed00:06
lkclprogrammerjake, anything under bug #254 i did say hasn't been approved yet00:15
lkcli put one in with a copy of the explicit instructions to Michiel so that he could review and action the changes (which include utf-8, which isn't strictly "3D") hence the need for him to review and approve it00:16
lkclcan you please re-submit, with *only* the Transcendentals for now? give it a few days i'll let you know when Michiel's reviewed00:18
lkclghostmansd[m], you can also put in #899
lkclbut don't put any others under 3D/Vulkan yet (anything connected to #254)00:18
lkclplease remove the "submitted" dates from #794 and #911 and wait until i've got confirmation00:20
lkcland redo *only* an RFP for #89900:20
lkclghostmansd[m], you can also put in EUR 2500 for #898
lkclbut put it as an RFP under the *parent*, #25200:23
lkclthen use the "notes" from this page
lkcli'll email you instructions00:23
programmerjakelkcl, done00:28
lkclnuisance the need to run things by michiel, but it means better auditing. sigh00:29
lkclgot it00:32
* lkcl checking00:32
lkclyeah all good00:32
lkclas #252 has had all its tasks 100% completed the RFPs can go in no problem00:33
lkcl#254 not so much00:33
programmerjakek, ttyl -- woke up really early (3am) so need sleep00:34
*** zemaye <zemaye!> has joined #libre-soc02:47
*** zemaye <zemaye!> has quit IRC03:05
*** yambo <yambo!~yambo@> has quit IRC06:04
*** yambo <yambo!~yambo@> has joined #libre-soc06:19
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC07:30
*** octavius <octavius!> has joined #libre-soc09:31
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC10:23
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc10:24
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC10:42
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc10:45
*** midnight <midnight!~midnight@user/midnight> has quit IRC11:12
*** midnight <midnight!~midnight@user/midnight> has joined #libre-soc11:13
lkclghostmansd[m], programmerjake 794 871 911 and 917 have now all been added to the MoU so RFPs for those can be submitted12:05
ghostmansd[m]Oh cool!12:06
ghostmansd[m]Thanks, will submit today12:07
lkcljust sending you an email on the "comments" to be included12:09
lkclyou must put the RFP under #252 (covering multiple bugreports) and under #911.12:10
lkclprogrammerjake, vulkan RFP approved by NLnet12:11
lkcland Video RFP approved as well12:11
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC12:35
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc12:36
*** philpax_ <philpax_!> has quit IRC13:01
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC13:01
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc13:02
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC13:06
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc13:07
*** ghostmansd <ghostmansd!> has joined #libre-soc13:54
*** ghostmansd <ghostmansd!> has quit IRC14:09
*** octavius <octavius!> has quit IRC14:17
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc15:22
lkclping ghostmansd[m] python3 src/openpower/sv/trans/  > /tmp/f15:26
lkclERROR: test_2_d_custom_op (__main__.SVSTATETestCase)15:26
lkcli posted more in #917 i think15:26
ghostmansd[m]Aha, checking15:27
lkclthat was just after the addition of SignedOperand15:28
ghostmansd[m]Yeah this one is signed15:28
ghostmansd[m]Likely DOperatorDX was broken, IIRC addpcis used it15:29
lkclprogrammerjake, you can put in.... err... #794 and #791 now15:29
lkclironically fmvis and fishmv are unsigned decimal15:29
lkclbut still use D-Field15:30
ghostmansd[m]We should sort these out on per-insteuction basis, then...15:30
ghostmansd[m]Because currently they use signed.15:31
ghostmansd[m]Or, well, on per-form format.15:31
*** octavius <octavius!> has joined #libre-soc15:31
ghostmansd[m]lkcl, by the way, these d0,d1,d2 are documented as signed15:31
ghostmansd[m]So we might need docs update15:32
ghostmansd[m]Or have different names for operands15:33
ghostmansd[m]Aaaaah yes15:36
ghostmansd[m]Ok I see the problem, this gets the same D operand as any other instruction with D operands15:36
lkclgood job there's unit tests, eh? :)15:37
ghostmansd[m]Yes, that's really helps a lot!15:37
ghostmansd[m]OK all but SPR is fine15:37
ghostmansd[m]I recall you did some tricks with it15:38
lkcli couldn't help but hear that in a sarcastic voice :)  "grumble grumble oh thanks a loooot" :)15:38
lkclyes, it works fine now15:38
lkclre-run pywriter noall sprset15:38
lkclwhatever errors you're getting will go walkies15:38
ghostmansd[m]Aaaaah yes this is always forgotten15:38
lkcli changed the pseudocode "n <- spr[blah] || spr[blah]"15:38
lkclto just15:39
lkcl"n <- spr"15:39
lkcloh you saw i added madd* yesterday?15:39
lkclthey passed first time15:39
ghostmansd[m]Pushed fix for addpcis15:40
programmerjakelkcl, put in rfp for #794 and #91115:40
ghostmansd[m]Hm, SPR still doesn't work on my side15:40
lkclprogrammerjake, yep got it, all good15:40
lkclghostmansd[m], 1 sec15:40
programmerjake#791 is the wrong bug id15:41
lkclghostmansd[m], you're on commit c22949b84c8d7087?15:41
*** ghostmansd <ghostmansd!> has joined #libre-soc15:42
ghostmansdThe one I just pushed15:42
ghostmansdprevious is c22949b84c8d708705357f8926016fd11104a4d615:42
lkclack 1 sec15:42
ghostmansdI see sprset.mdwn change, though15:42
lkclpython3 src/openpower/sv/trans/  > /tmp/f15:43
lkclRan 9 tests in 4.658s15:43
ghostmansd[m]Ah damn15:43
ghostmansd[m]I know, Talos version needs to be pulled15:43
lkcldoh :)15:43
ghostmansd[m]Yeah fine now15:44
ghostmansd[m]False alarm15:44
lkclwark-wark :)15:45
ghostmansd[m]Ok, back to D operand for addpcis/fishmv/fmvis15:45
ghostmansd[m]This should be signed for addpcis but not for FP, right?15:45
ghostmansd[m]so we have different D operand composed from dX chunks15:46
programmerjakenote on the latest commit that finished ci on openpower-isa.git master (c22949b8):15:47
programmerjakeFAILED src/openpower/decoder/test/
programmerjakeFAILED src/openpower/sv/trans/
ghostmansd[m]Yeah should be fixed in the recent master15:51
programmerjakei'd expect test_minor_19 to still be broken, lkcl can you fix it since you broke it by changing minor_19.csv to pattern rather than integer15:53
ghostmansd[m]lkcl, programmerjake, would you object if I add disassembly produced with sv_binutils_fptrans as one of the tests to
ghostmansd[m]This won't increase CI terribly, but still the test will take longer (all fptrans instructions with min/max operands permutations).16:10
ghostmansd[m]I'll need to tune sv_binutils_fptrans to make it reusable (it produces the output in binutils-like form), but still I find the idea attractive.16:11
*** ghostmansd <ghostmansd!> has quit IRC16:15
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC16:32
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc16:33
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC16:39
programmerjakesounds fine to me!16:40
programmerjakelkcl: yup, test_minor_19 failed for oprnpower-isa 008d1d2716:55
programmerjakeplease fix it soon16:56
ghostmansd[m]I can do it later today, approximately 4 hours later. Is the only issue that it should be pattern, not an integer?16:59
programmerjakeno, the issue is that the csv was changed to be a pattern rather than an integer and the test wasn't updated at the same time, so it fails to parse the csv properly17:00
*** octavius <octavius!> has quit IRC17:18
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC17:39
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc17:40
ghostmansd[m]Ok, got it!17:53
ghostmansd[m]I've just submitted tests for fptrans, now we have 10 tests, and one of them checks 232 instructions.17:55
lkcltests still good as of ae87619:03
*** octavius <octavius!> has joined #libre-soc19:08
programmerjakeat ae876:20:14
programmerjakeFAILED src/openpower/decoder/test/
*** octavius <octavius!> has quit IRC20:19
ghostmansd[m]Checking this20:37
ghostmansd[m]Just got to the laptop20:37
*** ghostmansd <ghostmansd!> has joined #libre-soc20:42
ghostmansd[m]Nah it's already pattern. Checking what went wrong exactly.20:42
ghostmansd[m]Ok it simply needs to be marked as opint in test it seems.20:48
ghostmansd[m]That said, we should drop these explicit opint eventually... I just don't have time for it now, but the idea is the same: we can iterate over insndb, each instruction has the "section" field where it's specified whether it's opint (or, actually, there's a enum of pattern/integer).20:50
ghostmansd[m]Pushed the fix, CI should be fine now.20:50
*** lxo <lxo!> has joined #libre-soc21:45
programmerjakeci passed, thx!21:48
*** ghostmansd <ghostmansd!> has quit IRC21:49
lkclhmmm that's a neat trick on test_9_fptrans sv/trans/test_pysvp64dis.py22:45
lkcli wonder if it can be replicated for auto-generation of every single instruction/22:45
lkclplus some automatic unit test assembler22:46
lkcl   asmop 1,122:46
lkcl   asmop 31,7 # 1st field is 5-bit, 2nd field is 3-bit22:46
lkcland all permutations in between, 0 1 2 4 8 16 then 3122:47
lkclhmm hmm22:48

Generated by 2.17.1 by Marius Gedminas - find it at!