Thursday, 2021-11-25

Veera[m]lkcl: in the meantime did you tried symbiflow00:01
rscsadoon_albader[m, lkcl: btw, what was the outcome regarding a possible follow-up with Red Hat?01:05
sadoon_albader[mI'll comment on the issue page in a biy02:23
sadoon_albader[mSo this is what I wrote lkcl07:26
sadoon_albader[m[libre-soc]( needs KVM-PR for instruction emulation and development, without KVM-PR we cannot develop new instructions for the POWER ISA.07:26
sadoon_albader[mIn addition, we use KVM-PR to run SVP64, VSX, and SIMD on the chip, which we cannot implement in hardware at this time and as such the software emulation will help greatly for compatability.07:26
sadoon_albader[mKVM-PR is essential for this project, and its removal will cause many issues for us.07:26
sadoon_albader[mIt also does not harm any other component of the system to keep KVM-PR disabled by default and KVM-HV enabled, as is currently the case on all Linux systems, instead of completely removing it.07:26
sadoon_albader[mI don't want to post it just yet so as to not speak on the project's behalf without you looking at it07:27
sadoon_albader[mLet me know07:27
programmerjakeone note: illegal instruction trapping should allow us to emulate vsx/vmx on libre-soc cpus even without kvm at all...though using kvm to do that may be waay easier07:31
lkclsadoon_albader[m, thank you13:05
lkclalso probably worth mentioning, Tiny Code Generator (TCG) in qemu depends on it.13:06
sadoon_albader[mAlright will add that13:31
sadoon_albader[mAnd send13:31
sadoon_albader[m<lkcl> "sadoon_albader, thank you" <- Anytime!13:31
sadoon_albader[mI'm quite sure I've used tcg without KVM-PR before though..13:35
lkcltoshywoshy knows the details: i'd only just heard about it on tuesday14:02
sadoon_albader[mI posted the comment, I'll add the TCG stuff later as an edit if we confirm it14:12
lkcla link back to the IRC chat here might be helpful to give context.14:12
lkcland the one from tuesday14:13
sadoon_albader[mPerfect, I'll link that14:14
sadoon_albader[mThere's a reply14:50
lkclahh nice idea by Laurent Vivier. that's fantastic15:31
sadoon_albader[mSo how would that work15:34
rscThe community likely needs to maintain a separate kernel module as a "kmod" RPM package, I guess16:36
sadoon_albader[m<rsc> "The community likely needs to..." <- That works too if they don't also disable it in qemu17:54
lkclha, got the 2-year-old out-of-order scoreboard engine up and running again21:54
lkclit even includes shadow-protection for branches21:57
lkcland memory-address (LD/ST) hazard protection21:57
Veera[m]lkcl: Can the bug 730 be closed22:01
lkclVeera[m], i need to do some more review.  the symbiflow one shold have been closed and RfP submitted 5 weeks ago22:03 needs quite a lot of tidying up22:03
lkcl 260                 if result < 0:22:04
lkcl 261                     e.intregs[3] = (result + 2**64) & ((2**64)-1)22:04
lkcl 262                 else:22:04
lkcl 263                     e.intregs[3] = result & ((2**64)-1)22:04
lkclsame thing there:22:04
lkclif result < 0: result += 2**6422:04
lkclthen the next line can just be22:04
lkcl e.intregs[3] = result & ((2**64)-1)22:04
Veera[m]why 5 weeks ago?22:05
lkclbecause it is completed22:05
lkcl5 weeks ago (appx)22:05
lkclsame with lines 276-27922:05
lkcland 271-27422:05
lkclbasically anywhere there is expression duplication, it needs to be removed to one single line22:06
Veera[m]if this bug 730 is closed, can it be submitted together so bank costs minimized22:06
lkclyes good idea22:07
lkclcarryout32 needs to be compared != 022:07
lkclthen = carry_out | (carry_out32<<1)22:07
lkcl>>31 is extremely confusing22:07
lkclline 31622:07
Veera[m]carry != 0 gives boolean True or False22:07
lkcl>>> True <<122:08
lkcl>>> False << 122:08
Veera[m]and gives error if not combined with (carry_out32<<1)22:08
lkclso it's perfectly fine.  python "<<" will convert it to 0/122:09
lkclyou are using the project standard python 3.7 so it is fine22:09
lkcl>>> x = True22:10
lkcl>>> x << 522:10
Veera[m]How rfp works with money shared bugs?22:10
lkcl>>> x = False22:10
lkcl>>> x << 522:10
lkclin the "notes" you put "shared with lkcl and {whoeverelse}22:11
lkclrun the budget-sync program and it will generate the notes for you22:11
lkclyou will see a vklr.mdwn file after running that program22:12
Veera[m]for bug 730 how much you are willing to allocate to each person (you have put total to EUR 800)22:14
lkclyes.  there is only another EUR 400 available from the parent, that's why22:18
lkclit should be more.22:18
lkclmaybe... EUR 100 for kyle, EUR 200 for me, and i think really... increase for you to... 700?22:19
lkcl100+200+700=1000 total22:19
lkclthat only leaves EUR 200 left for the parent task, hmmm22:19
lkclwe've basically run out there22:20
lkclfortunately there is another NLnet Grant, approved22:20
Veera[m]oh. ok.22:20
lkcland another budget can be allocated for more unit tests22:21
lkclas part of the verification / validation, including under the Test API22:21
Veera[m]when the Grant was approved?22:21
lkclkylel, if you're interested in that? ^22:21
lkcl4 weeks ago22:21
Veera[m]Whats the status of the Euro 23Million grant!22:22
lkclso although the milestones have to be written (and agreed), at least if work is done there will be payment for it22:22
lkclwe wait to hear from the EU.22:22
lkcltheir website says that they can take between 60 and 180 days to make a decision (!!)22:24
Veera[m]ran nohup budget_sync: it ends with error:23:21
Veera[m]cannot import name 'cached property'  from 'functools'23:21

Generated by 2.17.1 by Marius Gedminas - find it at!