Wednesday, 2023-04-19

*** JTL is now known as jittle00:21
*** jittle is now known as JTL00:21
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC01:20
*** gnucode <gnucode!~gnucode@user/jab> has quit IRC01:26
*** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc01:26
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc01:33
*** gnucode <gnucode!~gnucode@user/jab> has quit IRC02:39
*** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc02:40
*** gnucode <gnucode!~gnucode@user/jab> has quit IRC02:56
lkclprogrammerjake, its purpose is to be absolute and explicit07:50
lkcland also very specifically to avoid confusion caused by different meanings and changes of use to EXTS()07:50
lkclEXTSXL is not defined07:51
lkclwhich then needs explanation07:51
lkclwhich results in questions07:51
lkclwhich results in confusion and delay07:51
lkclthe temporary variable shortens the expressions07:52
programmerjakei didn't put EXTSXL in ls005, EXTS is entirely obvious and already exists07:53
lkclgood catch on the syntax error - as this is not intended to be executed it didn't matter so much07:53
lkclread up earlier07:53
lkclits use and meaning *changes*07:53
lkclthat requires explanation07:53
lkclexplanation results in confusion07:53
lkclconfusion results in delays07:53
lkclexactly: EXTSxl isn't in ls00507:54
lkclit's not part of Power ISA v3.107:54
lkclwhich is what the ISA WG will be comparing this against07:54
programmerjakewell, the expanded version I wrote just uses bit slicing, concatenation and repetition, nothing else07:54
lkclyou added a duplicate copy of the function which contained EXTSXL07:54
lkclwhich is longer07:54
lkcland requires explanation07:54
lkclof a concept that is not part of Power ISA v3.107:55
lkclwhich is what the ISA WG will ask about07:55
lkcli have very specifically thought this through07:55
lkclfrom the perspective of the ISA WG who know nothing about this07:55
lkcland are already confused07:55
lkclthat's why it's written the way it is07:55
lkclto avoid *their* confusion07:55
lkclwhether it "looks nice"07:56
lkclor "is shorter"07:56
lkclis completely irrelevant because that's not the purpose07:56
programmerjakeI *never* added EXTSXL.07:56
lkclthe purpose is: get the concept across to the ISA WG07:56
lkcli know you never added EXTSXL!07:56
programmerjakeEXTS is in v3.1b in book I 1.3.407:56
lkcljacob i've said it three times now07:57
lkclwe *CHANGE* the meaning of EXTS07:57
lkclsecond time of repeating:07:57
lkclplease LOOK07:57
lkclat the earlier examples07:57
lkclthe use of EXTS(SI) is *different* from the use of EXTS() in extsw07:57
lkclthat's going to cause MASSIVE confusion to the ISA WG when they read this07:57
lkclcan you acknowledge that i have a specific purpose for which this pseudocode is intended?07:59
programmerjakeso, i'm trying to say that the unexpanded version i wrote could have been deleted but the expanded version communicates exactly what you're were trying to afaict, so it didn't need to be reverted07:59
lkclyes it did because you removed the temporary variable07:59
lkclyou reduced the clarity07:59
programmerjakeok, i disagree07:59
lkclyou made unnecessary changes07:59
lkclmore than that07:59
lkcli had a particular speech in my head prepared08:00
lkclit doesn't even matter if you disagree!08:00
lkclyou are not the one presenting to the ISA WG!08:00
lkcli am!08:00
lkclyou changed what i was to be presenting *without consulting me*!08:01
lkclso i get half-way through the presentation and i go "wtf that's not the pseudocode i memorised"08:01
lkcland i've just wasted their time08:01
lkcland looked like a complete fool08:02
programmerjakeok, i still fixed the pseudocode which was factually intention was to fix the factual errors, sorry for changing it beyond that08:02
lkclplease *don't* change things that i have to present without asking me08:02
programmerjakei didn't realize you were presenting just then, sorry08:02
lkcland appreciated spotting the extra bracket08:02
programmerjakeso, i'd like to implement those extsb/h/w changes in the simulator and unit tests tomorrow, sound good?08:08
programmerjakeoh, also: if XLEN = 64 then RT <- in[0] * 56 || in[1:7] # 8->6408:10
programmerjakethat needs to be 57, not 5608:10
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has joined #libre-soc08:26
fedorafanare you into cpu development08:32
jnfedorafan: that's what happens here08:36
fedorafanvery interesting do you have knowledge about hardware backdoors08:37
programmerjakenot of any specific hardware backdoors, but yes I know how to create them if i wanted...08:38
programmerjakethough a lot of hardware has enough bugs that you don't necessarily need to bother, e.g. row-hammer08:39
fedorafancan you give me an simple explanation08:39
programmerjakeof hardware backdoors or row-hammer?08:40
jnwhat is your understanding so far, fedorafan?08:40
programmerjakehardware backdoors you just insert a circuit that changes something that causes software to misbehave in the desired fashion, e.g. if there was a special instruction you could run from user-level software that would write to the kernel's memory in order to grant the program root priveleges08:42
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has quit IRC08:44
programmerjakerow-hammer is where memory cells are soo small these days that the cells can be randomly changed by repeatedly writing to physically nearby memory cells, this can be exploited by getting some important kernel memory (e.g. page tables) to be right next to memory you control, which you can repeatedly write to, thereby changing the kernel memory in a way that can sometimes be exploited08:46
programmerjakethis is part of why ddr5 requires memory chips to have error correction internally, because they're small enough that cells can be often randomly flipped and the error correction catches and tries to fix that08:53
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has joined #libre-soc08:58
fedorafanprobably this is the wrong channel but i will idle for some hours before leaving09:02
fedorafanlets say 2 hours09:02
programmerjakeif you're looking for's irc chat, you came to the right place. we're in different countries all over, so aren't always online the same time09:04
programmerjakeso you may have to wait up to 24h or so for some people to reply09:05
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has quit IRC09:15
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has joined #libre-soc09:22
*** pangelo[m] <pangelo[m]!~pangeloma@2001:470:69fc:105::3ec5> has quit IRC10:00
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has quit IRC10:16
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has joined #libre-soc10:26
lkclprogrammerjake, yes sure go ahead, it's a good idea10:26
lkcl"hardware backdoors" can include an entire CPU, sitting idle, only 2,000 to 3,000 gates10:28
lkclin amongst billions, who's going to spot that?10:28
lkclparticularly when PHYs often *require* an actual entire CPU to do their job10:28
programmerjakelkcl: bugfix:
*** fedorafansuper <fedorafansuper!~fedorafan@user/fedorafan> has joined #libre-soc11:11
*** fedorafan <fedorafan!~fedorafan@user/fedorafan> has quit IRC11:11
lkclmagic. thx. night.11:18
fedorafansuperwrong place so ciao12:25
fedorafansuperoh oh oh oh12:25
*** fedorafansuper <fedorafansuper!~fedorafan@user/fedorafan> has left #libre-soc12:26
markosprogrammerjake, btw, the meeting was over last night so didn't have time to tell you about the dark energy uniformity, a few months ago I had read an article about exactly that13:01
markosturns out the universe is in fact anisotropic and one of the reasons is *perhaps* non-uniformity of dark energy13:02
markosinterestingly enough, about 20+ years ago there was an experiment done which measured the speed of light having tiny variations but only around specific angles13:03
markoswhich essentially made the universe behave like a huge crystal13:03
markosand even more interestingly, when I mentioned this experiment to one of my professors -who does Differential Geometry- he was extremely interested13:03
markoshe was working on Finsler Spaces, a anisotropic superset of Reimann Geometry where the speed of light is actually a vector and not a constant13:04
markosand that mathematical model was a good fit13:05
markosI don't have the experiment article at hand but I can try to find it13:05
markosturns out he's still working on that13:08
*** pangelo[m] <pangelo[m]!~pangeloma@2001:470:69fc:105::3ec5> has joined #libre-soc14:41
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC15:38
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc15:39
Ryuno-KiAndrJaen<markos> "he was working on Finsler Spaces..." <- *Riemann15:57
markosunfortunately, I'm not that old so that my professor was Riemann :)15:58
markosthat was late 90s, my professor was P. Stavrinos, and he was/still is active in Finsler Spaces from what I see15:59
markoshe's a mathematician not a physicist, but some of his colleagues were physicists and were interested in the cosmological application16:00
markosie, what if the universe was anisotropic, what would the mathematical model be? (<- Finsler Space) What experiment would prove that, etc etc16:01
Ryuno-KiAndrJaenmarkos: My analytics prof was physicist.16:03
Ryuno-KiAndrJaenBut to be frank, it's the first time I hear about Finsler Space.16:03
Ryuno-KiAndrJaenI left university like ten years ago but only discovered some parts of the math community on the Fediverse last year (mainly mathstodon).16:04
Ryuno-KiAndrJaenBefore I had contact to teachers at best.16:04
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC16:17
markosguys, I /might/ miss tonight's meeting, social reasons, I will join if I return in time19:04
programmerjakethx for sharing that link markos! interesting...19:28
*** toshywoshy <toshywoshy!> has quit IRC19:52
*** toshywoshy <toshywoshy!> has joined #libre-soc19:52
*** octavius <octavius!> has joined #libre-soc20:58
sadoon[m]Starting the process of rebuilding bootstrap debian for sffs by building without modifications first21:21
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC22:43
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc22:52
*** octavius <octavius!> has quit IRC23:41

Generated by 2.17.1 by Marius Gedminas - find it at!