| *** gnucode <gnucode!~gnucode@user/jab> has quit IRC | 02:54 | |
| *** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC | 03:00 | |
| *** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc | 03:01 | |
| openpowerbot | [irc] <programmerjake> lkcl: I closed and assigned budget to the ternlogi and grev[w][i] bugs (grevlut left for later task), so there's money available for you, see bugzilla emails | 03:36 |
|---|---|---|
| openpowerbot | [irc] <programmerjake> I left the full budget estimation for the parent task (bitmanip) for later, currently it only has the grev and ternlogi budget | 03:37 |
| openpowerbot | [irc] <lkcl_> the grev* instructions all have to go, sigh. grevluti is expensive but incredibly powerful | 03:47 |
| openpowerbot | [irc] <programmerjake> k | 03:49 |
| openpowerbot | [irc] <lkcl_> on its own, being able to express that many regular constants, and they're so clearly useful? | 03:52 |
| openpowerbot | [irc] <lkcl_> svp64:vector-immediate (when added later), hoo-boy :) | 03:52 |
| *** leah <leah!~leah@libreboot/developer/leah> has quit IRC | 04:34 | |
| *** leah <leah!~leah@libreboot/developer/leah> has joined #libre-soc | 04:36 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC | 04:51 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc | 05:29 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC | 05:37 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc | 06:48 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC | 07:35 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@broadband-90-154-80-239.ip.moscow.rt.ru> has quit IRC | 07:37 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc | 08:30 | |
| markos_ | programmerjake, I'm using this implementation, which already has a unit test: https://github.com/floodyberry/ed25519-donna | 08:40 |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC | 09:01 | |
| *** markos_ <markos_!~markos_@static062038151250.dsl.hol.gr> has joined #libre-soc | 09:16 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc | 09:16 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.160.71> has joined #libre-soc | 09:35 | |
| *** cesar <cesar!~cesar@2804:14d:688a:59ea:682d:e88b:7a7a:882f> has joined #libre-soc | 09:38 | |
| openpowerbot | [irc] <snowwolflair[m]> Am i in the correct place now? | 09:41 |
| lkcl | yes you are | 09:45 |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.160.71> has quit IRC | 09:57 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@broadband-90-154-80-239.ip.moscow.rt.ru> has joined #libre-soc | 09:58 | |
| openpowerbot | [irc] <lkcl_> snowwolflair[m], ping test with notification | 10:05 |
| *** octavius <octavius!~octavius@212.237.221.242> has joined #libre-soc | 10:12 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC | 10:25 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc | 10:38 | |
| *** cesar <cesar!~cesar@2804:14d:688a:59ea:682d:e88b:7a7a:882f> has quit IRC | 10:39 | |
| *** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC | 11:01 | |
| *** cesar <cesar!~cesar@2804:14d:688a:59ea:682d:e88b:7a7a:882f> has joined #libre-soc | 11:11 | |
| *** cesar <cesar!~cesar@2804:14d:688a:59ea:682d:e88b:7a7a:882f> has quit IRC | 11:12 | |
| openpowerbot | [irc] <snowwolflair[m]> 👍️ | 11:15 |
| openpowerbot | [irc] <snowwolflair[m]> cesar: thank you for the help | 11:15 |
| *** ghostmansd[pc] <ghostmansd[pc]!~ghostmans@broadband-90-154-80-239.ip.moscow.rt.ru> has quit IRC | 11:54 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has quit IRC | 12:09 | |
| *** markos_ <markos_!~markos_@user/markos/x-1838887> has joined #libre-soc | 12:10 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@broadband-90-154-80-239.ip.moscow.rt.ru> has quit IRC | 13:28 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.54.179> has joined #libre-soc | 13:29 | |
| openpowerbot | [irc] <lkcl_> octavius: you asked snowwolflair[m] (privately) for "where is PO9", you can now ask (publicly) on the IRC channel which will help engagement and discussion - see this commit https://git.libre-soc.org/?p=libreriscv.git;a=commitdiff;h=902ecc and the one after it | 13:37 |
| openpowerbot | [irc] <lkcl_> toshywoshy, i made some (minor) clarifications to the PO9 RFC https://git.libre-soc.org/?p=libreriscv.git;a=commitdiff;h=ebc54753 | 13:37 |
| openpowerbot | [irc] <lkcl_> octavius ^ (also relevant for your private question which for future reference can be public in this channel now that snowwolflair[m] is in it) | 13:38 |
| openpowerbot | [irc] <snowwolflair[m]> Yes that will help me handle any questions | 13:39 |
| openpowerbot | [irc] <lkcl_> the Makefile "pulls in" the relevant pages into RFCs https://git.libre-soc.org/?p=libreriscv.git;a=blob;f=openpower/sv/rfc/Makefile;hb=HEAD | 13:40 |
| openpowerbot | [irc] <lkcl_> and auto-generates Latex which then auto-generates PDFs | 13:40 |
| openpowerbot | [irc] <lkcl_> so you see here ls001.po9: https://git.libre-soc.org/?p=libreriscv.git;a=blob;f=openpower/sv/rfc/ls001.po9.mdwn;hb=HEAD | 13:40 |
| openpowerbot | [irc] <lkcl_> at the *very bottom* of that page is a markdown "inline include" that pulls in the entire po9_encoding page https://libre-soc.org/openpower/sv/po9_encoding/ | 13:41 |
| openpowerbot | [irc] <lkcl_> 81 | 13:41 |
| openpowerbot | [irc] <lkcl_> 82 [[!inline pages="openpower/sv/po9_encoding" raw=yes ]] | 13:41 |
| openpowerbot | [irc] <lkcl_> programmerjake and i spent some time writing pandoc plugins that "do approximately the same job as ikiwiki" so that we *do not* have to write markdown spec files... | 13:42 |
| openpowerbot | [irc] <lkcl_> ... oh and then **** around for months "hand-converting" them to latex | 13:43 |
| openpowerbot | [irc] <lkcl_> no | 13:43 |
| openpowerbot | [irc] <lkcl_> write once, treat the markdown as "Canonical" and create a pandoc plugin that *generates* RFCs, *generates* PDFs, *generates* quotes diffs quotes for the OPF ISA WG | 13:44 |
| openpowerbot | [irc] <lkcl_> ohhh that's interesting https://groups.google.com/g/comp.arch/c/C8HJVqcXGkQ/m/5Q0uIn-sAQAJ | 13:56 |
| openpowerbot | [irc] <lkcl_> IBM's z-Series team added something-like-AVX to the z-Series mainframe ISA?? | 13:56 |
| openpowerbot | [irc] <lkcl_> oink? | 13:56 |
| markos_ | lkcl, well, there is the Z Vector extension, has been there for years, it's very similar to VSX | 14:01 |
| markos_ | since z13, I've done the port to Eigen :-) | 14:02 |
| lkcl | iiinteresttiiiing... | 14:13 |
| lkcl | ah ok it's "VSX". does it have masks? (predication per-element)? | 14:13 |
| *** psydroid <psydroid!~psydroid@user/psydroid> has joined #libre-soc | 14:13 | |
| markos_ | no | 14:13 |
| lkcl | then it's a PackedSIMD Extension not a Vector Extension. | 14:13 |
| markos_ | yes exactly | 14:13 |
| lkcl | i reaallly wish people would get this through their heads. sigh | 14:14 |
| markos_ | :) | 14:14 |
| markos_ | well, it's still fixed-width vectors that are a pita to use but hey, it has worked for 20+ years now | 14:15 |
| markos_ | even compilers call them __vector(s) :) | 14:15 |
| octavius | lkcl, thanks for your response. | 15:26 |
| octavius | The question I asked was: | 15:26 |
| octavius | "How is PO9 opcode space going to be used (and if it is enough space)?" | 15:26 |
| octavius | "I seem to remember PO9 is 64-bit instruction space, but that Luke came up with a technique to put 32-bit instructions within this opcode space." | 15:26 |
| octavius | Now after reading the po9_encoding, discussion, and ls001 pages it seems that depending on opf's decision on where to allocate, this encoding space will be used for most of libre-soc's extensions (except svp64 management insns). | 15:26 |
| octavius | Also a question for markos_ (or ghostmansd[m]). Trying to run the mp3 "make tests" in openpower/media with openpower (latest git d530f465 on master). What branch do I need to use to have binutils (-mlibresoc) flag? | 15:38 |
| markos_ | octavius, keep in mind that I haven't really finished those | 15:39 |
| octavius | AH OK | 15:40 |
| octavius | oops | 15:40 |
| octavius | sorry caps lock | 15:40 |
| markos_ | but you would need the binutils as compiled in the dev-env-scripts | 15:40 |
| octavius | any examples I can run? | 15:40 |
| octavius | ok | 15:40 |
| markos_ | yeah, we were out of time for the mp3 programs and we just left them for another time | 15:41 |
| markos_ | av1, libvpx, chacha20, are ones that should work | 15:41 |
| markos_ | unless there were changes in binutils that break compilation | 15:41 |
| markos_ | but I'm not aware of any | 15:42 |
| octavius | Thanks markos_, will setup binutils and try it out :) | 15:55 |
| markos_ | np | 15:55 |
| openpowerbot | [irc] <lkcl_> octavius: the only one that works is the apply_window_float one. | 16:24 |
| openpowerbot | [irc] <lkcl_> neither ghostmansd nor konstantinos know about this because they were written by a contributor that sadly took a compromised vaccine a couple of years ago | 16:25 |
| openpowerbot | [irc] <lkcl_> they haven't been updated since, and use the *old* syntax (0.v instead of *0) | 16:25 |
| markos_ | well, I do have some half-finished code here, that I was working on using the new syntax, but it's in really early stages and nothing works -obviously | 16:26 |
| openpowerbot | [irc] <lkcl_> the only tests working the last time i tried them are the mp3_0 ones, which you have to "drill down" and actually read the README | 16:26 |
| markos_ | never committed it because it was terrible :) | 16:26 |
| openpowerbot | [irc] <lkcl_> doesn't matter - please commit it anyway! | 16:26 |
| openpowerbot | [irc] <lkcl_> that's *really* important. | 16:27 |
| openpowerbot | [irc] <lkcl_> you did _some_ work - someone else can carry it on, plus we can get another grant for it | 16:27 |
| openpowerbot | [irc] <lkcl_> please please please always always always *commit your work*. | 16:27 |
| markos_ | ok, I thought it didn't matter about this task but sure, no problem, I'll just put a WARNING: INCOMPLETE CODE :) | 16:27 |
| markos_ | ok, will commit it today even | 16:27 |
| markos_ | I actually did forget about this code entirely | 16:28 |
| openpowerbot | [irc] <lkcl_> the opensourceecology project has a hard rule that if you leave work for longer than 60 seconds that has not been committed and pushed, it's a "red flag" | 16:28 |
| openpowerbot | [irc] <lkcl_> (!) | 16:28 |
| markos_ | hahaha, 60 seconds :) | 16:28 |
| openpowerbot | [irc] <lkcl_> yes! | 16:28 |
| markos_ | try shifting that << 4 | 16:28 |
| openpowerbot | [irc] <lkcl_> their rules are designed for large-contributor-bases at weekend workshops | 16:28 |
| markos_ | and that would be close to my actual number | 16:29 |
| markos_ | I don't want to polute my git history with crap commits, basically | 16:29 |
| markos_ | anyway, I'll just commit this one now | 16:29 |
| openpowerbot | [irc] <lkcl_> > " polute my git history" | 16:30 |
| openpowerbot | [irc] <lkcl_> really don't worry about it | 16:30 |
| openpowerbot | [irc] <lkcl_> i routinely commit multiple single-line commits, up to 30 a day | 16:30 |
| openpowerbot | [irc] <lkcl_> or i do "atomic" commits if you know what i mean by that. "one purpose, one commit" is the general rule | 16:31 |
| openpowerbot | [irc] <lkcl_> this is all in HDL_workflow | 16:31 |
| openpowerbot | [irc] <lkcl_> if you find during the writing of the commit that you are tempted to use the word "and", then you should have committed what you had done *way* before | 16:32 |
| markos_ | yes, I understand, there are 2 lines of thought here, frequent commits whatever the result, and atomic commits that do not break compilation or CI | 16:32 |
| markos_ | I tend to do the latter | 16:32 |
| markos_ | this is from working in larger projects with many people | 16:33 |
| markos_ | if one commits something experimental that breaks compilation for everyone else | 16:33 |
| openpowerbot | [irc] <lkcl_> frequent commits whatever the result are a no-no because we work in master-branch typically | 16:33 |
| markos_ | that's really bad | 16:33 |
| openpowerbot | [irc] <lkcl_> yes absolutely | 16:33 |
| openpowerbot | [irc] <lkcl_> again this is in HDL_workflow already | 16:33 |
| openpowerbot | [irc] <lkcl_> but likewise "fixed X *AND* added Y *AND* solved Z *AND* worked-around problem W" - this is also a no-no | 16:34 |
| markos_ | also, a change might span several files, so I prefer to group my commits | 16:34 |
| markos_ | or rather not group the commits | 16:34 |
| openpowerbot | [irc] <lkcl_> because if X turns out to be a mistake, you have to revert X *AND* Y *AND* Z *AND* W | 16:35 |
| markos_ | but use a single commit that is around a single change, regardless of the number of files | 16:35 |
| openpowerbot | [irc] <lkcl_> which is a serious amount of work | 16:35 |
| markos_ | yes exactly that | 16:35 |
| openpowerbot | [irc] <lkcl_> hence the rule: "single-purpose, single-commit" | 16:35 |
| markos_ | completely agree | 16:35 |
| markos_ | I guess I'm just slower then | 16:35 |
| openpowerbot | [irc] <lkcl_> again - also in HDL_workflow - what i have done routinely is "care less" about stuff that is "leaf-node" (and has no unit tests at all) | 16:36 |
| openpowerbot | [irc] <lkcl_> slower and more meticulous :) | 16:36 |
| markos_ | I mean 30 commits a day sounds like I would be with 1kg coffee in my veins :D | 16:36 |
| openpowerbot | [irc] <snowwolflair[m]> Absolute minimum level of coffee to do anything useful 🤣 | 16:37 |
| openpowerbot | [irc] <lkcl_> it's an asperger's thing. "hard grind and bloody-minded persistence" over "slow meticulous thought" | 16:37 |
| markos_ | (well 1kg would probably mean I'd be dead :D really) | 16:37 |
| openpowerbot | [irc] <lkcl_> and your liver would be entirely empty of glucose :) | 16:38 |
| markos_ | but but caffeine!!! | 16:38 |
| openpowerbot | [irc] <lkcl_> caffeine triggers your liver to release glucose stores (thank you for that tip snowwolflair[m]) | 16:39 |
| openpowerbot | [irc] <snowwolflair[m]> 👍️ | 16:39 |
| markos_ | hm, I just did a git pull and I saw that some .py files in openpower/decoder/isa are not being tracked by git anymore, are they autogenerated? can I remove them? | 16:43 |
| openpowerbot | [irc] <programmerjake> yes | 16:54 |
| openpowerbot | [irc] <lkcl_> ok so "pywriter" and "pyfnwriter" auto-generate files from the markdown | 16:54 |
| openpowerbot | [irc] <programmerjake> run make, it will tell you all about it | 16:54 |
| * openpowerbot [irc] <lkcl_> leaves it to programmerjake to explain | 16:54 | |
| markos_ | lkcl, ah, my bad, I had not converted to new syntax yet, it's still the old syntax, what I did start doing was I added the C wrappers for the python simulator so I could do the same thing as I did for av1 | 16:54 |
| markos_ | programmerjake, thanks will do | 16:55 |
| openpowerbot | [irc] <programmerjake> basically, those are the isa pseudocode markdown files translated to python, I moved the canonical output location to its own folder, and added an error message if you don't clean up the old files | 16:55 |
| openpowerbot | [irc] <programmerjake> that way we can add new markdown files without worrying about conflicts with other random files in the folder | 16:56 |
| openpowerbot | [irc] <programmerjake> the only autogenerated file that stays in the old location is all.py | 16:57 |
| markos_ | perfect, thanks, worked like a charm | 16:58 |
| markos_ | nice touch with the warning too | 16:58 |
| openpowerbot | [irc] <programmerjake> do try and make it a habit to always run make after running git pull, i optimized it to be quite fast (few sec) | 16:58 |
| openpowerbot | [irc] <programmerjake> thx | 16:59 |
| markos_ | yup, it's much faster than before that's true | 16:59 |
| markos_ | nice job | 16:59 |
| openpowerbot | [irc] <programmerjake> maybe i'll make the python files hash the input files at load time and complain if they don't match what they were last time they were compiled | 17:01 |
| *** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc | 17:27 | |
| markos_ | programmerjake, fwiw, I have already begun drafting out the implementation, when it's done we can always look back and test it against other libraries, but for now, the included unit test is good enough | 17:37 |
| markos_ | I mean re ed25519 | 17:38 |
| markos_ | lkcl, no need for escalation, but it's just even if I wanted to test multiple implementations, there is just too little time | 17:38 |
| markos_ | we can worry about this later | 17:39 |
| markos_ | fwiw, if it's wrong it's wrong and it should be caught with this single unit test, once we get a working piece of code, adding more unit tests will be much easier | 17:40 |
| *** jn <jn!~quassel@user/jn/x-3390946> has quit IRC | 17:48 | |
| *** jn <jn!~quassel@2a0a-a540-c88c-0-20d-b9ff-fe49-15fc.ipv6dyn.netcologne.de> has joined #libre-soc | 17:49 | |
| *** jn <jn!~quassel@2a0a-a540-c88c-0-20d-b9ff-fe49-15fc.ipv6dyn.netcologne.de> has quit IRC | 17:49 | |
| *** jn <jn!~quassel@user/jn/x-3390946> has joined #libre-soc | 17:49 | |
| openpowerbot | [irc] <cesar[m]1> markos_, want to try a new bit file on your Nexys Video? I have a good feeling about this one. | 18:20 |
| markos_ | hi cesar, I just did, I wanted to tell you but I didn't see you online, I forget that many people use the matrix bridge | 18:22 |
| markos_ | so it did build and get programmed just fine | 18:23 |
| markos_ | green light is on, but I don't see anything else, what is this supposed to do? | 18:23 |
| markos_ | I missed the earlier message | 18:24 |
| openpowerbot | [irc] <cesar[m]1> Hmm, it's not supposed to be built, I had it pre-built. | 18:25 |
| openpowerbot | [irc] <cesar[m]1> Are you sure it's the latest one? | 18:25 |
| openpowerbot | [irc] <cesar[m]1> It's in the symbiflow directory. | 18:26 |
| markos_ | I called ./build.sh, shouldn't I have? | 18:32 |
| markos_ | build.sh does seem to build switchled.bit file | 18:32 |
| markos_ | ah sorry | 18:33 |
| markos_ | I was in the wrong dir | 18:33 |
| markos_ | programming now | 18:33 |
| markos_ | some of the leds are flashing now | 18:33 |
| openpowerbot | [irc] <cesar[m]1> Yay! | 18:34 |
| markos_ | the last 4 leds | 18:34 |
| openpowerbot | [irc] <cesar[m]1> It's supposed to be a binary counter. | 18:34 |
| markos_ | right | 18:34 |
| openpowerbot | [irc] <cesar[m]1> ... with 4 leds, exactly. | 18:34 |
| markos_ | yes, it seems to be working :) | 18:34 |
| markos_ | congrats! | 18:34 |
| markos_ | so you got the board to work, this is a nice milestone! | 18:35 |
| markos_ | next step: linux :D | 18:35 |
| openpowerbot | [irc] <programmerjake> now, for something more complex... | 18:35 |
| markos_ | joking :) | 18:35 |
| openpowerbot | [irc] <programmerjake> maybe that uart demo i wrote? | 18:36 |
| markos_ | sure, I hope I have a 2nd usb cable to use :) | 18:36 |
| markos_ | yes I do | 18:37 |
| openpowerbot | [irc] <programmerjake> in gitlab-build-utils branch of utils.git | 18:37 |
| markos_ | one is plugged in the prog port, the other will go to uart port | 18:37 |
| markos_ | well, as soon as cesar tells me how to do it | 18:37 |
| openpowerbot | [irc] <programmerjake> it may need some adjusting to get the right fpga pins and clock rate | 18:38 |
| *** octavius <octavius!~octavius@212.237.221.242> has quit IRC | 18:38 | |
| openpowerbot | [irc] <programmerjake> will need to be added to https://git.libre-soc.org/?p=utils.git;a=blob;f=uart_demo.py;h=76a8b0900a9c80c7a8307a242414f97ed2179f1c;hb=refs/heads/gitlab-build-utils#l242 | 18:39 |
| markos_ | ok, probably during the weekend | 18:42 |
| openpowerbot | [irc] <programmerjake> all of the adjusting should be in nmigen_boards, the only thing that should need to change in utils.git is adding the right platform class to the list of options | 18:44 |
| markos_ | well, cesar used symbiflow here | 18:44 |
| markos_ | and I see a nexys_video definition file | 18:45 |
| markos_ | that he had to add | 18:45 |
| markos_ | so to make it work on nmigen it will require some work still | 18:45 |
| openpowerbot | [irc] <programmerjake> nmigen should only need the board class, which has all the options for pins and toolchains | 18:46 |
| openpowerbot | [irc] <programmerjake> you could probably base it on the nexys4ddr which might be similar: https://gitlab.com/nmigen/nmigen-boards/-/blob/a949308a9ecac1cb96ad8e71436b4e0ee3f5070f/nmigen_boards/nexys4ddr.py | 18:51 |
| markos_ | nmigen_boards already has a board class, but it doesn't seem to work, at least it didn't when I tried with the cmds given by cesar and octavius | 18:53 |
| markos_ | we already found a problem in x3csprog where it masked one bit out in the board id value and it failed to detect it even | 18:54 |
| markos_ | and from what I see x3csprog is completely unmaintained so good luck having this bug fixed | 18:55 |
| markos_ | we could pick it up ourselves, or we could just pick a different tool to use | 18:55 |
| markos_ | but cesar's approach worked, symbiflow + openFPGAloader | 18:56 |
| markos_ | I guess it doesn't hurt using openFPGAloader to flash the board, but yes, it would be nice to have nmigen fixed | 18:57 |
| openpowerbot | [irc] <programmerjake> the board class supplies the command to program the fpga, so that should be easy to change | 18:58 |
| markos_ | yes, if you know what to change | 18:59 |
| openpowerbot | [irc] <programmerjake> it's right here: https://gitlab.com/nmigen/nmigen-boards/-/blob/a949308a9ecac1cb96ad8e71436b4e0ee3f5070f/nmigen_boards/nexys4ddr.py#L178 | 18:59 |
| openpowerbot | [irc] <programmerjake> except in the nexys video file | 19:00 |
| markos_ | yes I know | 19:00 |
| markos_ | that's what cesar has already done in his own tree :) | 19:00 |
| openpowerbot | [irc] <cesar[m]1> Thanks, programmerjake , I'll update the board class and give Symbiflow and openFPGAloader as defaults for the Nexys Video. | 19:01 |
| openpowerbot | [irc] <cesar[m]1> I'll also update the Symbiflow dev script to include support to the xc7a200. | 19:02 |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.54.179> has quit IRC | 19:12 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.42.200> has joined #libre-soc | 19:13 | |
| *** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc | 19:22 | |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.42.200> has joined #libre-soc | 19:34 | |
| ghostmansd | A weird question. We cannot have operands which cannot fit into 32 bits, even after decoding, can we? | 19:40 |
| ghostmansd | For vanilla instructions, the question makes no sense, since we basically would've been unable to fit such operands. But what about SVP64 and vector instructions? | 19:41 |
| ghostmansd | For SVP64 registers, we have 2**7 range. Anything else I'm missing? | 19:42 |
| ghostmansd | I think such large values cannot fit anywhere, even with prefixes. But I want to be sure. | 19:43 |
| *** tplaten <tplaten!~tplaten@195.52.183.118> has joined #libre-soc | 19:50 | |
| openpowerbot | [irc] <programmerjake> po1 immediates are 34 bit... | 20:04 |
| openpowerbot | [irc] <programmerjake> lkcl's vector immediate idea allows up to 512 byte immediates afaict | 20:05 |
| openpowerbot | [irc] <programmerjake> no, wait 2048 byte | 20:05 |
| openpowerbot | [irc] <programmerjake> 64 subvectors of 4x64 bits: 64*4*8=2048 | 20:06 |
| openpowerbot | [irc] <programmerjake> individual elements max out at 64-bit though | 20:07 |
| openpowerbot | [irc] <programmerjake> (though they'll likely be extended to 128-bit at some point in the more distant future) | 20:08 |
| ghostmansd | I mean encoding. | 20:16 |
| ghostmansd | Even if value of operand turns out to be bigger... is there chance that it's encoded more than 32 bits? | 20:16 |
| ghostmansd | These PO1 immediates... Do they kinda magically become 34 bit (like it happens with say GPR operand remap), or they MUST be 34-bit? | 20:17 |
| ghostmansd | I suspect the former. | 20:18 |
| ghostmansd | If the immediates go over the same process which happens for GPR/FPR/CR remap -- fine, I don't care. If this just extends vanilla PPC instruction, I'll handle it separately. | 20:19 |
| ghostmansd | If, however, we have a monolithic instruction with an operand bigger than 32 bits, this is more difficult. | 20:20 |
| ghostmansd | I assume we don't have these though, as Luke was very explicit about having custom 64-bit instructions which don't have conterparts in vanilla world. | 20:21 |
| *** tplaten <tplaten!~tplaten@195.52.183.118> has quit IRC | 20:22 | |
| *** rsc <rsc!~robert@fedora/rsc> has quit IRC | 20:22 | |
| openpowerbot | [irc] <programmerjake> the vector immediates are explicit up to 2048 byte immediates | 20:22 |
| *** rsc <rsc!~robert@linuxnetz.de> has joined #libre-soc | 20:22 | |
| *** rsc <rsc!~robert@fedora/rsc> has joined #libre-soc | 20:22 | |
| openpowerbot | [irc] <programmerjake> that said, i would just handle them as a special case of SVP64 | 20:23 |
| openpowerbot | [irc] <programmerjake> (realistically, i think they're a bad idea that shouldn't be implemented, but luke wants them...) | 20:24 |
| openpowerbot | [irc] <programmerjake> for now, you can assume immediates fit in 32 bits and change it later if necessary | 20:25 |
| openpowerbot | [irc] <programmerjake> since vector immediates are not ready for binutils or the simulator yet | 20:26 |
| ghostmansd | OK, got it! Thank you as always, Jacob! | 20:32 |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.42.200> has quit IRC | 21:18 | |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.42.200> has quit IRC | 21:18 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.57.65> has joined #libre-soc | 21:19 | |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.57.65> has joined #libre-soc | 21:20 | |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.57.65> has quit IRC | 21:50 | |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.57.65> has joined #libre-soc | 21:50 | |
| openpowerbot | [irc] <lkcl_> > maybe i'll make the python files hash the input files | 23:07 |
| openpowerbot | [irc] <lkcl_> please don't start implementing in python the job that's done by GNU make | 23:07 |
| openpowerbot | [irc] <lkcl_> that route ended up with scons. | 23:08 |
| openpowerbot | [irc] <lkcl_> and it was quite hilarious to see a patch submitted for python to replace autoconf with scons... which is written 100% in python. | 23:08 |
| * openpowerbot [irc] <lkcl_> facepalm | 23:08 | |
| openpowerbot | [irc] <programmerjake> the issue is that users don't realize they need to run make, and then wonder why it doesn't work | 23:09 |
| openpowerbot | [irc] <programmerjake> so the hash checking tells them to re-run make | 23:10 |
| lkcl | it is better to use make's built-in ability to detect timestamps than to piss about trying to replicate the features of make. | 23:12 |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.57.65> has quit IRC | 23:12 | |
| lkcl | please, really: don't spend time on replacing/duplicating the capabilities of a 40-year-old unix tool. | 23:12 |
| openpowerbot | [irc] <programmerjake> they won't realize they need to run make so make can detect stuff needs to be rebuilt. | 23:13 |
| lkcl | just... stop. please. listen. | 23:13 |
| lkcl | *don't* replicate what is in GNU make in python. | 23:14 |
| openpowerbot | [irc] <programmerjake> k, i'm thinking for 24h | 23:14 |
| lkcl | ok good. additionally there are much higher priorities than "optimising the last few seconds" out of the workflow. | 23:15 |
| lkcl | remember that the two grants cryptoprimitives and cavatools have *expired* | 23:15 |
| lkcl | we need to get the work in and the RFPs in as promptly as possible, not create more "make-work" distractions | 23:16 |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.57.65> has quit IRC | 23:16 | |
| openpowerbot | [irc] <programmerjake> i'll note what i proposed isn't about optimizing workflow time, but alerting users to prevent massive confusion | 23:18 |
| openpowerbot | [irc] <programmerjake> yeah, i'll look through those grants more soon, probably today | 23:18 |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has joined #libre-soc | 23:21 | |
| lkcl | > alerting users to prevent massive confusion | 23:23 |
| lkcl | that's what documentation and the IRC channel is for. if they're not communicating, it's not our/your problem. | 23:23 |
| lkcl | remember that we had some potential people wanting to help: it took them *eight months* to follow the goddamned instructions | 23:24 |
| lkcl | not one single one of them got on here and said, "hello i am at step X of the instructions: now what do i do?" | 23:24 |
| lkcl | one of them even managed to miss out the chroot part and destroyed their system by running the devscripts that did "apt-get remove systemd" | 23:25 |
| lkcl | we do not need people "helping" if they are in fact just plain stupid and cannot communicate. they will be a nightmare drain on our extreme-limited resources | 23:26 |
| lkcl | focus on *now*, please. | 23:26 |
| lkcl | not "what problems some random as-yet-non-existent *future* and *potential* developer *might* run into if they didn't ask for help and/or didn't follow the build instructions" | 23:27 |
| *** ghostmansd <ghostmansd!~ghostmans@176.59.56.241> has quit IRC | 23:33 | |
| *** ghostmansd[m] <ghostmansd[m]!~ghostmans@176.59.56.241> has joined #libre-soc | 23:54 | |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!