Wednesday, 2023-07-26

*** toshywoshy <toshywoshy!> has quit IRC01:00
*** toshywoshy <toshywoshy!> has joined #libre-soc01:03
programmerjakeis there anything i need to be there for at the meeting tomorrow? my aunt's moving and i've been asked to help01:39
programmerjakelkcl ^01:42
programmerjake(in case you're wondering, the previous moving plans fell through)01:44
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC08:03
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc08:28
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC09:08
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc09:08
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC09:12
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc09:13
lkclprogrammerjake, oh whoops!09:19
lkclmarkos_, fantastic!09:19
lkclsadoon[m], yes bootstrapping from ppc64le is a very fortunate way to break out, usually debian circular build-dependencies are hell to break by new porters, typically requiring months of patient systematic hand-editing of build packages and partial installs09:22
lkclghostmansd[m], wow amazing!09:22
lkclprogrammerjake, the main things we need to go over is the formation of the Libre-SOC Foundation, and then patent and OpenPOWER Foundation "inbound" agreements09:31
lkclwhich we've mentioned maaany times before, so there's not actually anything "new"09:31
lkclsadoon[m], look up the joel shenki microwatt 5.7 linux kernel config.  there is *no need* to reinvent or rediscover the (many) options09:32
lkclso please don't waste time doing that - just use "make menuconfig_microwatt" (or whatever it is)09:33
lkclit's documented in our wiki already the sequence of commands09:34
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC09:49
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc09:49
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC09:53
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc09:53
sadoon[m]<lkcl> "sadoon, look up the joel..." <- That is indeed helpful but unfortunately we've discovered disabling VSX and co. in the kernel doesn't guarantee that userspace can't use them09:58
sadoon[m]Toshaan tested it before09:59
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC10:06
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc10:07
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC10:29
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc10:30
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC10:37
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc10:38
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC11:07
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc11:07
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC12:45
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc12:45
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC13:15
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc13:15
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC13:22
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc13:23
lkclsadoon[m], i mentioned that kconfig so that you don't waste time re-inventing a perfectly-well-tested kconfig that has an associated initramfs/rootfs with busybox *pre-compiled to SFFS level*13:24
lkcl(which was taken down very quickly but i put a copy on
lkcleveryone is (should be) keenly aware that "just because the kernel is compiled to SFFS does not mean that userspace binaries are"13:25
lkcli simply didn't want you wasting precious time on something that's already been solved13:25
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC13:39
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc13:39
programmerjakewell, this is the first time i remember you going over the details of the Libre-SOC Foundation, iirc you mentioned it previously but not really in detail, so I think it's some combination of you discussing it with toshaan a bunch, me missing last wed meeting, and/or me not paying close enough attention in the I think I need to attend today, I'll just have to take a break from helping moving14:36
programmerjakei guess my main concern is if we're assigning patents to the libre-soc foundation, imho they need to be at least licensed back to the inventors such that we can use them in non-Libre-SOC projects if we choose, with sub-licensing to others allowed like GPLv3 or Apache 2.0 require14:46
programmerjakee.g. I'm planning on eventually designing a foss cpu that implements programmable hw translation from x86, risc-v, arm, powerpc, etc. and I want to be able to do that independently of Libre-SOC if i choose14:46
programmerjakeplus it would be nice if i can implement svp64 in the powerpc translation part of that cpu, doing that requires that I obtain a license to whatever Libre-SOC patents are necessary14:50
programmerjakeimho it would be best from a free software perspective if the libre-soc foundation had royalty-free licence grants to any foss project that wants to use them15:00
sadoon[m]<lkcl> "sadoon, i mentioned that kconfig..." <- Oh yeah fully agree there16:54
sadoon[m]Currently doing some qemu/gdb stuff to find out exactly what we need to make sure that code is sffs compliant16:55
sadoon[m]qemu's power6 configuration explicitly disables VSX like programmerjake  mentioned and linked yesterday16:56
sadoon[m]I could very easily build qemu to do that with power9 as well to maintain ISA 3.0 while disabling VSX, it's good progress and we can look into the other extensions after16:56
markos_sadoon[m], yesterday programmerjake pasted a configuration file for qemu for Power profiles17:00
markos_the idea was to duplicate the Power9 profile and just have VSX and stuff disabled17:01
markos_but I do not remember the actual file17:01
sadoon[m]PCR_VSX_DIS         = PPC_BIT(1), /* VSX disable (bit NA since POWER8) */17:24
sadoon[m]From qemu, this might suggest that disabling it will have no effect17:24
sadoon[m]There's also17:24
sadoon[m]PCR_VEC_DIS         = PPC_BIT(0), /* Vec. disable (bit NA since POWER8) */17:24
sadoon[m]If this works, running debian 12's ffmpeg should fail17:27
sadoon[m]If ffmpeg still works, I have no clue :(17:27
sadoon[m]Yep, unforunately ffmpeg still works, VSX is still enabled17:47
programmerjakelooks like it'll work if you don't use kvm and add a new target cpu without the PPC_ALTIVEC flag
sadoon[m]Will try that thanks18:08
sadoon[m]I'm not using kvm anyhow18:08
sadoon[m]Makes it easier to test18:08
sadoon[m]A bit slower to boot but not so bad18:08
programmerjakeactually, do try kvm with the pcr bit disabled, power9 hw apparently does respect that bit (at least according to my understanding of the PowerISA spec.)18:17
sadoon[m]trying now :)18:19
programmerjakehmm, maybe qemu/linux isn't passing the pcr bit through to the hw? idk. adding a new cpu targdt to qemu seems most useful anyway since it works on x86 too18:21
sadoon[m]That's what I'll try next18:21
programmerjakedo make sure you filter out all the other isa support flags libre-soc won't have18:22
sadoon[m]I'll first try just disabling VSX since qemu compilation doesn't take long, baby steps :)18:23
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC18:33
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc18:34
*** octavius <octavius!~igloo@> has joined #libre-soc20:06
octaviusApologies chaps, I'll be late for the meeting.20:06
octaviusIf I will manage to join, I'll be muted in the car20:07
octaviusThe main point is that markos, toshywoshy, sadoon, programmerjake mentioned yesterday is for me to document the FPGA environment setup and programming, and hopefully boot linux kernel on arty a7. Luke managed to do it before on Versa ECP5, so I'll have to do some archive hunting to see what needs to be done. The arty a7 has on board ddr3; not a lot20:09
octaviusbut perhaps enough for just the kernel20:09
sadoon[m]if you have 32-64MB it'll be more than enough20:11
sadoon[m]I remember my arm+FPGA board using 33MB with systemd20:12
*** octavius <octavius!~igloo@> has quit IRC20:12
sadoon[m]I disabled have_vsx and have_altivec universally in qemu's TCG translator, let's see20:12
sadoon[m]Every translation checks those booleans before executing the respective code20:13
*** octavius <octavius!~igloo@> has joined #libre-soc20:16
octaviusThanks sadoon[m]20:16
sadoon[m]no problem :)20:18
sadoon[m]the meeting is at 9PM UTC correct?20:18
octaviusYes, in 40min20:19
sadoon[m]Wait.. 40min?20:19
sadoon[m]That's with daylight savings then?20:20
sadoon[m]9pm UTC and I'm at UTC+3, it's 10:20PM here :)20:20
octaviusOk, it might not be UTC, but it is in 40min20:20
sadoon[m]Better 11PM than 12AM yeah?20:20
octaviusPerhaps it's worth suggesting to move the meeting (unless there are any Australians who plan to attend)20:21
*** octavius <octavius!~igloo@> has quit IRC20:22
sadoon[m]ffmpeg still works.. ugh20:22
sadoon[m]I'm trying something different, trying to get gcc to compile code with VSX/Altivec enabled, a for loop multiplying array values by each other, this should vectorize unless I'm missing something argh20:26
sadoon[m]It could very well be that ffmpeg has built in checks to see if Altivec or VSX are avialable and just use good old software to process the video20:26
sadoon[m]Actually I'm 99% sure it does20:27
programmerjakeif you're trying to vectorize code, altivec supports f32, but not f6420:45
sadoon[m]Scratch all that, I'm using some simple exponent compare functions in VSX20:46
programmerjakei'd use a altivec insn since if that traps, vsx likely will too20:46
programmerjakei may be a few min late for the meeting, i'm just about to my aunt's house20:48
sadoon[m]<programmerja> "i'd use a altivec insn since..." <- Neither did. I'm doing something horribly wrong. Need to dig much further into qemu's code22:03
sadoon[m]And then once we can verify that we do indeed have an sffs compliant OS, I can move onto the task of patching qemu so we can have a proper test solution22:06
sadoon[m]I really wanna get both done ASAP because you guys talking about chacha20 makes me want to pull my old poly1305 verilog code and marry it with that :D22:06
lkcloo! oo!
lkcloctavius: i never managed to get DDR3 working (ever).  i did however manage to get HyperRAM up and running on ls2, and with QTY 2of the larger HyperRAM PMODs that *may* just - barely - be enough to run linux. but only just.22:24
lkclsadoon[m], please please do remember that the budget for this task is a "fixed price".  if you do 4 months of work because you did 3 months of extremely valuable work on qemu, *it's still the same amount of money for you*!22:26
lkclyou really really must keep an eye on what you are planning to do, think through "how long is this going to take", and then think "do i really want to do X amount of work for Y amount of NLnet donations"22:27
lkclif the answer is NO, please please for goodness sake say so and we can put in *another* Grant request to cover what you need (and presumably would love) to do22:27
lkclwe absolutely cannot have you - or anyone - doing stuff where they feel they are being massively undervalued because they did far too much work for far too little money22:28
lkclplus, i know that toshywoshy did manage to reconfigure qemu for the PowerEL-SFFS porting, for NGI POINTER, already :)22:29
lkclanyway - night all, i have yet another meeting tomorrow, near oxford.22:29
sadoon[m]<lkcl> "sadoon, please please do..." <- Most definitely. We were actually discussing that yesterday which is why I only allocated half a day today to trying to mess with qemu to make sure things went smoothly with the gentoo build22:55
sadoon[m]I think testing sffs will need a separate task at this point22:55
sadoon[m]<lkcl> "plus, i know that toshywoshy did..." <- Unfortunately for some weird reason his method is not working with me. Could be the debian qemu package missing that option? Again, something to look into later22:56

Generated by 2.17.1 by Marius Gedminas - find it at!