Sunday, 2023-02-19

*** gnucode <gnucode!~gnucode@user/jab> has quit IRC03:12
lkclsadoon[m]: :)03:25
lkclhey if you implemented 914 instructions you'd have noo problem running standard ppc64 gentoo releases.03:26
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC05:43
*** Knar <Knar!~Knar@> has quit IRC06:34
*** tplaten <tplaten!~tplaten@> has joined #libre-soc07:22
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc07:26
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC07:51
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC07:58
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc08:32
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc08:59
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC09:10
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc09:15
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC10:06
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc10:06
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC10:20
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc10:21
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC11:07
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc11:10
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC11:42
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc12:10
*** mx08_ is now known as mx0812:13
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC12:27
sadoon[m]Finding libre-soc basically made that project idea of mine irrelevant12:28
*** tplaten <tplaten!~tplaten@> has quit IRC12:31
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC12:50
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc13:56
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc13:56
*** ghostmansd <ghostmansd!> has joined #libre-soc14:20
*** ghostmansd <ghostmansd!> has quit IRC14:41
*** tplaten <tplaten!~tplaten@> has joined #libre-soc15:20
*** tplaten <tplaten!~tplaten@> has quit IRC16:05
*** tplaten <tplaten!~tplaten@> has joined #libre-soc16:17
*** tplaten <tplaten!~tplaten@> has quit IRC16:26
*** tplaten <tplaten!~tplaten@> has joined #libre-soc16:31
*** tplaten <tplaten!~tplaten@> has joined #libre-soc16:44
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC16:51
*** tplaten <tplaten!~tplaten@> has quit IRC17:17
*** tplaten <tplaten!~tplaten@> has joined #libre-soc17:36
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc17:36
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC18:24
markosprogrammerjake, I'm all for adding rust support, but I'd first spend some time defining the intrinsics scheme for SVP6418:56
markosand C definitely takes precedence over Rust18:56
markoshaving said that, I could not possibly tell you not to do it, it's just that it wouldn't be my first priority18:57
markosmy point is that imho, adding compiler support depends on defining C intrinsics, unless you want to depend on auto-vectorization for everything -I wouldn't19:09
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has joined #libre-soc19:41
*** ghostmansd[m] <ghostmansd[m]!~ghostmans@> has quit IRC20:02
programmerjakesince that task doesn't have enough funding to be complete, my plan was to only add simple fixed-length and variable length vectors support, and rely completely on translation from the ISA-independent vector support20:34
programmerjakealso, cranelift doesn't currently have a huge amount of optimizations, iirc it doesn't have autovectorization at all20:35
programmerjakeit's designed to compile quickly, not as much to produce the fastest code20:36
programmerjake(though it can generate very fast code when the input is already optimized)20:36
programmerjake^ basically how wasm is supposed to work20:37
markosso it doesn't depend on intrinsics being defined?20:37
programmerjakeno, not for the amount i'm planning on completing for that grant20:38
programmerjakei'm visiting family today, so ttyl20:39
markospossibly no objections from me then, my only fear was/is adding support for "temporary" rust intrinsics without having proper C intrinsics defined in gcc/llvm20:40
programmerjakeportable-simd relies entirely on ISA-independent intrinsics, so no issues there either20:41
programmerjake(though we are temporarily using x86 intrinsics in portable-simd for dynamic swizzle until rustc and llvm gain support)20:42
markosgo for it then :)20:43
markoswould also be a good way to find possible problems in the process20:44
markoswe probably wouldn't be able to borrow code to gcc/llvm, but we might use some of the ideas20:45
programmerjakethe main technical problem needing solving is register allocation of register ranges, basically everything else is pretty straightforward20:45
programmerjakethat's part of why i'm already working on it as part of bigint-presentation-code.git20:46
markoswe really should also start doing a high level design of the intrinsics, I'd hate to fall in the same trap as the Risc-V people, where they had to drop and redesign their whole C intrinsics support from scratch, pissing everyone off in the process20:47
programmerjakegcc would also need non-power-of-2 vector support added, it currently assumes vectors are a power-of-2 size20:48
markosyes, that's going to meet some resistance, unless it's only a soft-limitation -ie it doesn't break everything if you try to use non-power-of-2 vector sizes20:49
markosI haven't looked at it in detail, no idea how hard-coded it is20:49
programmerjakeit breaks dynamic swizzle's semantics, gcc defines out-of-bounds indexes to wrap, requiring remainder operations for non-power-of-220:50
programmerjakeprobably many other places assume power-of-220:51
programmerjakeany way, gtg, my brother's picking me up in a few min20:51
markossure, have fun20:51
*** ghostmansd[m] <ghostmansd[m]!> has joined #libre-soc21:01
markoslkcl, trying to understand your vertical first algorithm for chacha20, would appreciate it if we did some call one of the next days to understand it better, currently am in the process of integrating it in an asm version of xchacha_hchacha2022:01
markosI get the big picture, but want to make sure I understand the exact role of svindex/svshape2 in the beginning22:05
*** gnucode <gnucode!~gnucode@user/jab> has joined #libre-soc22:21
*** markos <markos!~Konstanti@> has quit IRC22:22
*** markos <markos!> has joined #libre-soc22:24
*** ghostmansd[m] <ghostmansd[m]!> has quit IRC22:41
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has quit IRC23:09
*** lxo <lxo!~lxo@gateway/tor-sasl/lxo> has joined #libre-soc23:23

Generated by 2.17.1 by Marius Gedminas - find it at!