15:37 <rick42> beneroth: hi! o/ creepy fb at it again, eh?

15:37 <rick42> hey pil folks

15:37 <Regenaxer> Hi rick42!

15:37 <rick42> hi Regenaxer! :)

15:38 <rick42> Regenaxer: so, llvm-based dev making good progress?

15:39 <Regenaxer> Yep

15:39 <Regenaxer> Takes time though

15:39 <beneroth> yo rick42 o/

15:39 <rick42> Regenaxer: understood

15:40 <rick42> good to hear!

16:01 <Regenaxer> It is a big task

16:01 <Regenaxer> Most time I spend to first learn the implementation language

16:05 <rick42> Come on, Regenaxer, you are a genius. This is child's play for you. Admit it! :) just kidding, I get it. You have to learn something that someone else wrote. To get really good at it takes time, as you "crawl into the mind" of the people who developed it for insight. And then after that you have to apply that to the task at hand whith is non-trivial.

16:05 <rick42> which*

16:06 <Regenaxer> No, I have to learn something *I* wrote

16:06 <Regenaxer> But it is a new lang

16:06 <Regenaxer> Have to learn to think in it

16:06 <rick42> ... then you have to crawl into your own mind ... from years ago :)

16:06 <rick42> T

16:08 <rick42> Nothing that time and beer can't solve. :)

16:14 <Regenaxer> Yes, but not years ago. Last month

16:14 <Regenaxer> and the lang is still changing

16:15 <Regenaxer> It *looks* like Pil, but is a subset, with different behavior

16:15 <Regenaxer> It maps to LLVM-IR, so I have to keep in mind bith

16:21 <Regenaxer> OK, to explain it better

16:21 <Regenaxer> look at this: http://ix.io/1UWx

16:21 <Regenaxer> The same code fragment in three versions

16:22 <Regenaxer> The Java version in Ersatz is 95% the C version in pil32

16:26 <beneroth> good example :)

16:37 <rick42> i take it pil21 is new version. i see what you mean. thanks

16:38 <Regenaxer> :)

16:44 <Regenaxer> The 'loop' function in that example is a typical case

16:44 <Regenaxer> It looks kind of similar, but has an optional 'T' to indicate that it returns a value

16:45 <Regenaxer> and it does not support (T ...) or (NIL ...)

16:45 <Regenaxer> but (? ...) for exit

16:45 <Regenaxer> Different is also that (? ...) exits also from inside nested expressions

16:45 <beneroth> sounds logical, but also confusing

16:46 <Regenaxer> unlike T/NIL which must be on top level

16:46 <Regenaxer> indeed

16:46 <Regenaxer> (cond T ...) is similar

16:46 <Regenaxer> ie also without T

16:47 <Regenaxer> (if ...) always returns something

16:47 <Regenaxer> and (when ...) and (unless ...) never

16:47 <Regenaxer> Same for while and until

16:47 <Regenaxer> Thats confusing

16:48 <Regenaxer> but it is critical whether it returns something

16:48 <Regenaxer> because *if* it returns a value, it must be the same type for all branches

16:48 <Regenaxer> LLVM is a strongly typed system

16:48 <Regenaxer> tedious!!

16:49 <Regenaxer> pilAsm was the complete opposite

16:49 <Regenaxer> total freedom! :)

17:11 <rick42> :) freeeeeeeeedom!

17:11 <Regenaxer> aye

17:23 <beneroth> [OT] Intel Skylake and related CPUs have a "unpredictable behaviour" bug in the conditional jump instruction (JCC)

17:23 <beneroth> https://www.phoronix.com/scan.php?page=article&item=intel-jcc-microcode&num=1

17:54 <Regenaxer> uh

17:55 <Regenaxer> "jumps spanning 32-byte boundaries" are very frequent

18:01 <beneroth> well luckily conditional jumps are not one of the most important building blocks *sarcasm*

18:01 <beneroth> Intel has a huge QA problems for years now

18:01 <Regenaxer> T