08:44 <whitequark> fpgatools are intimidating

08:45 <larsc> you are just not though enough

08:45 <whitequark> a huge undocumented codebase working with a complex undocumented format, and on top of that written in C

09:54 <rjo> lekernel: hey. in migen.genlib.fsm, i think the entering and leaving signals should be combinatorially driven. as it is now, they are true if you have already entered or left a state.

09:55 <rjo> lekernel: or they should be renamed accordingly.

10:55 <wpwrak> whitequark: C means that not only you can use the code :)

10:55 <wpwrak> whitequark: wolfgang may also have some diagrams. maybe ping him to see if he has something he can share ?

10:56 <whitequark> wpwrak: it also means I'm not writing it, unless someone pays me a really big sum of money ;)

10:56 <whitequark> diagrams... maybe.

14:10 <lekernel> rjo, yes; _florent_ is using that semantics ...

14:11 <lekernel> we could use after_entering or after_leaving

14:15 <lekernel> rjo, still need to answer your email too. been busy with the mixxeo (and MBE)

14:17 <lekernel> whitequark, another problem with fpgatools is they are very buggy (and difficult to debug for the reasons you mentioned), even minor changes like moving the LUT by 1 position in the example design will crash the program

14:20 <_florent_> lekernel, rjo: I can easily change my code if you want to use after_entering, after_leaving

14:58 <whitequark> lekernel: I've been thinking of rewriting fpgatools in something more sane and extensively documenting it in the process

14:58 <whitequark> but even though I knew it would be a gigantic task, I severely underestimated it

15:01 <whitequark> plus, I'm definitely not going to pull it alone, and if the only language available for cooperation is C... meh

15:02 <lekernel> cooperation? does Wolfgang answer your email?

15:03 <lekernel> the upper level (Migen to EDIF) should be very reasonable to do, though

15:03 <wpwrak> C is simple, efficient, and a lot of people understand it. it sounds like a rather bad idea to try to move to something else for something like this

15:03 <lekernel> and already useful given the number of bugs that Xst has

15:04 <lekernel> (and amount of bloat)

15:12 <larsc> rewrite it in php

15:13 <wpwrak> no, in Adobe Flash :)

15:17 <larsc> but it must be able to run with gnash ;)

15:19 <larsc> the best is probably an iterative approach, first rewrite it in php, once you are done throw everything away in restart in flash, then next haskel...

15:22 <wpwrak> where's the theoretical purity in this ? start with a turing machine. then adapt for a two-dimensional turing machine. next, generalize for n dimensions. find proof that the result is optimal for 23 dimensions and implement that.

15:23 <larsc> that sounds somewhat stupid

15:23 <larsc> ;)

15:24 <larsc> I mean you want to get things done, dont you?

15:25 <ysionneau> do it in php then use hiphop to turn it into C++

15:25 <ysionneau> then compile C++ to js using emscripten

15:25 <lekernel> as long as a paper can get published, many researchers are happy enough

15:27 <larsc> right, I forgot wpwrak is a phd ;)

15:43 <wpwrak> ;-)

18:34 <rjo> lekernel, _florent_: ack. i'll whip up a patch for {before,after}_{leaving,entering}. given my confusion, i like the explicit but longer names. _florent_, you would have to rename.

22:34 <GitHub75> [migen] sbourdeauducq pushed 1 new commit to master: http://git.io/4s0n7Q

22:34 <GitHub75> migen/master 7e4024b Robert Jordens: genlib/fsm: rename {entering,leaving} to after_{entering,leaving}, add before_{entering,leaving}

22:35 <lekernel> rjo, thx