IRC channel logs

<matrix_bridge><gtker> oriansj: Why did you end up creating the "// CONSTANT BLA 1" syntax instead of just treating "#DEFINE BLA 1" as a constant?

<ekaitz>janneke: sent you a documentation patch for Mes, if it needs some extra work in the docs to accomodate that kind of explanation

<ekaitz>I also plan to add documentation about the Mes interpreter internals, once it's reworked

<ekaitz>for the moment i tried to be the less intrusive as possible with the rest of the document

<matrix_bridge><Andrius Štikonas> gtker I guess early versions of cc_x86 had # as a comment and it was processed very early (in tokenizer)

<janneke>ekaitz: thanks

<ekaitz>stikonas: interesting how my if->switch change makes gcc -O0 20% faster but it's unoticeable in M2Planet

<stikonas>hmm, I guess gcc can optimize things better

<stikonas>still I guess it's not bad that gcc build is faster

<stikonas>and switch should be easier toread

<stikonas>than lots of if/else if /else blocks

<ekaitz>yeah of course

<ekaitz>structurally speaking the changes I did are only in the garbage collector

<ekaitz>so it makes sense that the impact is kind of small in both cases anyway

<ekaitz>stikonas: in M2-Planet I can rely on checking the first field of structs vs integers?

<stikonas>what do you mean?

<ekaitz>ugh superhard to explain

<stikonas>any code example somewhere?

<ekaitz>let me write something

<stikonas>I guess that's why I'm confused

<ekaitz>yeah.. haha

<ekaitz>ugh i don't remember how it was

<ekaitz>i know it was written somewhere in a book, let me find

<ekaitz>okay i think I got something

<ekaitz>the proper question would be: can I cast pointers in M2Planet?

<ekaitz>are the struct fields in order and all that?

<stikonas>there is no casting in M2-Planet

<stikonas>struct fields are in order though

<stikonas>casting, e.g. (void*) int_pointer is not allowed yet

<stikonas>but you could assign them I think

<stikonas>e.g. void_pointer = int_pointer;

<ekaitz>oh

<ekaitz>good

<ekaitz>yeah this is enough i think

<ekaitz>so i can reassing a pointer to a struct to an int pointer and it will get the first field of the struct (if it's an int)

<stikonas>I think this should work

<stikonas>but best to test it

<ekaitz>okay great

<stikonas>well, if something doesn't work, feel free to report

<stikonas>we fixed that switch issue that you had...

<ekaitz>:)

<ekaitz>uugh i thought this thing was going to be easier but Mes compares things by pointer

<ekaitz>very interesting, i will need to rewrite many things :)

<stikonas>I'm not too familiar with mes... I mostly looked at meslibc and mescc

<ekaitz>it's a very simple interpreter, pretty short, but in a couple of places it gets interesting

<ekaitz>and now M2-Planet supports pointer arithmetic, right?

<stikonas>ekaitz: partiallty

<mihi>ekaitz: gcc will compile switch to a jump table even with -O0: https://godbolt.org/z/6McTcasTo I'm pretty sure M2-Planet does not.

<stikonas>yeah, looks like mes/src/*.c has similar size as M2-Planet these days

<ekaitz>mihi: oh! yeah that must be a huge improvement in performance

<ekaitz>stikonas: yes! it's very small

<ekaitz>the complexity really is in mescc i think

<stikonas>and mescc also needs nyacc

<stikonas>ekaitz: as for pointer arithmetic, it is supported in += -= ++ and --

<stikonas>but not for + and -

<stikonas>++ and -- are unreleased though, not part of 1.12

<stikonas>so if you can, port pointer arithmetic cases to += or -=

<ekaitz>okay

<stikonas>form the current #if __M2__ blocks

<stikonas>I think there aren't that many of those

<stikonas>there is include/mes/mes.h:#define M2_CELL_SIZE sizeof (struct scm)

<stikonas>I guess e.g. something like cell_zero = cell_arena + M2_CELL_SIZE; could now be written cell_zero = cell_arena; cell_zero += 1;

<stikonas> cell_zero = cell_arena + 1 wouldn't work yet

<ekaitz>okay

<ekaitz>good

IRC channel logs

2025-03-25.log