Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Mar 22 19:17
    raravena80 commented #100
  • Mar 22 19:14
    satabin commented #100
  • Mar 22 19:12
    raravena80 commented #100
  • Mar 04 08:48
    monperrus commented #94
  • Jan 04 21:01
    raravena80 commented #100
  • Jan 04 20:20
    satabin commented #100
  • Jan 04 18:45
    raravena80 opened #100
  • Nov 26 2020 07:00
    monperrus commented #99
  • Nov 25 2020 21:39
    satabin edited #99
  • Nov 25 2020 21:38
    satabin converted_to_draft #99
  • Nov 25 2020 21:38
    satabin labeled #99
  • Nov 25 2020 21:38
    satabin labeled #99
  • Nov 25 2020 21:38
    satabin labeled #99
  • Nov 25 2020 21:38
    satabin labeled #99
  • Nov 25 2020 21:38
    satabin labeled #99
  • Nov 25 2020 21:38
    satabin labeled #99
  • Nov 25 2020 21:38
    satabin opened #99
  • Nov 19 2020 10:11
    Jacarte commented #98
  • Nov 19 2020 10:10
    monperrus commented #98
  • Nov 19 2020 10:10
    monperrus commented #98
Lucas Satabin
@satabin
I'll try again with their clang
and then I'll have to make it run in the CI, and I am already crying about the number of push I'll need for that :sob:
Orestis Floros
@orestisfl
2020-07-14-111244_692x202_scrot.png
Lucas Satabin
@satabin
this looks really good
Lucas Satabin
@satabin
thanks for helping improving swam!
Orestis Floros
@orestisfl
Thank you for writing swam :)
Javier Cabrera
@Jacarte
Hi Lucas
How are you? I hope you are doing good with this covid thing. It has been a while :D
Lucas Satabin
@satabin
HI Javier
doing ok given the circumstances
hope you're doing good as well
Javier Cabrera
@Jacarte
I am ok thanks
Lucas sorry to bother you
but I am having an issue with wasm decoder/encoder binary in SWAM
Lucas Satabin
@satabin
Is that the ticket you opened?
Javier Cabrera
@Jacarte
yep
Lucas Satabin
@satabin
ok, I wanted to take care of this this weekend but I had a crazy one and couldn't work on it
I'll try to take a shot at it tonight
I hope it is not a limitation in scodec (the binary codec library I am using)
if this is it, then I will need some more time to change the current approach
but it looks more like a bug
Javier Cabrera
@Jacarte
It seems it is, but you are the expert :D
Lucas Satabin
@satabin
is the file big?
Javier Cabrera
@Jacarte
1.9 Mb, it is not even that big
Lucas Satabin
@satabin
let's see what I can find
Javier Cabrera
@Jacarte
Thanks
take care !
Lucas Satabin
@satabin
I wanted to change my approach here for a while
this might be an opportunity to push me in doing it for real
Javier Cabrera
@Jacarte
Let me know if I can help :D
I would be glad
Lucas Satabin
@satabin
thanks, I'll let you know
Lucas Satabin
@satabin
Hello there, it's been a while. @Jacarte quick update from my side: I am working on implementing the bulk memory operations proposal. I'd say I'm halfway there and expect it to be implemented by the end of the week
once this is done, I will then do the multi memory one, so that you can safely use memory to store the coverage bitfields/counters/data whatever without inerfering with the standard program memory
Javier Cabrera
@Jacarte
Hi @satabin, thanks !, I am looking forward to this :). Let me know if you need something from me here
zeroexcuses
@zeroexcuses
Are there benchmarks of swam vs wasmer / wasmtime ?
Lucas Satabin
@satabin
Hi @zeroexcuses I didn't perform any benchmark personally but I hink Swam cannot compete with wasmer. Swam is a quite naive bytecode interpreter. I intend in doing such benchmarks some time and identify the big overhead and try to improve these points. Maybe @Jacarte you know or did this kind of benchmarks
zeroexcuses
@zeroexcuses
@satabin : Thank you for the honest & straight forward, not-written-by-PR answer. Much appreciated.
Lucas Satabin
@gnieh:matrix.org
[m]
@zeroexcuses: no problem. Swam is before anything else a project for testing out things around WASM and streaming compilation. I will try to make it more efficient, I have some ideas. However not a lot of time currently :/
Javier Cabrera
@Jacarte
Hi @zeroexcuses and @satabin. Sorry for the late answer :( I agree with @satabin, SWAM cannot compete with wasmer, however, we need some numbers to properly say so, unfortunately, I have not done a benchmark on this. In any case, I would propose PolyBench to measure it, https://github.com/MatthiasJReisinger/PolyBenchC-4.2.1. Cheers
zeroexcuses
@zeroexcuses
@gnieh:matrix.org @Jacarte : I'm fairly new to wasm. Can I ask why the swam code base is (according to cloc) 17k loc? Naively, I expect wasm to have the complexity of form-interpreter + scheme interpreter -- so I'm clearly way off. I'm interested in where this complexity is from.
Lucas Satabin
@gnieh:matrix.org
[m]
A lot of loc are pattern matching for pretty-printing and tree transformations
wasm has a rich set of instructions, so every time we need to do something with them, it results in big pattern matches
plus there is some infrastructure code to try to make it as easy to use (automatic derivation of type classes for instance)
the text parser is also super big
I want to rewrite it at one point, I am pretty sure I can make it more simple
zeroexcuses
@zeroexcuses
Ah, I see, https://webassembly.github.io/spec/core/appendix/index-instructions.html is definitely 10x the # instructions I expected.
Lucas Satabin
@gnieh:matrix.org
[m]
and it's growing ^^
zeroexcuses
@zeroexcuses
wtf, sqrt is a builtin instruction, lol
Lucas Satabin
@gnieh:matrix.org
[m]
wasm is intended to be compiled to lower level things, and specialized instructions are intended to help compiling efficiently to dedicated instructions when available, they have this kind of instructions in wasm, so that you don't loose too much semantics and make efficient compilation easier