Fork me on GitHub
#cljs-dev
<
2016-06-10
>
rohit07:06:24

@danielcompton: thank you! i’ve decided for the time being to copy over all the benchmarks from cljs repo and later focussing on the quality of the results themselves.

rohit07:06:49

do you expect the code size to have changed over time significantly?

danielcompton07:06:04

Generated code size that is

danielcompton07:06:29

Not really sure, but it's good to keep an eye on it. There have been regressions in the past

rohit07:06:13

@danielcompton: oh cool. i think it should be possible to add that.

dnolen12:06:50

@danielcompton: no, tickets should only report how to reproduce with ClojureScript only

rohit12:06:18

I’ve just updated my benchmarks to include all the benchmarks from cljs repo. I had to run the benchmarks through simple benchmarking code to achieve it though. https://ducky427.github.io/clojurescript-benchmarks

dnolen12:06:24

just include the minimal source to repro and describe the problem - don’t link to anything

rohit12:06:11

it can take 10-15 seconds to draw all the charts

dnolen12:06:27

@jr the protocol should not be a var - you should create an example that always shows it to be the case

dnolen12:06:37

if it’s not always the case I want to see an example of how it can happen

mfikes14:06:11

@rohit that looks really cool! If possible, perhaps a different color choice for JSC and SM. Nearly impossible to see the differences. (Red/green color blind.)

rohit14:06:16

@mfikes: definitely possible. I’ll change those colours.

mfikes14:06:20

@rohit: I think there are online tools you can use to simulate the various forms of color blindness and see if the choices look differentiable.

rohit14:06:40

@mfikes: awesome. let me research that.

mfikes14:06:00

(Sorry for the superficial feedback, but that’s the first thing I noticed. Will dig deeper into the actual content.)

rohit14:06:32

@mfikes: no no. all feedback is welcome! 🙂

rohit14:06:07

@mfikes: next week i may run these benchmarks through benchmark.js. those take a really long time to run. ~10sec per benchmark and I have about 150 benchmarks in there. so about 30min per suite. I am benchmarking against 14 versions.

dnolen14:06:24

I didn’t realize the reversed param order was part of the original implementation

bhauman14:06:36

yeah thats what I thought

dnolen14:06:38

don’t want to change that - which one of these patches actually apply then?

bhauman14:06:50

let me check

dnolen14:06:04

@bhauman: you may need to rebase master due to other changes

mfikes14:06:12

@rohit Perhaps if there is an actual server to run the benchmarks, they can take a long time to run without any real consequence.

bhauman14:06:34

@dnolen: the first one, but I can rebase

rohit14:06:42

@mfikes: thats my plan. now that you’ve given me jsc built for linux, i can do that.

rohit15:06:23

@mfikes: I’ve changed the colours to something which hopefully won’t be an issue.

mfikes15:06:04

@rohit Yes. Much better 🙂

rohit15:06:40

@mfikes: colorbrewer to the rescue! 😄

richiardiandrea15:06:53

wow a lot of graphs great!

dnolen18:06:59

@bhauman: just bumping about that patch if you want it in a release - cutting one today

bhauman18:06:21

@dnolen: cool thanks, on it

bhauman18:06:03

@dnolen: the first patch by date, the last patch in the list applies cleanly

bhauman18:06:21

and works 🙂

dnolen18:06:33

@bhauman: thanks applied

dnolen19:06:49

@jr if you want a fix for your Var issue in the release going to need more feedback as requested above

dnolen19:06:04

this is the last ticket in my queue

dnolen19:06:14

my suspicion is that Schema adds some :protocol meta, if this is not a symbol we now throw it away

jr19:06:26

👍 trying it now

jr19:06:15

dnolen: unfortunately that still errors with the same exception. I’m still trying to put together reproducible repo, but have been told not to utilize my work time on it 😞

jr19:06:48

I’ll get you all the information you need for the next release

jr19:06:02

er next next release

dnolen19:06:50

@jr k thanks - it’s obviously sanitization issue

dnolen19:06:05

it’s preferable that garbage just not get written out

dnolen19:06:12

if we read back a var it’s meaningless

anmonteiro21:06:17

I’m running into an error with :simple compilation that doesn’t occur with any other type of optimizations

anmonteiro21:06:39

has anything like this ever been reported?

anmonteiro21:06:05

the error is $s$$.substring is not a function and seems related to cljs.pprint

dnolen21:06:09

@anmonteiro: maybe there’s been something like that in the past yes

anmonteiro21:06:16

from what I can gather from the stacktrace

dnolen21:06:34

cljs.pprint has never had any serious testing around production usage

dnolen21:06:53

if you’re shipping cljs.pprint than you’re treading into waters where very little time has been spent

anmonteiro21:06:56

weird thing is it works in :advanced, so it’s good to go!

anmonteiro21:06:26

It’s devcards that uses pprint

anmonteiro21:06:44

I’m trying to compile some cards with optimizations, and ran into this thing

anmonteiro21:06:50

I’ll use :advanced anyways