Gah, I'm late!

Anyway, this is my hackathon project:

http://digger.k3.1azy.net/trend/

Succinctly, it is the lovechild of Digger and Mozilla's areweslimyet.com.

It measures stats about D built from D's entire GitHub history, as well as those of programs built with said D versions. Currently only two programs are tested (empty program and "hello world"), so please send PRs for meaningful benchmarks. Adding new programs is very easy: http://j.mp/1I7ELEc.

There is a bunch of cool things happening under the hood about which I might or might not do a full blog post later. Currently it's still collecting data from recently added benchmarks, so coverage is spotty for some tests - should be fleshed out in a few days.

Enjoy!

On Sunday, 3 May 2015 at 13:04:41 UTC, Vladimir Panteleev wrote:
> Gah, I'm late!
>
> Anyway, this is my hackathon project:
>
> http://digger.k3.1azy.net/trend/
>
> Succinctly, it is the lovechild of Digger and Mozilla's areweslimyet.com.
>
> It measures stats about D built from D's entire GitHub history, as well as those of programs built with said D versions. Currently only two programs are tested (empty program and "hello world"), so please send PRs for meaningful benchmarks. Adding new programs is very easy: http://j.mp/1I7ELEc.
>
> There is a bunch of cool things happening under the hood about which I might or might not do a full blog post later. Currently it's still collecting data from recently added benchmarks, so coverage is spotty for some tests - should be fleshed out in a few days.
>
> Enjoy!

I got a parsererror for data/data.json, "Unrecognized token '?'"

On Sunday, 3 May 2015 at 13:20:03 UTC, cym13 wrote:
> I got a parsererror for data/data.json, "Unrecognized token '?'"

Please try Chrome or Firefox.

On Sunday, 3 May 2015 at 13:04:41 UTC, Vladimir Panteleev wrote:
> Gah, I'm late!
>
> Anyway, this is my hackathon project:
>
> http://digger.k3.1azy.net/trend/
>
> Succinctly, it is the lovechild of Digger and Mozilla's areweslimyet.com.
>
> It measures stats about D built from D's entire GitHub history, as well as those of programs built with said D versions. Currently only two programs are tested (empty program and "hello world"), so please send PRs for meaningful benchmarks. Adding new programs is very easy: http://j.mp/1I7ELEc.
>
> There is a bunch of cool things happening under the hood about which I might or might not do a full blog post later. Currently it's still collecting data from recently added benchmarks, so coverage is spotty for some tests - should be fleshed out in a few days.
>
> Enjoy!

Cool project, can't wait to see benchmarks/compile times for larger programs get tracked.

On Sunday, 3 May 2015 at 13:04:41 UTC, Vladimir Panteleev wrote:
> It measures stats about D built from D's entire GitHub history, as well as those of programs built with said D versions. Currently only two programs are tested (empty program and "hello world"), so please send PRs for meaningful benchmarks. Adding new programs is very easy: http://j.mp/1I7ELEc.

You can find a few self-contained benchmarks at https://github.com/D-Programming-Dash. This is the compiler performance CI project I started a while ago, but unfortunately couldn't quite finish yet due to life/academia getting in the way. My focus for this project was a bit different, though. I primarily collected real-world test cases instead of trying to focus on the most "meaningful" ones, and there is quite a number of them. The history tracking display mode is not written yet, and the UI lacks a lot of tooltips, etc.

(In case anybody is curious what the current state looks like: http://dash.klickverbot.at/dash1/compare/gdc_main:release_boundschecked@current..ldc_master:release_boundschecked@current. It seems that the system hit a GitHub API rate limit from which it failed to recover some time in January, though, so there are no recent results. Also, the server currently runs in debug mode and I didn't put any effort into front-end optimizations yet, so expect it to be slow.)

 — David

On 5/3/15 6:04 AM, Vladimir Panteleev wrote:
> Gah, I'm late!
>
> Anyway, this is my hackathon project:
>
> http://digger.k3.1azy.net/trend/
>
> Succinctly, it is the lovechild of Digger and Mozilla's areweslimyet.com.
>
> It measures stats about D built from D's entire GitHub history, as well
> as those of programs built with said D versions. Currently only two
> programs are tested (empty program and "hello world"), so please send
> PRs for meaningful benchmarks. Adding new programs is very easy:
> http://j.mp/1I7ELEc.
>
> There is a bunch of cool things happening under the hood about which I
> might or might not do a full blog post later. Currently it's still
> collecting data from recently added benchmarks, so coverage is spotty
> for some tests - should be fleshed out in a few days.
>
> Enjoy!

This is awesome! Any chance to get the guilty commit more precisely? For example I see a large jump recently, but any of 3 dozens commits (or a combination thereof) could have caused it.

Andrei

On 5/3/15 12:04 PM, Andrei Alexandrescu wrote:
> This is awesome! Any chance to get the guilty commit more precisely? For
> example I see a large jump recently, but any of 3 dozens commits (or a
> combination thereof) could have caused it.

Oh, it looks like the zoom feature already does that. Apparently a jump from 492KB to 1,379KB was caused by this? https://github.com/D-Programming-Language/tools/pull/130

Andrei

On Sunday, 3 May 2015 at 19:04:20 UTC, Andrei Alexandrescu wrote:
> This is awesome! Any chance to get the guilty commit more precisely? For example I see a large jump recently, but any of 3 dozens commits (or a combination thereof) could have caused it.

Keep zooming in until you start seeing individual commits. If the lines begin to disappear, there is as of yet insufficient data, check back in a few hours.

I wrote a little help page here:

https://github.com/CyberShadow/TrenD/wiki

On Sunday, 3 May 2015 at 19:06:50 UTC, Andrei Alexandrescu wrote:
> On 5/3/15 12:04 PM, Andrei Alexandrescu wrote:
>> This is awesome! Any chance to get the guilty commit more precisely? For
>> example I see a large jump recently, but any of 3 dozens commits (or a
>> combination thereof) could have caused it.
>
> Oh, it looks like the zoom feature already does that. Apparently a jump from 492KB to 1,379KB was caused by this? https://github.com/D-Programming-Language/tools/pull/130

Not sure how you got that. The program-hello-binarysize spike from May 2014 was caused by https://github.com/D-Programming-Language/dmd/pull/2561 .

On Sunday, 3 May 2015 at 13:04:41 UTC, Vladimir Panteleev wrote:
> There is a bunch of cool things happening under the hood about which I might or might not do a full blog post later.

http://blog.thecybershadow.net/2015/05/05/is-d-slim-yet/