[weaddercaps]

Chrome came with a brand new implementation using JIT (just-in-time compiling).

So then it'll use more memory and be just as slow the first time you run a website, won't it? Does it only use JIT for methods? How does it handle lambdas?

We just need to kill IE 6-8 quickly.

cold day in hell. The fact that we STILL HAVE IE6 shows that they aren't gonna go away.

[extessarere]

So then it'll use more memory and be just as slow the first time you run a website, won't it? Does it only use JIT for methods? How does it handle lambdas?

Magic

I haven't looked into the specific implementation of it, just the high-level design. It's not just as slow the first time you run it because the conversion itself is highly optimized, and it doesn't just run it literally as it's written. It aggressively optimizes the JS, just as when you compile C++ or something with optimizations turned on. Because it's done intelligently from start to finish, it's wickedly fast.

cold day in hell. The fact that we STILL HAVE IE6 shows that they aren't gonna go away. IE6 is still supported, and therefore in use, because it was bundled with Windows XP so MS has to legally keep supporting it until XP is unsupported.

[kazinopartnerkae]

gcc isn't even that clever. The IBM C/C++ compiler smokes the **** out of gcc in terms of performance. It's not even close. Unfortunately it's PowerPC only. But they do insane amounts of research and innovation in compiler design & optimization. Intel's C/C++ compiler is by far the best for x86...I think Visual Studio is 2nd, then gcc.

gcc is awful because it's lenient with the standards. Though they're trying, slowly, to tighten up. I can't remember all the devs I had *****ing at me thinking the compiler was broken -- "But this C++ code works with gcc!", and they never took the answer "gcc shouldn't. Your code is not standards compliant" very well. *******s.

[EliteFranceska]

Firefox but thinking of switching to Chrome.

[Darnisg]

I actually had a mini argument with the dude interviewing me today for a (mostly) C++ position. It was about the use of virtual destructors (which is asked in every ****ing C++ interview EVER...if you interview for a C++ job and you didn't prepare for this question or know it intuitively, you deserve to not get the job).

I told them the reason why you should have virtual destructors for any class with any virtual members is because if it's not virtual, it's undefined behaviour to the C++ runtime how to proceed with destructing any derived class if it is deleted via a base class object.

He tried to "correct" me that it will simply always leak memory by never calling the destructors of any derived class, if you delete it via a pointer to its base class. I spent a good two minutes arguing this was incorrect (very politely), and in fact it is an undefined behaviour that is simply inconsistent from compiler/runtime to compiler/runtime. But that for all intents and purposes one should assume it would leak memory, and simply always use virtual destructors for any classes with virtual members.

About 30 minutes after the call, I got an email from the guy telling me I was right.

[seekfrieddy]

Why did Ben vote for IE6-8? He really uses all three of them? I'm surprised they don't upgrade when you install IE8. They do upgrade. My laptop is IE 8, but the other computers like this backup box is still IE 6.

[luspikals]

Note to self: Never argue with Asher about browser technology, and never employ him either.

[joanbertis]

Note to self: Never argue with Asher about browser technology, and never employ him either.

[Andrew1978]

I'm probably the only CS geek on the planet who can't ****ing stand Python.

I have to admit, it is INCREDIBLY slow with objects (probably because it type checks CONSTANTLY), mixing tabs and spaces is clinically proven to cause people to GO POSTAL, and it is extremely difficult to test all of your code and edge cases can be a huge ****ing ***** but oh my god it's so simple!

[arerrurrY]

I have to admit, it is INCREDIBLY slow with objects (probably because it type checks CONSTANTLY), mixing tabs and spaces is clinically proven to cause people to GO POSTAL, and it is extremely difficult to test all of your code and edge cases can be a huge ****ing ***** but oh my god it's so simple!

I did go postal because of the spaces/tabs many times. It looks the ****ing same on my screen, but it *****es with cryptic errors. Jesus ****ing christ, a masochist designed the language.

My company launched a python-powered extremely dynamic website (Plone is the CMS), and immediately after putting it in production the site slowed to a CRAWL. It was HIDEOUSLY slow. Patheticly slow. Unbelievably slow.

I had to "fix" the problem by implementing aggressive HTTP proxy caches (Varnish), hand-tuning browser cache settings, implementing memcached (which was a MySQL perf problem more than Python), and basically get the code to run once, cache for hours...and never run again. Because it was so ****ing slow.

Even that didn't fix it. I had to break it up into an Amazon EC2 array with a load balancer. It'd constantly poll the load levels on the server, and once it sustained at a high level it'd automatically spawn a new server (boots, configs, and installs in ~3 mins) and attaches to the load balancer to handle the extra load. During our lunch time peak traffic, the site runs on 12 ****ing servers...it shrinks down to 1 automatically at its lowest traffic (3am).

Anyway...that was a ****ing month of a nightmare, all thanks to Python.

[vNZsk39B]

stuff

You're using python for an enterprise level website. You're doing it wrong.

It is at its core a scripting language, and a damn fine one at that.

[RalfDweflywex]

If only they had let you use Perl.

SP

[wMceqj7F]

The current experiment uses python as the interpreter with most modules written in C++. It is in some ways worse than using cint in ROOT, but it is nice not to have the confusion of having a C interpreter.

JM

[Alliopeti]

I've never even heard of embedded Perl. What a futile idea.

SP

[forexsoft]

Behold the evil that is cint:

http://root.cern.ch/drupal/content/cint

JM

[rasiasertew]

[Zdmlscid]

That is what I wante dto suggestto google years ago.

JM