When will MATLAB 2014b release?

I noticed that in the past few years the "b" versions were released almost all in early September but this year it looks like it's being delayed. Please let us know if anyone has an insider's info.

Thanks Babak

Answer 1

[as hz]

Released

Answer 2

[Image Analyst]

The Mathworks employees are very tight lipped about release dates. I've never heard of them giving specific dates in advance.

Answer 3

[Bingzhi Zhao]

I am wondering about that as well

Answer 4

[Rob Campbell]

This release is unusually late. HG2 update is a big deal, I'd say... http://undocumentedmatlab.com/blog/hg2-update

Answer 5

[Royi Avital]

MATLAB R2014b is out.

I'm so disappointed Mathworks does nothing to improve the language speed.

Moreover, they are still stuck with CUDA instead of moving on to OpenCL for the GPU support.

What are they waiting for?

Julia is much faster and gaining traction and its Toolboxes are improving by day.

Very Disappointing.

[Oleg Komarov]

I disagree. This is a major release with many features other than the hg2. Improved version control integration, datetime(), datastore and mapreduce!

[Royi Avital]

And each one of them works slow, very slow...

They need overhaul of their JIT engine.

Things have evolved a lot on this frontier in the last few years, MATLAB should as well.

[Oleg Komarov]

Julia's benchmarks of Matlab code are based on basic constructs of Matlab code. While this exposes the weaknesses of the Matlab's JIT, it does not take into account that e.g. the sort() function is already provided as compiled and this is true for many applications. Therefore, it is not completely fair to say that the quicksort implementation in Matlab is super slow, because who's gonna use it in that way ever when I can call sort()?

Another point, is that it's run under an older version of Matlab.

An additional point, is that it avoid vectorized code!

Basically, Matlab is not only JIT and forcing a JIT only comparison is NOT fair.

[Royi Avital]

I have asked here once about fast implementation of `im2col`.

If I had Julia, all I needed to do is write it like C and get C like performance.

I can't do that on MATLAB.

Yes, MATLAB have many built in functions which are pretty fast, but what happens when you do something by yourself?

For instant, think of implementing Non Local Means.

Julia only starts its journey.

They haven't implementated vectorization in their compiled code and already faster than MATLAB.

The gap will only get widen in the future unless this becomes priority for Mathworks.

[Oleg Komarov]

Fair point about im2col and all the functions in Matlab code. The Mex API is an alternative, although it's definitely not a seamless paradigm. However, I do think that with increasing size/features in the Julia language, the gap will increase at a decreasing pace.

[Royi Avital]

I don't know.

My wish is the M language will become an open standard.

Mathworks will focus on IDE + Tooblox and someone will build really fast JIT Machines fro that language.

At the long run, it must happen.

For now, I'll take a major improvement in the JIT engine by Mathworks.

They must do it.