• C++ Programming for Financial Engineering
    Highly recommended by thousands of MFE students. Covers essential C++ topics with applications to financial engineering. Learn more Join!
    Python for Finance with Intro to Data Science
    Gain practical understanding of Python to read, understand, and write professional Python code for your first day on the job. Learn more Join!
    An Intuition-Based Options Primer for FE
    Ideal for entry level positions interviews and graduate studies, specializing in options trading arbitrage and options valuation models. Learn more Join!

Why not Matlab instead of C++?

Joined
10/20/09
Messages
32
Points
16
I'm just starting to learn Matlab, and I'm coming to wonder, why isn't this used instead of C++ in MFE programs? Since it has some of the finance toolboxes, it seems like it would be a great choice (now I don't actually know anything of these tool boxes, so I could be totally wrong), but everyone uses C++. Why?
 
In the words of Jim Johnson:

"Speed, speed, can someone say speed?"
Michael Strahan: "SPEED!"
"Speed! 65 yards for the touchdown!"

(talking about Dolphins vs. Saints...of course the Saints won that one...well, what can I say...hey Drew, in terms of keeping your undefeated streak up...break a leg!)
 
Speed in one thing but that it's not necessary the main reason. Luck is something else since a lot of the systems has been developed in C++ so there is expertise already.

Search QN, there are a some discussions on this topic.
 
Showing my naiveness, is speed really that big of a concern? Do these companies not have enough money to spend on high performance computers, or do the calculations take too much from computer resources?
 
Computer hardware are so cheap these days, it's not even a big factor like years ago. The consideration is much more on the implementation and optimization of the code and how much you can do with the language of choice.
Two links just off the top of the google search I found
MATLAB - Choosing Hardware
AFJ's PhD: C++ versus MatLab

Like it said, you can simply price a call option in Matlab by typing blsprice but you have no control over how it is implemented. On the other hand, you can implement your BS pricing in C++ using a few options.

A tool for each purpose.
 
Showing my naiveness, is speed really that big of a concern? Do these companies not have enough money to spend on high performance computers, or do the calculations take too much from computer resources?

You can never have too much speed. Plus, the difference is on orders of magnitude. MatLab is about as fast as R. C++ is fast enough to blow a hole in something.
 
MatLab is about as fast as R.
This is misleading. One real problem with Matlab is that it is a proprietary product. It means that it will be as fast as the optimization used by the company who sells it.

Since R is open source, you can compile it using any compiler, library and any level of optimization. We found that using the dedicated Math Libraries to recompile R is worth the effort. We recompiled R using ATLAS and the matrix operations were sped up considerably.
 
Jacket: MATLAB on the GPU

For those interested in accelerating MATLAB code with minimal effort, I invite you to check out GPU-accelerated MATLAB which we released earlier this year at AccelerEyes. We offer Jacket which provides a really simple mechanism for automatically porting standard MATLAB code to the GPU. Using the GPU, you can "blow a hole" in even C++, directly in the MATLAB environment, for any computation involving matrices of size 32x32 or larger.

Anyway, it may be useful to this thread... check it out here: http://www.accelereyes.com
 
This looks impressive. The main question is, how much is it?
 
Didn't I see you suggest a Tesla GPU for someone? You must have known of some equivalent of this product (unless this is impressive over its competitors), does this "blow a hole in C++" or what?
 
The impressive part is the way to translate Matlab to CUDA code, not the performance of the GPU. We already use GPUs for calculations but we write low level CUDA code.
 
Maybe ppl wants to control during compile time and runtime, lol. I remember my coworker once said, the reason he didn't use Java was the garbage collector would get him lazy. :)
 
Jacket does look like making MATLAB programming for CUDA very easy.

Yeah, melonakos, how much is the single computer license? I do not want to go through the registration process, and why pricing is a secret. Any specials for students, like MATLAB has?
 
These specialty pricing libraries are so competitive/secretive they don't put pricing online. Ever.
You have to give your contact info away in order to ask about it. The next step would be for their sale manager to bug you and/or your manager to death. I had a few product inquiries with several firms before at my last job and they all wanted to send the whole sales team to meet with my manager who can write the check, not me the one who decides which software to purchase.
It's the same deal for companies that sale to Wall Street firms.
Try to see if you can find any pricing on Numerix, Quantifi website.
 
Jacket Pricing Link

Hello everyone,

Thanks for the feedback. The following is a direct link for pricing on Jacket:

http://www.accelereyes.com/doc/AccelerEyesProductsAndPrices.pdf


It's priced based on customer type - just like standard addon MATLAB packages.

Let me know if there is anything we at AccelerEyes can do to be useful to you. If I can personally help, you can reach me by email to john.melonakos a t accelereyes.com

BTW, free, fully-functional trials are available on our website too.

Best,

John
 
If you have vector- and matrix-based calculations, MatLab will be first choice.
Nevertheless if you have loops (if, while, for, etc.) in your code, MatLab will be very slow.
So for a mathematician it is very visceral to use MatLab, but MatLab is not really a programming language.
 
I swear I searched the accelereyes's website high and low but I did not see this pricing link. Andy's email probably invoked their conscience ;)

$100 for student license - I am in. melonakos, are there any limitations of student license/product, e.g. size of the matrix as MATLAB has?

Thank you.
 
Details of Jacket trial version

Hi const451,
There are absolutely no functional limitations of the trial version of Jacket, apart from the 15-day limitation. Hope this helps.

--
Product Manager
AccelerEyes LLC
 
Hi vishy, I was asking about the student version, are there any limitations in student version?

Thanks!
 
Back
Top