-
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!
CUDA Certification
- Thread starter Ryu Hoshi
- Start date
DominiConnor
Quant Headhunter
- Joined
- 9/6/06
- Messages
- 1,051
- Points
- 93
I worry about any certification that you can pass by reading a programming guide and some PDFs.
Daniel Duffy
C++ author, trainer
- Joined
- 10/4/07
- Messages
- 10,162
- Points
- 648
I agree with Dominic.
To write code in CUDA you need to have a real deep knowledge of C and multithreading.
Exams prove one thing: you can do exams. It does not mean you can program and certainly not multi-threading.
Many developers have been scorched by believing that MT was easy. They went back to single-threaded programming. And speedup is not a given. The end result can be very disappointing for all involved.
It's a rare skill as far as I can see and even more rare if it is combined with finance knowledge.
IMO it's best to go off and program the stuff; I believe Mark Joshi has some CUDA code ..
To write code in CUDA you need to have a real deep knowledge of C and multithreading.
Exams prove one thing: you can do exams. It does not mean you can program and certainly not multi-threading.
Many developers have been scorched by believing that MT was easy. They went back to single-threaded programming. And speedup is not a given. The end result can be very disappointing for all involved.
It's a rare skill as far as I can see and even more rare if it is combined with finance knowledge.
IMO it's best to go off and program the stuff; I believe Mark Joshi has some CUDA code ..
- Joined
- 6/23/09
- Messages
- 244
- Points
- 28
I attended a GPU seminar at Baruch in the Spring put on by the CUNY HPC lab , which had sessions with some top guys at NVIDIA and a consulting firm. It was pretty cool.
They recommended this 2010 book:
Amazon.com: Programming Massively Parallel Processors: A Hands-on Approach (9780123814722): David B. Kirk, Wen-mei W. Hwu: Gateway
It it seems to give some good context outside of just the CUDA manuals. I wonder if anyone else has impressions of it?
They recommended this 2010 book:
Amazon.com: Programming Massively Parallel Processors: A Hands-on Approach (9780123814722): David B. Kirk, Wen-mei W. Hwu: Gateway
It it seems to give some good context outside of just the CUDA manuals. I wonder if anyone else has impressions of it?
DominiConnor
Quant Headhunter
- Joined
- 9/6/06
- Messages
- 1,051
- Points
- 93
Cuch makes a good point, it is difficult, but that is in some ways is a good thing.
There may be a seriously good market in GPU tech centred around CUDA with banking experience. It may be quite soon.
I make the unintuitive claim that it being hard is actually good because to earn a good living with a skill, it has to have some barrier to entry, and the best is that only you and the right sized set of people have the talent to do it.
Occasionally the barrier is due to law protecting a profession or the cost of gaining the skill. That's less good because the law can be changed, and occasionally technology drops the cost hard.
Of course it must not be too hard, else the lack of skilled workers to do it will deter people from using it.
You need the barrier, else as soon as the price goes up, people rush to learn it, drive the price down, and since they have invested in this skill and will hang around, the price will keep on going down for a while.
In my my view this is the market for Java skills.
To be more lucrative you need people with money who want this work done, and who will make money from your activity, and since people are competitive, they will pay more for an edge than more boring cost saving activities that offer equivalent returns.
Also there is a subtle point, not covered well in the literature, which is that employers must feel that the skill 'deserves' higher pay, typically because they see it as more difficult.
It is tempting to bring Java in again since it's virtue is ease of use, but a better example is Excel VBA. Most employers sees it as trivial so won't pay good money, then spend hours bitching about the hours they've lost because their spreadsheets don't work.
So CUDA has good barriers, and employers with deep pockets.
There may be a seriously good market in GPU tech centred around CUDA with banking experience. It may be quite soon.
I make the unintuitive claim that it being hard is actually good because to earn a good living with a skill, it has to have some barrier to entry, and the best is that only you and the right sized set of people have the talent to do it.
Occasionally the barrier is due to law protecting a profession or the cost of gaining the skill. That's less good because the law can be changed, and occasionally technology drops the cost hard.
Of course it must not be too hard, else the lack of skilled workers to do it will deter people from using it.
You need the barrier, else as soon as the price goes up, people rush to learn it, drive the price down, and since they have invested in this skill and will hang around, the price will keep on going down for a while.
In my my view this is the market for Java skills.
To be more lucrative you need people with money who want this work done, and who will make money from your activity, and since people are competitive, they will pay more for an edge than more boring cost saving activities that offer equivalent returns.
Also there is a subtle point, not covered well in the literature, which is that employers must feel that the skill 'deserves' higher pay, typically because they see it as more difficult.
It is tempting to bring Java in again since it's virtue is ease of use, but a better example is Excel VBA. Most employers sees it as trivial so won't pay good money, then spend hours bitching about the hours they've lost because their spreadsheets don't work.
So CUDA has good barriers, and employers with deep pockets.
- Joined
- 7/9/10
- Messages
- 141
- Points
- 28
It is already happening. Indian IT vendors like Infosys and HCL are developing expertise in CUDA. People take CUDA classes in India and get jobs in these companies.
http://www.nvidia.co.in/page/cuda_consultants.html
Infosys is working on the following
Converting existing applications to CUDA to
enable GP GPU
• Options Pricing, Risk Modeling, Algorithmic
Trading, Intra-day Risk management, Firm-wide
Risk Management, Options pricing using Monte
Carlo (MC) simulations, etc.
• Seismic – Oil & Gas
• Aerodynamics and aero acoustics
• Numerical Analysis
• Scientific Computing (e.g. computational
electromagnetic, fluid dynamics)
• Healthcare & Life Sciences
• Bioinformatics/ work bench – gene
pattern matching
DominiConnor
Quant Headhunter
- Joined
- 9/6/06
- Messages
- 1,051
- Points
- 93
Thanks for the link Marina, and although I didn't know that, it surprise me not at all.
There is not an Indian market for skills, or an American one, Britsh one or (say it quietl)y European one.
There are skills, and people who want to hire those skills, and given the low price entry point for a usable GPU rig there is no barrier to entry there.
So of course we will see Indian s/w firms doing this, and they will see the same effect, in that getting people who can do CUDA to an acceptable standard is harder than getting Java people.
To make it clear, I'm not saying that CUDA is a good way to defend your earning power against Indians, I'm saying that it is a good way of defending it against other programmers regardless of where they happen to be. That is as useful for an Indian to know as a Brit or American.
There is not an Indian market for skills, or an American one, Britsh one or (say it quietl)y European one.
There are skills, and people who want to hire those skills, and given the low price entry point for a usable GPU rig there is no barrier to entry there.
So of course we will see Indian s/w firms doing this, and they will see the same effect, in that getting people who can do CUDA to an acceptable standard is harder than getting Java people.
To make it clear, I'm not saying that CUDA is a good way to defend your earning power against Indians, I'm saying that it is a good way of defending it against other programmers regardless of where they happen to be. That is as useful for an Indian to know as a Brit or American.
