- September 29th, 2014, 12:22 pm
- Forum: Careers Forum
- Topic: Tier 2 Model Validation - Good place to start?
- Replies:
**26** - Views:
**7361**

<t>QuoteOn the other hand, would it impair my future career, considering I will be paid less, would have to work my up to not just FO, but also to a tier 1 bank, and then maybe algo trading? Dont bank on this career plan mv->fo derivs->algo. Its extremely unlikely to happen, and it will become more ...

- September 18th, 2014, 2:29 pm
- Forum: Programming and Software Forum
- Topic: Gaining Exposure to Production Quant C++ Library?
- Replies:
**54** - Views:
**7848**

<t>Lol. I wasnt suggesting that he doesnt learn new c++ features. I was just saying that for example using 'auto' variables in the interview might make the interviewer think that he has recently learnt c++. The usual assumption is that few years of c++ experience is not enought to be good at it so t...

- September 18th, 2014, 8:25 am
- Forum: Programming and Software Forum
- Topic: Gaining Exposure to Production Quant C++ Library?
- Replies:
**54** - Views:
**7848**

<t>QuoteOriginally posted by: MaxCohenSo for a C++ quant developer ignoring the math. What is the standard job specs then?In my experience within banks only speaking of quant devs (developers who are part of the quant team):In derivatives area, typically C++ (to a high level), Excel/VBA, sometimes C...

- August 13th, 2014, 11:32 am
- Forum: Technical Forum
- Topic: Volatility Arbitrage Intraday
- Replies:
**8** - Views:
**4836**

<t>From a theoretical point of view, I cant see how 1 or 2 would be a good strategy to put on without effective gamma hedging. I'm assuming you know the theory behind replicating a realised vol play so I wont go into that. To answer your questions on the arb, consider the following two scenarios:- s...

- August 13th, 2014, 7:48 am
- Forum: Programming and Software Forum
- Topic: Tridiagonal solver parallel algorithm
- Replies:
**19** - Views:
**5347**

<t>One question for the pde buffs...Lets say a generic diffusion equation: dU(x,t)/dt=beta(x,t) * d2U/dx2, so that the diffusion constant beta is not constant. To ensure stability (convergence) in explicit FD scheme, I use the max{beta(x,t), for all x,t} to calculate the number of time-steps require...

- August 12th, 2014, 4:21 pm
- Forum: Programming and Software Forum
- Topic: Tridiagonal solver parallel algorithm
- Replies:
**19** - Views:
**5347**

<t>QuoteOriginally posted by: CuchulainnQuoteThe trickier problem is in knowing a priori which numerical systems are well-conditioned enough to be solved with floats, not doubles.The techniques for this are documented but it is lost in action I suppose because double is so pervasive. Why float??My g...

- August 12th, 2014, 4:16 pm
- Forum: Technical Forum
- Topic: Volatility Arbitrage Intraday
- Replies:
**8** - Views:
**4836**

<t>In 1 and 2 you've got increasing short gamma position which is very difficult to manage towards expiry. I think naked short gamma positions have been losing money in last 6 months atleast. I would be more inclined to build strategies on 3 (rolling tenor every week or so) so that you have more sta...

- August 12th, 2014, 2:34 pm
- Forum: Programming and Software Forum
- Topic: Tridiagonal solver parallel algorithm
- Replies:
**19** - Views:
**5347**

<t>Thats an interesting article. I'll try that approach of setting a binary convenient dx term rather than 0.01.After reading further about it, it looks like the intermediate double calculations are due to the compiler storing variables and constants in registers. GPUs typically have thousands of re...

- August 12th, 2014, 10:08 am
- Forum: Programming and Software Forum
- Topic: Tridiagonal solver parallel algorithm
- Replies:
**19** - Views:
**5347**

<t>Thanks Cuch. I'll try that after the project as I dont have much time to change the code too much at this time. It'll be interesting to see the performance difference. I spent a lot of time yesterday debugging numerical differences in local vol and heston calculations on gpu v cpu! All my variabl...

- August 8th, 2014, 4:01 pm
- Forum: Programming and Software Forum
- Topic: Tridiagonal solver parallel algorithm
- Replies:
**19** - Views:
**5347**

Thanks Alan Yes thats the matrix system I was trying to solve. Thats very interesting method, I didnt know about this method. I will try it in matlab. For academic purpose, it will be very easy to parallelise it aswell.

- August 8th, 2014, 3:46 pm
- Forum: General Forum
- Topic: Interview question help
- Replies:
**5** - Views:
**4577**

<t>You can think of a N year swap with annual resets. If you parallel bump the libor or whatever else curve by 1bp to get dv01, each yearly payoff will be 1bp higher i.e. 1bp*5mio=$500 each. Now discount each payoff by the appropriate disc factor. So in simple terms, the dvo1 of N year swap is aroun...

- August 8th, 2014, 10:25 am
- Forum: Programming and Software Forum
- Topic: Tridiagonal solver parallel algorithm
- Replies:
**19** - Views:
**5347**

<r>Thanks for replying guys.Costeanu - that was a good link but notations are difficult to follow. It looks similar to reduction algorithms but on LU decomposed matrix. outrun - N-number of diagonal elements ~ Number of serial steps. You're right, my statement doesnt make sense. If I do some variant...

- August 4th, 2014, 8:29 am
- Forum: Programming and Software Forum
- Topic: Tridiagonal solver parallel algorithm
- Replies:
**19** - Views:
**5347**

<t>Hi guysI'm writing code for my uni project. I understand tridiagonal solvers are inherently serial in nature and best parallel algorithms offer O(logN) complexity in theory. However, using cuSparse library I only start to see around 1.3x speedup when the diagonal size is around 10k elements. In t...

- July 24th, 2014, 3:16 pm
- Forum: Careers Forum
- Topic: Internal vs External Job Application
- Replies:
**56** - Views:
**9881**

<t>QuoteOriginally posted by: SprinterQuoteOriginally posted by: ashkarI dont advise anyone to go on a bankwide job search ... try in all teams trading, structuring, dev, quant, middle office etc etc anything will do approach (like you).I only think about applying to jobs which makes sense given my ...

- July 24th, 2014, 1:49 pm
- Forum: Careers Forum
- Topic: Internal vs External Job Application
- Replies:
**56** - Views:
**9881**

<t>I dont advise anyone to go on a bankwide job search ... try in all teams trading, structuring, dev, quant, middle office etc etc anything will do approach (like you).I only think about applying to jobs which makes sense given my experience and network which means 2-3 teams in the bank. Anyway it ...

GZIP: On