-
Notifications
You must be signed in to change notification settings - Fork 0
2012 09 20
dybber edited this page Sep 20, 2012
·
2 revisions
-
ICFP
-
Access to GPU-machine
-
Proposed schedule
-
21. September: installation of:
- Feldspar
- Repa 3
- Accelerate
- DPH
- Obsidian
- Nikola
- Required OpenCL/CUDA toolkits etc.
-
27. September (inden næste møde): implementation of Binomial pricer:
- Sequential (Unboxed Vectors)
- i Feldspar
- i Nikola
- i Accelerate
- (i Repa - gratis)
- (i DPH - gratis) Lookup a hand-optimised OpenCL/CUDA implementation.
-
4. October: implementation of Longstaff and Schwartz:
- Sequential
- i Feldspar
- i DPH
- i Repa
-
11. October: implementation of Longstaff and Schwartz:
- På Nikola
- På Accelerate
-
18. October: deadline for survey (including written report section)
-
1. November: Decision deadline (how to proceed) and scheduling of tasks for november/december
-
November/December: Programming
-
Januar/Februar: finishing, tuning and writing the report
-
1. march 2012 Hand-in
-
-
Finance and understanding of methods
- Geoffrey Mainland welcomes examples and collaboration on Nikola
- Feldspar authors would also like examples
- Ken has sent a link to the american-options github-repository to Manuel Chakravarty, and they say that the newest Github-version of Accelerate should handle the example much better.
- Comments on the schedule:
- As we already have many of Binomial pricers implemented, one week is probably too much for that. Though, we also have to set up a good environment for performing benchmarks, understand the math etc.
- Rolf has a sequential C++ implementation of the Binomial pricer we can look at.
- We should also look at the possibility of parallelism at the level of portfolios, and make pricers that simultaneously could price e.g.: American, European, Asian, Bermuda in parallel
- Ken likes the 18. October deadline for the survey
- We should discuss what the time between 18. October and 1. November should be used for