Home
Forums
New posts
Search forums
Online Courses
2022 Rankings
2022 MFE Programs Rankings Methodology
Reviews
Latest reviews
Search resources
Tracker
What's new
New posts
New media
New media comments
New resources
New profile posts
Latest activity
Log in
Register
What's new
Search
Search
Search titles only
By:
New posts
Search forums
Menu
Log in
Register
Install the app
Install
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!
Home
Forums
Quant discussion
Computing
Nvidia - Cuda Toolkit for options pricing
JavaScript is disabled. For a better experience, please enable JavaScript in your browser before proceeding.
You are using an out of date browser. It may not display this or other websites correctly.
You should upgrade or use an
alternative browser
.
Reply to thread
Message
<blockquote data-quote="DailyVaR" data-source="post: 16590" data-attributes="member: 527"><p>I feel vindicated today. I found out that IBM is dropping the Cell processor as reported by this Ars Electronica <a href="http://arstechnica.com/hardware/news/2009/11/end-of-the-line-for-ibms-cell.ars" target="_blank">article</a>.</p><p></p><p>One particular point that I brought up appears in the article.</p><p></p><p></p><p>Boys and girl(s)*, its 2010. Do we want to spend the next two years before the end of the world writing cache coherency routines in your analytic code? I heard one IBM developer implemented some code that steals some memory from each SPE to build a coherent cache, but seriously this should be built into the hardware.<span style="color: Silver"></span></p><p><span style="color: Silver"></span></p><p><span style="color: Silver"><span style="font-size: 9px">---------- Post added at 02:04 PM ---------- Previous post was at 01:41 PM ----------</span></span></p><p><span style="color: Silver"></span></p><p><span style="color: Silver"></span></p><p></p><p>Probably never. If you look at every single NVidia CUDA example they implement everything two ways: once for the CPU and once for the GPU. At the end of each example, they compute the run-times so they can compute the speed-up. Also, there is a simple test to see if the GPU result is close to the CPU result within some epsilon. If the numbers were exact, they wouldn't need an epsilon. Just my observation.</p><p></p><p>Also, in all their MC simulation code, they always round up the user specified number of sims to a power of 2 that will occupy all the "units." In CUDA, doing extra work, sometimes, doesn't take anytime at all if you are waiting anyways. For example, you might want to run with 1,000 sims but you will get 1,024. Kinda annoying if you are trying to match MC-based options prices.</p></blockquote><p></p>
[QUOTE="DailyVaR, post: 16590, member: 527"] I feel vindicated today. I found out that IBM is dropping the Cell processor as reported by this Ars Electronica [URL="http://arstechnica.com/hardware/news/2009/11/end-of-the-line-for-ibms-cell.ars"]article[/URL]. One particular point that I brought up appears in the article. Boys and girl(s)*, its 2010. Do we want to spend the next two years before the end of the world writing cache coherency routines in your analytic code? I heard one IBM developer implemented some code that steals some memory from each SPE to build a coherent cache, but seriously this should be built into the hardware.[COLOR="Silver"] [SIZE=1]---------- Post added at 02:04 PM ---------- Previous post was at 01:41 PM ----------[/SIZE] [/COLOR] Probably never. If you look at every single NVidia CUDA example they implement everything two ways: once for the CPU and once for the GPU. At the end of each example, they compute the run-times so they can compute the speed-up. Also, there is a simple test to see if the GPU result is close to the CPU result within some epsilon. If the numbers were exact, they wouldn't need an epsilon. Just my observation. Also, in all their MC simulation code, they always round up the user specified number of sims to a power of 2 that will occupy all the "units." In CUDA, doing extra work, sometimes, doesn't take anytime at all if you are waiting anyways. For example, you might want to run with 1,000 sims but you will get 1,024. Kinda annoying if you are trying to match MC-based options prices. [/QUOTE]
Verification
Post reply
Home
Forums
Quant discussion
Computing
Nvidia - Cuda Toolkit for options pricing
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
Accept
Learn more…
Top