[Rcpp-devel] Presentation of Rcpp in Meielisalp

[Dirk Eddelbuettel]

On 29 June 2011 at 15:35, Dirk Eddelbuettel wrote:
| 
| On 29 June 2011 at 22:02, Tama Ma wrote:
| | Dear Dirk and Romain,
| | 
| | 	I will be giving a talk on the power of the Rcpp wrapper tomorrow morning in the Meielisalp financial engineering workshop organized by Dielthem.
| | 
| | 	Hereby, I attach the link of my presentation slides (in pdf), ie. 
| | 		http://www.phys.ethz.ch/~pingnang/numerical-option-pricing-with-Rcpp.pdf 
| | 
| |         Could I have your opinion on the slides, or is there something in it which you strongly do not allow me to speak on?
| 
| Well, this arrived just minutes ago, and I really only glanced at it. So
| really quickly and as an unedited stream of consciousness:
| 
| 
| i)    Your talk, your responsibility. You didn't need to ask here what we think.
| 
| ii)   The visual style consistently applies the worst elements of Powerpoint
|       and stresses those. What a horror.  But hey, it's your talk and it is in
|       front of a bunch of geeks.  But I'd learn Latex Beamer if I were
|       you. And try to exercise some restraint.
| 
| iii)  You mention the standard API / inline / Rcpp progression but you fail to
|       use Rcpp + templated headers via inline!  You _never_ have to leave the
|       R prompt in that case.  That is the one thing you could consider
|       changing. Inline and Rcpp rock. The list archives has examples of Rcpp
|       and templated headers for MCMC, what you do here is exactly the same.
| 
| iv)   Great example of Rcpp modules use with the first option pricer. Nicely
|       done, and a good motivation for why I should switch RQuantLib to use
|       more 
| 
| v)    I got lost in what the 2nd or 3rd prices did anew. Your slides are too
|       'noisy', and its late here (and I happen to be on the same timezone as
|       you this week -- and wish I could have come to the Alp!)
| 
| vi)   OpenMP came to gcc/g++ earlier than 4.4.*, I think it was 4.2.*
| 
| vii)  You don't talk much about Rcpp and a glue between R and C++. I think you
|       should.
| 
| viii) Below my (almost 10 year old) picture you have 'RCpp'. We typeset it
|       'Rcpp'.  
| 
| ix)   Good luck, and let us know what the audience says. I'd espcially
|       interested in comments by Brian Ripley who has helped us a lot in build
|       issues at different points but who otherwise seems to be somewhat
|       critical of Rcpp.
| 
| x)    There is no tenth point.

xi)     I fully agree with your code reuse, STL/Boost, OO recommendations.

xii)    Some of the STL example are overdone. You do not need Boost lambda to
        scale a vector by a constant.

xiii)   Boost accumulators is wonderful for what it does, but I would not call
        it a 'statistics library'.

Cheers, Dirk
 
| | 	Thanks for any advice in advance. have a nice day!
| 
| Likewise. Good luck,  Dirk
| 
| | 
| | 	Thank you very much.
| | 
| | Best regards,
| | Tama Ma 
| | (たまま / 馬秉能）
| | 
| | pingnang at phys.ethz.ch
| | www.phys.ethz.ch/~pingnang
| | 
| | Contact Number: 
| | +41 44 633 76 55  (Office)
| | +41 78 640 50 15  (Mobile)
| | +41 44 633 11 15  (Fax)
| | 
| | HIT K 31.3
| | Institut für Theoretische Physik
| | Wolfgang-Pauli-Strasse 27
| | ETH Hönggerberg
| | 8093 Zürich
| | Switzerland
| | 
| | _______________________________________________
| | Rcpp-devel mailing list
| | Rcpp-devel at lists.r-forge.r-project.org
| | https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
| 
| -- 
| Gauss once played himself in a zero-sum game and won $50.
|                       -- #11 at http://www.gaussfacts.com
| _______________________________________________
| Rcpp-devel mailing list
| Rcpp-devel at lists.r-forge.r-project.org
| https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel

-- 
Gauss once played himself in a zero-sum game and won $50.
                      -- #11 at http://www.gaussfacts.com