Comments on: Why expression templates matter ?

By: alpmestan

alpmestan — Fri, 18 Dec 2009 22:32:30 +0000

You can create a matrix class being built combinating many “blocks” (the storage, etc), well something like a policy-based matrix, or policy/traits-based, something like that. But then we rather enter the generative programming world, totally, whereas here it was just about having an AST for a minimal DSEL for matrixes. Very minimal.

But yeah, absolutely nicolas, you’re right.

By: nicolas66

nicolas66 — Fri, 18 Dec 2009 21:10:40 +0000

Nice post. But a problem arise when you decide to do operations on generic types: how to choose the right type between matrix and matrix? Traits can bring an elegant solution to this problem but I don’t know if it’s possible to reduce the number of classes (e.g for symetric cases).

By: Gwenaël Dunand

Gwenaël Dunand — Mon, 23 Nov 2009 20:55:56 +0000

Very nice blog dude.
Keep writing !!

By: uberVU - social comments

uberVU - social comments — Wed, 04 Nov 2009 01:54:15 +0000

Social comments and analytics for this post…

This post was mentioned on Twitter by Alp Mestan: Why expression templates matter ?: http://wp.me/pEB41-W…

By: and

and — Mon, 02 Nov 2009 17:52:44 +0000

Awesome post! I have been using Eigen recently, a linear algebra library, and they stated support for expression templates. I didn’t know what that meant, but I assumed it was along the lines of this. It is nice to see a simple and clear implementation, thanks.

By: Twitter Trackbacks for Why expression templates matter ? « Alp Mestanogullari's Blog [alpmestan.wordpress.com] on Topsy.com

Mon, 02 Nov 2009 13:45:44 +0000

[...] Why expression templates matter ? « Alp Mestanogullari's Blog alpmestan.wordpress.com/2009/11/01/why-expression-templates-matter – view page – cached Imagine we are using a matrix library, with a naive implementation of +, that simply adds column and row-wise each coefficient. Thus, for a, m1, m2, m3 being for example square matrices of order n,… (Read more)Imagine we are using a matrix library, with a naive implementation of +, that simply adds column and row-wise each coefficient. Thus, for a, m1, m2, m3 being for example square matrices of order n, if we want to compute the sum of m1, m2 and m3 and to put the result in a, corresponding to the following code : a = m1 + m2 + m3 (Read less) — From the page [...]

By: alpmestan

alpmestan — Sun, 01 Nov 2009 22:53:42 +0000

Yeah indeed. My code is just a beginning. We should have a temporary object created in the operator=, and then putting this one’s data in this’ data.

I consider writing about expression templates / DSELs for a less common task, in this one I’ll take care of avoiding such “bugs”. By the way, if you have any idea for the “domain” of the DSEL, don’t hesitate. And thanks for reading & commenting !

By: rmn

rmn — Sun, 01 Nov 2009 20:56:35 +0000

First off, nice post – I enjoyed reading.

This will work nicely for the operation you showed, but since you’re not copying the pbjects you’re operating on, if I write something like

a = a*b; // or something else with the lhs in it

I’ll get the wrong results because a is changed in the process.
But then again, as an example for expressions templates, this is a good post.