|
From: | Paul F. Dietz |
Subject: | Re: [Gcl-devel] generic function performance |
Date: | Sat, 04 Jan 2003 08:56:36 -0600 |
User-agent: | Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2) Gecko/20021126 |
Peter Wood wrote:
Hi In another post, I showed how we can make some functions (like typep, subtypep and coerce) be generic, using #'pcl::make-specializable. It is really a big advantage to be able to use this technique since we can automatically reuse the present code, with small modifications, and build some fancy-ansi stuff on top. However, I suggested that it might have negative performance implications. Well I've been doing some informal testing, and I'm not sure that the performance impact is very significant.
I most often see TYPEP in form where the type is a constant form. TYPEP should specialize that at compile time (perhaps using a compiler macro?). Anyway, GCL's performance is (IMO) secondary at this point vs. correctness and spec compliance. Make it right, then make it fast. Paul
[Prev in Thread] | Current Thread | [Next in Thread] |