> Clearly we're spending way much time in the GC with the vxml variant.
Thanks a lot! This is a nice tutorial for statprof.
What you have found out, was already my guess but I had no proof. The
VXML version does not utilize the fact that the nodes are already
destructured. During each visit of a
On Wed, 28 Dec 2022, Sascha Ziemann wrote:
Here are the profling informations:
--8<---cut here---start->8---
scheme@(guile-user)> ,profile (vxml:traverse-pre (sxml->vxml xcb)
remove-whitespace) #:loop 1
% cumulative self
time seconds seconds p
i have no idea, at the opposite, in one of my program i experienced a
vector implementation faster than the list one.
Regards,
Damien
On Wed, Dec 28, 2022 at 2:23 PM Sascha Ziemann wrote:
> I spent my Christmas vacation playing with XML and tree traversal. And
> I have observed some strange ti
On Wed, Dec 28, 2022 at 02:22:16PM +0100, Sascha Ziemann wrote:
> I spent my Christmas vacation playing with XML and tree traversal. And
> I have observed some strange timings, while measuring the execution
> time of different algorithms. My test file is the XCB XML schema:
[vector version slightl
