Re: [lmi] Calculation summary speed

[Greg Chicares]

On 2006-10-23 14:02 UTC, Greg Chicares wrote:
> 
> as a test, here are some rough timings (in milliseconds):
> 
>           total  calculate prepare format
> old:        175        150       0     25 no xslt: all C++
> new:        950        150     300    500 current cvs
> modified:   430        150      30    250 cvs + patch below

Here are some precise timings--the average of five runs:

           total  calculate prepare format
old':        915        158     301    456 before latest change
new':        549        154     299     96 current cvs

Thus, Evgeniy's changes committed Oct 27 15:52:19 2006 UTC
yield a tremendous improvement on my machine, in the area
that most needed more speed.

If the "extremely coarse sketch" snipped from the quoted email
contains a viable idea, then we have the big win that I was
hoping for. That "sketch" reduced "prepare" time enormously.
If we combine the best parts of "modified" and "new'" above:

           total  calculate prepare format
modified:    430        150      30    250 "coarse sketch"
new':        549        154     299     96 xsl optimization
the winner:  278        152      30     96

then we get 278 milliseconds: quadruple the speed of the
initial implementation, and only a tenth of a second slower
than what's in production.

I think the facility to add supplemental-report columns isn't
fully working as intended yet, so I'm taking this latest set
of changes as a proof-of-concept for xslt optimization, not
a final release.