Re: A logic code error (possibly?)

Dear Vasilis.

Thank you for confirming a bug and making a correction. However, it is
not clear that the fix is complete -- or at least the work-around is not
sufficient. The error messages are gone, but the output appears to be
unchanged.

Attached is a figure of the wedge region which shows a strip, parallel
to and outside the 'pipe', that is missing. This strip crosses the beam
path of interest. I tried both WHAT(5) = 2 and 1. The repaired code
may be necessary.

I had mentioned in an earlier message that the code could replace an ARB
with planes if it needed to. I did not know that it already did that.
I see no reason to remove the ARB, BOX, or WEDGE bodies from the user.
It is much more reliable by far for the code to convert to planes than
for the user to do so (even allowing for an occasional bug). The user
input is easier to make and to read and understand.

I was stunned at how extensively the code expanded my definition of the
World region. I would love to reduce it. Part of the problem is the
need to extend bodies beyond their actual dimensions so they can be cut
by planes to avoid precision complications. A 'tolerance' parameter to
let the code make small adjustments would be really nice.

Regards,
Joe Comfort

Vasilis Vlachoudis wrote:
> Dear Joseph,
>
> thank you for the bug report. There was indeed an error in the bounding
> box routine for the ARB. Internally the ARB's are converted to planes,
> as the code stores the plane coefficients in the same memory location
> with the vertices, the bounding box information was lost. As result the
> optimizer was canceling the Wedge since it was appearing(wrongly) not to
> have any overlap with the HK1D2a.
> The code is fixed and a patch will be provided with the next respin.
> In the mean time to have it working you can disable the bounding box
> optimization by setting the GEOBEGIN.WHAT(5)=2
>
> However you are heavily using parenthesis in the description of the
> World, which expands to several thousand terms and with bounding box
> optimization off would result in slow execution.
>
> Cheers
> Vasilis
Received on Tue Jun 29 2010 - 23:07:46 CEST