41CL Quiz: Determinant of 30x30 anti-Identity matrix

[J-F Garnier]

Thanks for your detailed explanations about the determinant adjustment on the 48G series.

So I changed my mind, I recognize that it is not just a trick to artificially round the determinant results in integer matrix problems, but is generally an improvement of the accuracy, in a way similar to the much older 2^3 accuracy fix mentioned by Valentin.

You quoted Paul McClellan, do you have any references to articles or forum posts where he may have described more on the subject?

Quote:I've always wondered if the algorithm as I described above could deliver a wrongly rounded result, in which case I would have to agree with you.

Yes, and the fact that the adjustment can be disabled with flag -54 may suggest that, in some situations, it may be better not to adjust the determinant.

J-F