(
Jump to the original submission)
Fri 12 Oct 2012 07:34:11 AM UTC, comment #6:
I like your version, so I pushed it :
http://hg.savannah.gnu.org/hgweb/octave/rev/f0dfdc7faa71
I am closing this bug.
|
Thu 11 Oct 2012 08:04:27 PM UTC, comment #5:
Carlo, take a look at the attached changeset (eps.cset). I followed your lead, but I believe I simplified things by not creating a second copy of the input matrix. This routine isn't a high performance one so I also used normal indexing, x(i), rather than changing over to a Fortran vector, C pointers, and x[i] notation.
(file #26742)
|
Thu 11 Oct 2012 05:22:04 PM UTC, comment #4:
Rik, is it OK then if I push?
|
Thu 11 Oct 2012 04:04:07 PM UTC, comment #3:
Ignore the last comment. I see that the Matlab behavior has already been verified on the maintainers list.
|
Thu 11 Oct 2012 03:35:17 PM UTC, comment #2:
Could someone with Matlab verify what they do? The documentation doesn't seem to contemplate that eps can be called with a matrix as an argument. We could, of course, choose to extend it in this logical way.
|
Thu 11 Oct 2012 07:27:45 AM UTC, comment #1:
The attached changeset should fix the problem.
c.
(file #26737)
|
Wed 10 Oct 2012 06:11:21 PM UTC, original submission:
eps(z) returns eps(z(1)) for matrix argument z; it should return a matrix the size of z with the (i,j) element = eps(z(i,j))
|
(Note: upload size limit is set to 16384 kB, after insertion of the required
escape characters.)
Attach File(s):
Comment:
Depends on the following items: None found
Items that depend on this one: None found
Follow 7 latest changes.