Re: revised quadcc preserve single?

[Nicholas Jankowski]

got it. I'll work up a quick patch for affected functions.

On Mon, Oct 16, 2017 at 6:57 PM, Rik <address@hidden> wrote:

On 10/16/2017 03:17 PM, Nicholas Jankowski wrote:

Ok, thanks. Obviously this function can do whatever it wants since there is not a Matlab implementation, but because we use it as the underlying integrator for those other functions, they will no longer produce compatible output for single inputs. We could easily just add a line to those other functions to recast output as single since there is already an issingle check. Just didn't know if quadcc  behave like the other quads in this regard

Yes, for the moment I think we should just work around quadcc in the m-files by casting to single.  The tolerances will already be correctly reduced so, even though the quadcc function is using doubles, it won't be wasting time calculating unnecessary precision only to have it discarded.

--Rik

On Oct 16, 2017 6:04 PM, "Rik" <address@hidden> wrote:

On 10/16/2017 02:42 PM, Nicholas Jankowski wrote:
> can someone with a compiled version of the new quadcc check whether a
> single type integration limit produces a single output? The one bundled
> with the windows 4.2.1 release does not, which shows up in the recently
> pushed versions of integral2 and 3. A compatibility check against MATLAB
> suggests it should

It does not.  As far as I know, quadcc always performs its operations using
doubles.

--Rik