CineonInput is not a public class, and it only exposes certain lower-level
building blocks (read_native_things) that are used with the public read_things
with some extra intervention in many cases (including data conversion). You
might be able to make it work (except for the fact that it's not exposed in any
public headers, but I wouldn't recommend it.
I think our best bet is still to try to figure out why it's crashing in the
first place. Catching it in a debugger to narrow down to a specific line (or at
least function) would be immensely helpful.
-- lg
> On Apr 23, 2019, at 2:17 PM, Mathieu Prevot <[email protected]> wrote:
>
> Hello.
>
> I'm having DCI 4k 10bits images (video), and I'm looking to read one or many
> of these files as fast as possible (I'm using a cache for fast re-access,
> reading those).
>
> I can read with ` ImageInput::open(filename);` in a debug build (with a good
> computer, NVMe 970 evo plus, 9900k, 32GB, 2080ti), but it take 4 seconds to
> open the image and convert it to a uint8 uint16 or float32 image; and working
> with a release build (with -O2 or more) fails (cf. "openfile memory
> exception" thread).
>
> I was thinking about using directly `class CineonInput`. Any thought about
> this ?
>
> My goal is to load those images and have a quick and accurate display or use
> (for processing) in Directx12 (or cuda), if not in WPF (only 8bits).
>
> I could convert images to uncompressed 16bits tiff files, but I'd rather not.
>
> Mathieu
> _______________________________________________
> Oiio-dev mailing list
> [email protected]
> http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
--
Larry Gritz
[email protected]
_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
