I just checked in with the cg supe, and I also had a look in exrheader -
they're definitely regular zip1 scanlines, not within tiles either.
Apparently that's an option in Mantra but it's not what they're doing here.

I understand the current limitations with rendering multichannel exr's and
avidly wait for multipart files as well as multithreaded reads. But as I
said we are getting much smaller files (and corresponding speedups in Nuke)
just by re-rendering them.



On 3 May 2013 17:03, Nathan Rusch <[email protected]> wrote:

>   Yeah, Zip1 is just a compression scheme for a block of data. With a
> tiled file, the compression gets applied on a per-tile basis, so each
> scanline in each tile would be compressed separately; for a bucket size of
> 32, that would be zipping 32 pixels together.
>
> Since there is no multithreaded decompression of (non-deep) EXR files in
> Nuke, you would definitely see a performance hit when trying to read the
> sequence in, especially with 36 channels, and it stands to reason that the
> compression ratio would be comparably horrible as well.
>
> -Nathan
>
>
>  *From:* Michael Garrett <[email protected]>
> *Sent:* Friday, May 03, 2013 1:58 PM
> *To:* Nuke user discussion <[email protected]>
> *Subject:* Re: [Nuke-users] EXR file size Mantra vs Nuke re-render
>
> Hey - that was the one thing I thought of but didn't mention - I have so
> little experience with them that I assumed Zip1 always rendered scanline.
> I'll check that out!
>
> On 3 May 2013 16:55, Nathan Rusch <[email protected]> wrote:
>
>>   Sounds like Mantra is generating tiled files.
>>
>> -Nathan
>>
>>
>>  *From:* Michael Garrett <[email protected]>
>> *Sent:* Friday, May 03, 2013 1:37 PM
>> *To:* Nuke user discussion <[email protected]>
>> *Subject:* [Nuke-users] EXR file size Mantra vs Nuke re-render
>>
>>  I just got some multichannel Zip1 compressed 16 bit half EXR files
>> rendered with Mantra which were incredibly slow to read off the network.
>>
>> I re-rendered them like for like in Nuke 7.0v6 to see if there was a
>> performance gain. What I found is that the size of each frame went from
>> ~170MB/frame to ~43MB/frame.  That's definitely rendering all channels
>> (about 36 individual channels), 16-bit half, Zip1 compressed in both source
>> and destination images.
>>
>> I know that Nuke has not yet rolled in the multi part image optimisation
>> that's in EXR 2.0, I'm certainly looking forward to that. But I'd be
>> interested to kow howit could be possible that the file size has shrunk so
>> much in the instance I described above.
>>
>> Cheers,
>> Michael
>> ------------------------------
>> _______________________________________________
>> Nuke-users mailing list
>> [email protected], http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>
>> _______________________________________________
>> Nuke-users mailing list
>> [email protected], http://forums.thefoundry.co.uk/
>> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>>
>
>
> ------------------------------
> _______________________________________________
> Nuke-users mailing list
> [email protected], http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>
>
> _______________________________________________
> Nuke-users mailing list
> [email protected], http://forums.thefoundry.co.uk/
> http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users
>
_______________________________________________
Nuke-users mailing list
[email protected], http://forums.thefoundry.co.uk/
http://support.thefoundry.co.uk/cgi-bin/mailman/listinfo/nuke-users

Reply via email to