Nathan what is the block size?  The entire width?    I don' t know for some
reason I was under the impression that the array is little blocks of scan
lines.  Or is it that each block is the entire X ?

Scan Lines

For scan line blocks, the line offset table is a sequence of scan line
offsets, with one offset per scan line block.

In the table, scan line offsets are ordered according to increasing scan
line y coordinates.

Randy S. Little
http://www.rslittle.com/
http://www.imdb.com/name/nm2325729/




On Mon, Mar 31, 2014 at 5:15 PM, Nathan Rusch <[email protected]>wrote:

>   When using ZIPS with scanline-based EXRs, each scanline is compressed
> into its own block within the file, so the library has to read and
> decompress the whole line, even if only part of the resulting data is
> actually used by Nuke.
>
> -Nathan
>
>
>  *From:* Randy Little <[email protected]>
> *Sent:* Monday, March 31, 2014 2:08 PM
> *To:* Nuke user discussion <[email protected]>
> *Subject:* Re: [Nuke-users] Faster EXR format for Nuke read.
>
>  I don't believe that it has to read the whole line.  Other wise changing
> just the width of the ROI wouldn't change redraw speed.   I can't remember
> but I thought it was discussed before about that very thing.
>
>  Randy S. Little
> http://www.rslittle.com/
> http://www.imdb.com/name/nm2325729/
>
>
>
>
> On Mon, Mar 31, 2014 at 4:52 PM, Dorian Fevrier <
> [email protected]> wrote:
>
>> Hi Deke, thanks for feedbacks!
>>
>> For now, nothing is write in the stone, that's why I ask for the faster
>> way to read EXRs.
>>
>> Files are located on server.
>> Nuke version is 8.0v3
>> About the number of layer/channels, it depend on your answer. :) I guess
>> we will have one file per layer and per AOV, so the end files will only
>> have rgba without any sub layers/channels.
>> What is the problem with backward compatibility? I didn't get it. Files
>> write with OpenEXR 1.7 will be slower to open in Nuke 8x?
>>
>> I copy paste the problem I see with Zip1 (From my answer to Jeremy):
>>
>>
>> - when you zoom on a part of the image you are working, you need to load
>> and uncompress the whole line of every file you use to generate it.
>> - when you unzoom and should only display 1 pixel for 2 or 2 pixel every
>> 3 (horizontally I mean), it need to load the whole line while only 2 pixels
>> of 3 are needed.
>>
>> And I add: Tell me if I'm wrong but Nuke seems to work with cache more
>> granularly with NO_COMPRESSION EXRs than any compression method because
>> cache seems to be generated on the fly only for visible areas while Zip1
>> make it need to reload the whole line when another area is needed.
>>
>> Regards,
>>
>> Dorian
>>
>> Le 03/31/2014 03:12 PM, Deke Kincaid a écrit :
>>
>> So this depends on a few factors.  A few questions first.  Where are the
>> files located(local raid, local disk, or server)? Which version of nuke?
>> How many layers/channels are you writing?  If you are using Nuke 8 can the
>> file lack backward compatibility (multipart channels written for exr 2.0).
>>
>> The reason I ask is on a network Zip1 compressed tend to be best and a
>> good mix of compression and speed.  Locally on a fast fusion io then
>> uncompressed is actually faster though they take up a lot more space.  If
>> you use lots of channels in a single file then writing the layers as
>> separate parts are suggested or if the file is all mattes then write each
>> channel as a separate part.
>>
>>  --
>>  Deke Kincaid
>> Creative Specialist
>> The Foundry
>> Skype: dekekincaid
>> Tel: (310) 399 4555 <%28310%29%20399%204555> - Mobile: (310) 883 4313
>> Web: www.thefoundry.co.uk
>> Email: [email protected]
>>
>>
>> On Mon, Mar 31, 2014 at 10:23 AM, Dorian Fevrier <
>> [email protected]> wrote:
>>
>>>  Hi (again for some of you!) Nuke community!
>>>
>>> I try to find the faster EXR formats for the exr Nuke reader in a case
>>> of a full CG compositing and it's quite hard to get valid informations.
>>>
>>> Could someone confirm me this:
>>>
>>> One channel per file (RGBA)
>>> Compression: Imf::NO_COMPRESSION
>>> Order: Imf::INCREASING_Y (I've never been sure about this)
>>> A dataWindow well defined.
>>>
>>> Have I miss something?
>>>
>>> Big thanks in advance and Hello again!
>>>
>>> Regards,
>>>
>>> Dorian
>>>
>>> _______________________________________________
>>> 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 [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