[Sidefx-houdini-list] Problems with Merge's performances

Jerry jerry at axyzfx.com
Mon Dec 1 09:48:09 EST 2008


I looked into that with Oodini's file, and even with the same attributes 
on each input
the problem persisted.
> I wonder if both inputs have the same number of attributes present? If 
> not the Merge SOP may be initializing attributes on some of the input 
> geometry. That may be marking the second input dirty and forcing it to 
> cook. If this is the case try initializing the missing attributes 
> yourself to see if that fixes the problem. Either that or remove the 
> needless attributes that result from the Particle System. Just make sure 
> the two inputs have the same attributes present. Why the first input 
> doesn't cook is probably the first input being optimized.
>
>
> The best approach is to make sure the same attributes exist on all 
> inputs in to the Merge SOP then see if the upstream SOPs continue to 
> cook needlessly.
> The second best approach if the first isn't feasible (data memory sizes, 
> speed hits, etc.) is to not use the Merge SOP but to extract the one 
> network leaf to a new object and be done with it. It's a trivial step 
> with the shelf these days.
>
>
>
> -jeff
>
> Oodini wrote:
>   
>> Jerry a écrit :
>>   
>>     
>>> I don't see any reason why this would be the case.
>>> I had a look at your scene and checked the performance monitor output. 
>>> There is definitely a difference in what is being
>>> cooked depending on the order of the merge inputs.
>>>     
>>>       
>> Thanks for confirmation. I sent a bug report to SESI.
>> _______________________________________________
>> Sidefx-houdini-list mailing list
>> Sidefx-houdini-list at sidefx.com
>> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>>   
>>     
>
> _______________________________________________
> Sidefx-houdini-list mailing list
> Sidefx-houdini-list at sidefx.com
> https://lists.sidefx.com:443/mailman/listinfo/sidefx-houdini-list
>
>
>   





More information about the Sidefx-houdini-list mailing list