[Bf-cycles] Cycles tables

[Mohamed Sakr]

Hi Sergey,

I'm having a nightmare with this bug, so I need your help.
what happens in the first rendering:
kernel_globals is filled correctly inside device update, path trace is
running fine "scene is empty, no meshes, no lights, no objects, just params
are passed".

in the second rendering:
kernel_globals is filled correctly inside device update, path trace
kernel_globals got all data = NULL... , any idea how this happens "and they
are just filled before path trace"?

cheers,
Mohamed Sakr

On Wed, Apr 13, 2016 at 11:01 AM, Mohamed Sakr <3dsakr at gmail.com> wrote:

> Hi Sergey,
>
> it fails to sample Sobol sampler.
>
> On Wed, Apr 13, 2016 at 8:54 AM, Sergey Sharybin <sergey.vfx at gmail.com>
> wrote:
>
>> It depends on what texture it's failing to sample.
>>
>> Most likely you either not calling `device_update()` somewhere or you
>> don't call kernel_tex_copy().
>>
>> On Wed, Apr 13, 2016 at 4:57 AM, Mohamed Sakr <3dsakr at gmail.com> wrote:
>>
>>> Hi,
>>>
>>> I'm implementing an exporter similar to Cycles standalone.
>>> got a little problem,
>>> first render call runs fine "simply an empty scene"
>>> second render call (should be with a non-empty scene" fails, it always
>>> crashes inside tables "many times inside sobol table, once inside film
>>> offset"
>>>
>>> template<typename T> struct texture  {
>>> ccl_always_inline T fetch(int index)
>>> {
>>> kernel_assert(index >= 0 && index < width);
>>> return data[index];
>>> }
>>>
>>> #ifdef __KERNEL_SSE2__
>>> ccl_always_inline ssef fetch_ssef(int index)
>>> {
>>> kernel_assert(index >= 0 && index < width);
>>> return ((ssef*)data)[index];
>>> }
>>>
>>> ccl_always_inline ssei fetch_ssei(int index)
>>> {
>>> kernel_assert(index >= 0 && index < width);
>>> return ((ssei*)data)[index];
>>> }
>>> #endif
>>>
>>> T *data;
>>> int width;
>>> };
>>>
>>> data is nullptr, sometimes width is wrong, etc..
>>> what does reset these values? "or where/when are they
>>> deleted/initialized, with scene or session"
>>>
>>> thanks.
>>> cheers,
>>> Mohamed Sakr
>>>
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> https://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>>
>>
>>
>> --
>> With best regards, Sergey Sharybin
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> https://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20160414/3254b0bd/attachment.htm