[Bf-cycles] OpenCL on AMD GPUs

Lenny Wang lenny at multicorewareinc.com
Mon Jan 12 06:20:22 CET 2015


The original OpenCL megakernel seems working fine with the latest 14.12
omega driver on AMD GCN GPUs.  Tried on some popular models available on
web.

Could someone point me to the repo that holds scenes for testing?

On Sun, Jan 11, 2015 at 3:20 PM, Sergey Sharybin <sergey.vfx at gmail.com>
wrote:

> Hi,
>
> No, not really. Would be cool to hear AMD developers' progress about their
> kernel split work here in the list :)
>
> On Tue, Dec 2, 2014 at 6:46 PM, Nazim Mer <nazim.mer at gmail.com> wrote:
>
>> Hello Blenderheads,
>>
>> Has any one received an update in regards to the progress of the AMD's
>> developers approach.
>>
>> Regards,
>>
>> Nazim
>>
>> On 2 December 2014 at 12:17, Sergey Sharybin <sergey.vfx at gmail.com>
>> wrote:
>>
>>> Nirved, AMD developers announced here in the list they'll be working on
>>> splitting the Cycles megakernel to smaller kernels, similar to wavefront
>>> path tracer paper.
>>>
>>>
>>> So my guess would be that it's more productive to collaborate on that
>>> work instead of initializing new one.
>>>
>>> On Mon, Dec 1, 2014 at 3:33 AM, storm <kartochka22 at yandex.ru> wrote:
>>>
>>>> Great. How are you going to disassemble binary blob, then fix AMD
>>>> proprietary OpenCL compiler bugs, assemble/patch binary compiler or make
>>>> your own? Or you going Luxrender (Dade) way to toss lines in random
>>>> order untill it can render few specific scenes, and hope that next month
>>>> Catalyst AMD proprietary blob will not runie it?
>>>>
>>>> В Вс, 30/11/2014 в 19:46 +0200, Nirved пишет:
>>>> > Hello,
>>>> >
>>>> >
>>>> > I'm going to work on finally making cycles work on my hardware - R9
>>>> > 290, Linux.
>>>> >
>>>> >
>>>> > The project has four phases:
>>>> >
>>>> > 1. Parse CPU cycles code.
>>>> >
>>>> > 2. Interpret, estimate performance in best/worst case.
>>>> >
>>>> > 3. Grok code, use statistics from phase 2; divide & conqueror,
>>>> > transform & conqueror. This requires some AI.
>>>> >
>>>> > 4. Produce OpenCL code.
>>>> >
>>>> >
>>>> > Each phase will take about a week.
>>>> >
>>>> >
>>>> > I expect the project to be done before 6th January; if lucky before
>>>> > 24th December.
>>>> >
>>>> >
>>>> > Regards
>>>> >
>>>> > Nirved
>>>> >
>>>> > _______________________________________________
>>>> > Bf-cycles mailing list
>>>> > Bf-cycles at blender.org
>>>> > http://lists.blender.org/mailman/listinfo/bf-cycles
>>>>
>>>>
>>>> _______________________________________________
>>>> Bf-cycles mailing list
>>>> Bf-cycles at blender.org
>>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>>
>>>
>>>
>>>
>>> --
>>> With best regards, Sergey Sharybin
>>>
>>> _______________________________________________
>>> Bf-cycles mailing list
>>> Bf-cycles at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>>
>>>
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>>
>
>
> --
> With best regards, Sergey Sharybin
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20150111/370187a8/attachment.htm 


More information about the Bf-cycles mailing list