[Bf-committers] [Bf-blender-cvs] SVN commit: /data/svn/bf-blender [57543] trunk/blender/intern/cycles: Cycles: optimization for BVH traveral on CPU's with SSE3, using code from Embree.

[Jürgen Herrmann]

I updated my Benchmark with the new builds and the results are astonishing!

http://download.shadowrom.de/Benchmark.pdf

More than 35% performance gain *yay*

-----Ursprüngliche Nachricht-----
Von: bf-committers-bounces at blender.org [mailto:bf-committers-bounces at blender.org] Im Auftrag von Jürgen Herrmann
Gesendet: Dienstag, 18. Juni 2013 15:34
An: bf-blender developers
Betreff: Re: [Bf-committers] [Bf-blender-cvs] SVN commit: /data/svn/bf-blender [57543] trunk/blender/intern/cycles: Cycles: optimization for BVH traveral on CPU's with SSE3, using code from Embree.

I just tested r57548 (MSVC 2012) on my Core i5 system and got a 23,38% speed up ;) BMW test scene rendered in 530,60 seconds that is nearly the speed of MinGW build r57169 (512,67 seconds)
 
Great work, thanks Brecht!
I'll test on my other systems later today.

Am 18.06.2013 um 15:26 schrieb Jürgen Herrmann <shadowrom at me.com>:

> Hi Brecht,
> 
> It compiled fine on VS 2012 could you please change your compile fix patch to compile with the previous settings in MSVC11?
> The exception is only needed for VS2008.
> 
> Am 18.06.2013 um 15:21 schrieb Brecht Van Lommel <brechtvanlommel at pandora.be>:
> 
>> Looks like a compiler bug. I didn't get that here but I committed 
>> some more changes to try to fix it. If it doesn't fix things can you 
>> tell me if this is 32 or 64 bit and cmake or scons?
>> 
>> On Tue, Jun 18, 2013 at 12:09 PM, Thomas Dinges <blender at dingto.org> wrote:
>>> Hi Brecht,
>>> great work, thank you! On Linux I get a nice speedup. :)
>>> 
>>> Windows fails to compile for me though (vc2008).
>>> 
>>> Linking library ==> 'bf_intern_cycles_sse2.lib'
>>> kernel_sse3.cpp
>>> d:\blender_dev\code\trunk\intern\cycles\kernel\svm\svm_noise.h(38) :
>>> fatal error
>>> C1001: Internal compile error.
>>> (Compilerdatei
>>> "f:\dd\vctools\compiler\utc\src\p2\main.c[0x00000000604E0000:0x00
>>> 000000604E0000]", Zeile 182)
>>> 
>>> Any idea? I tried a full clean build already.
>>> 
>>> Thanks,
>>> Thomas
>>> 
>>> Am 18.06.2013 11:36, schrieb Brecht Van Lommel:
>>>> Revision: 57543
>>>>          http://projects.blender.org/scm/viewvc.php?view=rev&root=bf-blender&revision=57543
>>>> Author:   blendix
>>>> Date:     2013-06-18 09:36:06 +0000 (Tue, 18 Jun 2013)
>>>> Log Message:
>>>> -----------
>>>> Cycles: optimization for BVH traveral on CPU's with SSE3, using code from Embree.
>>>> 
>>>> On the BMW scene, this gives roughly a 10% speedup overall with 
>>>> clang/gcc, and 30% speedup with visual studio (2008). It turns out 
>>>> visual studio was optimizing the existing code quite poorly 
>>>> compared to pretty good autovectorization by clang/gcc, but hand written SSE code also gives a smaller speed boost there.
>>>> 
>>>> This code isn't enabled when using the hair minimum width feature 
>>>> yet, need to make that work with the SSE code still.
>>> 
>>> --
>>> Thomas Dinges
>>> Blender Developer, Artist and Musician
>>> 
>>> www.dingto.org
>>> 
>>> _______________________________________________
>>> Bf-committers mailing list
>>> Bf-committers at blender.org
>>> http://lists.blender.org/mailman/listinfo/bf-committers
>> _______________________________________________
>> Bf-committers mailing list
>> Bf-committers at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-committers
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
_______________________________________________
Bf-committers mailing list
Bf-committers at blender.org
http://lists.blender.org/mailman/listinfo/bf-committers