[Bf-committers] MSVC official builds, optimisation and render errors

[Simon Clitherow]

Ok, just compiled the render module with /fp:fast and the problem is 
gone.  Would it cause a problem if
only this module gets /fp:fast while the others stay /fp:precise?


Eugene Reilly wrote:
> Throwing my hat into the ring - I just tested the test file in the bug
> report, and my builds [1] using MSVC 8 (Visual Express 2005) don't
> have any dots like in the image on BA.  I use use /O2 /fp:fast (except
> on the elbeem and src projects), SSE2, and largeaddressaware.  So
> though /O2 alone might be the culprit, it seems that this combination
> doesn't cause a problem for me.  Of course, using these flags for a
> release wouldn't be the way to go, but at least we can start tracking
> down what's really happening.
>
> [1] http://www.graphicall.org/builds/builds/showbuild.php?action=show&id=419
>
> Eugene (etr9j)
>
>   
>> Message: 9
>> Date: Wed, 25 Jul 2007 18:31:26 +0100
>> From: Simon Clitherow <aphex at nildram.co.uk>
>> Subject: Re: [Bf-committers] MSVC official builds, optimisation and
>>        render errors
>> To: bf-blender developers <bf-committers at blender.org>
>> Message-ID: <46A788EE.7020705 at nildram.co.uk>
>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>
>> Hi,
>>
>> Tested this today with VS2005 and the error occurs when using /O2
>> optimization.  When using no optimization or /O1, the error is gone. VC
>> is not performing any other fancy optimizations beyond these flags.
>> I've played around with floating-point precision flags, inline expansion
>> flags, etc, but nothing seems to touch it.
>>
>> So, the question is - do we resort to /O1 for this windows release?
>>
>> Does anyone have thoughts on this?  When did this error first appear?
>>
>> Simon
>>     
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>
>
>