I've managed to maintain bit exactness in the past, as have others.

An equivalent test is 1 part in 10^4 difference in the output
synthesised speech.

If you can't demonstrate equivalence after optimisation, then pls don't
check in the code.

- David

On Tue, 2013-01-29 at 02:45 -0500, Albert Cahalan wrote:
> On Tue, Jan 29, 2013 at 12:00 AM, David Rowe <[email protected]> wrote:
> 
> > The best way is make sure the output (e.g. bit stream, decoded speech
> > samples) after any optimisation is bit exact with what it was before
> > optimisation.  Pls demonstrate this before checking in any modified
> > code.
> 
> This is already unreliable on x86. You're at the mercy of gcc's
> register allocator. If values stay in registers, you get 80-bit intermediates.
> If values are spilled to the stack, you get 64-bit or 32-bit intermediates.
> Unremarkable changes to nearby integer code can change code
> structure, causing or preventing FPU register spills. Any change of
> the compiler version will also change the FPU register spills.
> 
> So, even if I make zero changes to anything, this standard can't be met.
> It's already being violated.
> 
> ------------------------------------------------------------------------------
> Master Visual Studio, SharePoint, SQL, ASP.NET, C# 2012, HTML5, CSS,
> MVC, Windows 8 Apps, JavaScript and much more. Keep your skills current
> with LearnDevNow - 3,200 step-by-step video tutorials by Microsoft
> MVPs and experts. ON SALE this month only -- learn more at:
> http://p.sf.net/sfu/learnnow-d2d
> _______________________________________________
> Freetel-codec2 mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/freetel-codec2



------------------------------------------------------------------------------
Master Visual Studio, SharePoint, SQL, ASP.NET, C# 2012, HTML5, CSS,
MVC, Windows 8 Apps, JavaScript and much more. Keep your skills current
with LearnDevNow - 3,200 step-by-step video tutorials by Microsoft
MVPs and experts. ON SALE this month only -- learn more at:
http://p.sf.net/sfu/learnnow-d2d
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2

Reply via email to