[FFmpeg-devel] [RFC] wma encoder regtest
Tue Feb 17 16:14:41 CET 2009
On Tue, Feb 17, 2009 at 03:38:19PM +0100, Benjamin Larsson wrote:
> Michael Niedermayer wrote:
> > On Tue, Feb 17, 2009 at 02:32:30PM +0100, Benjamin Larsson wrote:
> >> I suggest we disable the wmaencoder regtest. Rounding the sine tables
> >> and bypassing the fft didn't work on all platforms. IMO the decode test
> >> of encoded data is enough.
> >> And regardless of the result I wont spend any
> >> more time in this matter.
> > good i was alraedy confused by you continously posting patches that could
> > not have worked
> > if you want a fixed point FT write the 5 lines of c code that implements one
> > and use a fixed point sin()
> > using a float fft and float sin with funny rounding in the middle
> Last patch did not use a fft at all. The permute and fft call was a nop.
> Comparable to your suggestion of using memcpy instead. Feel free to fix
> the regtest in whatever way you prefer.
until i succeed in reproducing the problem i cant do anything
once i reproduced it i have to investigate why they differ
once i know why they differ (this may be a bug or rounding or whatever)
i can think about how to solve it.
if anything you or someone else did was part of documenting how to reproduce
and analysis the cause
of the issue id like to hear it as it might safe me some time.
but 99% of what i read where hacks based on guesses of where the problem is
and suggestions of what to break and disable similarly based on wild guesses
summary reading the whole wma discussion was a waste of time i could have
spend debuging it properly myself in the first place.
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
Opposition brings concord. Out of discord comes the fairest harmony.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: Digital signature
More information about the ffmpeg-devel