[opus] Make check failure on clone from 31 January

[Marcello Caramma (mcaramma)]

gcc-4.4.real (Debian 4.4.5-8) 4.4.5

If I compile with -O1 the test is succesful
[...]
nfft=36 inverse=1,snr = 86.619364
[...]
nfft=60 inverse=1,snr = 88.728534
[...]
If I compile with -O2 the test fails

[...]
nfft=36 inverse=1,snr = 0.217968
** poor snr: 0.217968 **
[...]
nfft=60 inverse=1,snr = 12.874762
** poor snr: 12.874762 **
[...]

Marcello


________________________________________
From: le.businessman at gmail.com [le.businessman at gmail.com] on behalf of Tristan Matthews [tristan at sat.qc.ca]
Sent: 05 February 2014 18:40
To: Marcello Caramma (mcaramma)
Cc: opus at xiph.org
Subject: Re: [opus] Make check failure on clone from 31 January

On Wed, Feb 5, 2014 at 11:05 AM, Marcello Caramma (mcaramma)
<mcaramma at cisco.com> wrote:
> Hi,
>
> Apologies if this is a known issue, but running make on revision e3187444692195957eb66989622c7b1ad8448b06 fails one of the tests when using fixed point configuration (floating point is ok) on my linux x86.
> Note that libopus1.1, as extracted from the tar ball, is OK.
>
> Specifically, the tests that fail are in celt/tests/test_unit_mdct:
> nfft=32 inverse=0,snr = 85.341197
> nfft=32 inverse=1,snr = 96.285703
> nfft=256 inverse=0,snr = 86.726515
> nfft=256 inverse=1,snr = 87.308386
> nfft=512 inverse=0,snr = 82.577935
> nfft=512 inverse=1,snr = 87.103849
> nfft=1024 inverse=0,snr = 84.950837
> nfft=1024 inverse=1,snr = 87.105776
> nfft=2048 inverse=0,snr = 84.749139
> nfft=2048 inverse=1,snr = 87.243518
> nfft=36 inverse=0,snr = 84.424292
> nfft=36 inverse=1,snr = 0.217968
> ** poor snr: 0.217968 **
> nfft=40 inverse=0,snr = 83.806898
> nfft=40 inverse=1,snr = 87.400185
> nfft=60 inverse=0,snr = 84.981677
> nfft=60 inverse=1,snr = 12.874762
> ** poor snr: 12.874762 **
> nfft=120 inverse=0,snr = 82.344266
> nfft=120 inverse=1,snr = 84.955920
> nfft=240 inverse=0,snr = 83.273247
> nfft=240 inverse=1,snr = 85.040479
> nfft=480 inverse=0,snr = 83.702884
> nfft=480 inverse=1,snr = 84.168293
> nfft=960 inverse=0,snr = 84.481005
> nfft=960 inverse=1,snr = 83.929637
> nfft=1920 inverse=0,snr = 83.974087
> nfft=1920 inverse=1,snr = 84.177282
> FAIL: celt/tests/test_unit_mdct
>
> Running bisect on the repository gives:
>
> e43a0abe0a908603a71b0c35e8c2307a77a7211f is the first bad commit
> commit e43a0abe0a908603a71b0c35e8c2307a77a7211f
> Author: Jean-Marc Valin <jmvalin at jmvalin.ca>
> Date:   Fri Dec 27 00:10:54 2013 -0500
>
>     Removes the separate 1/8N rotation in the (I)MDCT and unmerges the MDCT sizes
>
>     Undoes commits f7547a4e and 72513f3c
>
>
> This is what I used to configure:
> ../configure --prefix=$HOME/opus_install  --enable-fixed-point
>
> These are the details of my system:
>
> uname -a
> Linux mcdebian 2.6.32-5-686 #1 SMP Mon Feb 25 01:04:36 UTC 2013 i686 GNU/Linux

What compiler are you using?

Best,
Tristan