[Bug 31458] New: Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

Bug #: 31458
           Summary: Fallout 3: significant performance regression
           Product: Wine
           Version: 1.5.5
          Platform: x86
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: directx-dsound
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]
                CC: [e-mail]
    Classification: Unclassified
   Regression SHA1: 275dfb83f2371a476b615b260465afb3cdf67f32

Created attachment 41350
  --> http://bugs.winehq.org/attachment.cgi?id=41350
Remove the windowed-sinc resampler.

I haven't played Fallout 3 in a while and decided that I finally wanted to try
and finish it up.  Unfortunately, I discovered a performance regression that's
so bad that the game is virtually unplayable (the game freezes up almost
constantly).  A regression test returns the following:
====
275dfb83f2371a476b615b260465afb3cdf67f32 is the first bad commit
commit 275dfb83f2371a476b615b260465afb3cdf67f32
Author: Alexander E. Patrakov 
Date:   Mon May 7 00:06:13 2012 +0600

dsound: Added a windowed-sinc resampler.

:040000 040000 8ecb8b2fbc5c65c2066aaf9a5e8c58104996c5ab
3768c38fdafc656c5e18e14ad70884ea7864b758 M    dlls
:040000 040000 680c52a0ebe164ca9b6a42d40c6521a91ae86e96
83407150c9beff0aa69d2d2e017b522543ac347f M    tools
====

Unfortunately this does not remove cleanly off of latest git, so I've attached
a patch that removes the resampler (based against wine-1.5.10).

[Bug 31458] Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

--- Comment #1 from Alexander E. Patrakov  2012-08-12 23:44:51 CDT ---
I think that outright removing the resampler for all applications is not the
right thing. It was added for a reason - very bad sound quality in media
players.

Bug 30639 has a patch (possibly outdated) that switches to a low-quality
resampler only if enough voices are playing for performance concerns to become
relevant. If that patch also resolves your problem, please mark your bug as a
duplicate. Thanks.

[Bug 31458] Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

--- Comment #2 from Alexander E. Patrakov  2012-08-12 23:51:34 CDT ---
Or rather, could you please look at these patches instead:

http://www.winehq.org/pipermail/wine-patches/2012-May/114356.html

http://www.winehq.org/pipermail/wine-patches/2012-May/114357.html

[Bug 31458] Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

Erich Hoover  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #41350|0                           |1
        is obsolete|                            |

--- Comment #3 from Erich Hoover  2012-08-13 07:29:51 CDT ---
Created attachment 41353
  --> http://bugs.winehq.org/attachment.cgi?id=41353
Simpler removal of windowed-sinc resampler.

(In reply to comment #2)
> Or rather, could you please look at these patches instead:
>
> http://www.winehq.org/pipermail/wine-patches/2012-May/114356.html
>
> http://www.winehq.org/pipermail/wine-patches/2012-May/114357.html

Well, these patches work - but I'm not sure that this game needs the resampler
at all.  Maybe I'm just not in tune with such things, but it doesn't sound like
the audio is messed up when there's no resampler.  I've attached a simpler
patch that I used to test this.

[Bug 31458] Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

--- Comment #4 from Alexander E. Patrakov  2012-08-13 08:36:31 CDT ---
Given that the game uses a lot of CPU when a high-quality resampler is used, it
does need a resampler. Even as simple as zero-order-hold (which is what was
there before I added a windowed-sinc=based one, and which is what your patch
effectively does).

I say so because there is a condition in the code to disable resampling when
the input and output sample rates match.

As for your patch - I completely disagree. Please listen to a sample from 26984
in foobar2000.

[Bug 31458] Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

Erich Hoover  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |DUPLICATE

--- Comment #5 from Erich Hoover  2012-08-13 09:07:45 CDT ---
(In reply to comment #4)
> Given that the game uses a lot of CPU when a high-quality resampler is used, it
> does need a resampler. Even as simple as zero-order-hold (which is what was
> there before I added a windowed-sinc=based one, and which is what your patch
> effectively does).
> ...

Ok, I'll trust you - I honestly can't tell the difference in the audio between
the two cases, but it may just be that explosions sound like explosions no
matter how they're resampled ;)

*** This bug has been marked as a duplicate of bug 30639 ***

[Bug 31458] Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

--- Comment #6 from Alexander E. Patrakov  2012-08-14 00:10:37 CDT ---
(In reply to comment #5)
> Ok, I'll trust you - I honestly can't tell the difference in the audio between
> the two cases, but it may just be that explosions sound like explosions no
> matter how they're resampled ;)

It is indeed the case. To hear the difference, you need some tonal content -
e.g. background music, an alarm, birds or even human speech.

[Bug 31458] Fallout 3: significant performance regression

http://bugs.winehq.org/show_bug.cgi?id=31458

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #7 from Austin English  2012-08-21 17:51:35 CDT ---
Closing.