[Webkit-unassigned] [Bug 75528] Optimize the multiply-add in Biquad.cpp::process

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=75528





--- Comment #37 from Raymond Toy <rtoy at chromium.org>  2012-03-15 10:39:35 PST ---
(In reply to comment #36)
> (In reply to comment #33)
> > (In reply to comment #31)
> > > This patch seems to crash in debug:
> > > 
> > > [49220:49236:17280644243061:ERROR:process_util_posix.cc(142)] Received signal 11
> > >     base::debug::StackTrace::StackTrace() [0x851b56]
> > >     base::(anonymous namespace)::StackDumpSignalHandler() [0x80b739]
> > >     0x7f352f9d1af0
> > >     WebCore::Biquad::process() [0x256d73e]
> > >     WebCore::HRTFKernel::HRTFKernel() [0x25ac2e4]
> > >     WebCore::HRTFKernel::create() [0x25ab549]
> > >     WebCore::HRTFElevation::calculateKernelsForAzimuthElevation() [0x25aa703]
> > >     WebCore::HRTFElevation::createForSubject() [0x25aa9db]
> > >     WebCore::HRTFDatabase::HRTFDatabase() [0x25a987e]
> > >     WebCore::HRTFDatabase::create() [0x25a9630]
> > >     WebCore::HRTFDatabaseLoader::load() [0x257382d]
> > >     WebCore::databaseLoaderEntry() [0x25737ac]
> > >     WTF::threadEntryPoint() [0x1f509e5]
> > >     WTF::wtfThreadEntryPoint() [0x793cb8]
> > >     start_thread [0x7f3531e369ca]
> > >     0x7f352fa8470d
> > > None
> > 
> > Hi,
> > That`s strange that I cannot reproduce the issue (by several different machines). Could you provide more details about your platform?
> 
> The stack trace looks totally bogus. (I'm pretty sure the HRTFKernel doesn't call Biquad).
> 
> But I can reproduce the crash on my system.  The stack trace is totally trash too.  More work needed....

Ok.  The issue is that on my debug build, I get a 64-bit build so loading the source address into edx loses the top 32 bits, which are non-zero.  We then try to load from that non-existent address and crash.

Don't understand why this doesn't happen on a release build, but this is definitely a bug.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.