Your message dated Sun, 27 Oct 2013 03:30:47 +1030
with message-id <[email protected]>
and subject line Re: Bug#700602: speex: Build with floating point/asm
optimization on armhf.
has caused the Debian Bug report #700602,
regarding speex: Build with floating point/asm optimization on armhf.
to be marked as done.
This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.
(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact [email protected]
immediately.)
--
700602: http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=700602
Debian Bug Tracking System
Contact [email protected] with problems
--- Begin Message ---
Package: speex
Version: 1.2~rc1-7
Severity: wishlist
Tags: patch
User: [email protected]
Usertags: origin-ubuntu raring ubuntu-patch
Dear Maintainer,
In Ubuntu raring, we enabled asm optimizations on armhf to improve the
resampling performance with PulseAudio.
*** /tmp/tmpxt7lqr/bug_body
In Ubuntu, the attached patch was applied to achieve the following:
* Build with float on armhf (LP: #1125295)
Thanks for considering the patch.
-- System Information:
Debian Release: wheezy/sid
APT prefers raring-updates
APT policy: (500, 'raring-updates'), (500, 'raring-security'), (500, 'raring')
Architecture: amd64 (x86_64)
Foreign Architectures: i386
Kernel: Linux 3.8.0-6-generic (SMP w/4 CPU cores)
Locale: LANG=en_AU.UTF-8, LC_CTYPE=en_AU.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
diff -u speex-1.2~rc1/debian/rules speex-1.2~rc1/debian/rules
--- speex-1.2~rc1/debian/rules
+++ speex-1.2~rc1/debian/rules
@@ -56,9 +56,9 @@
arch_objs = $(objdir)
ifeq ($(DEB_HOST_ARCH_CPU),arm)
-objdir = $(objdir_fixedpoint)
ifneq ($(DEB_HOST_ARCH),armhf)
+objdir = $(objdir_fixedpoint)
EXTRA_CONFIG_FLAGS = --enable-arm4-asm
endif
diff -u speex-1.2~rc1/debian/changelog speex-1.2~rc1/debian/changelog
--- End Message ---
--- Begin Message ---
On Mon, Feb 18, 2013 at 01:46:44PM +0100, David Henningsson wrote:
> On 02/18/2013 01:09 PM, Luke Yelavich wrote:
> >On Fri, Feb 15, 2013 at 07:54:48AM GMT, Ron wrote:
> >>You realise that's _not_ actually what your patch does, right?
> >>
> >>>In Ubuntu, the attached patch was applied to achieve the following:
> >>>
> >>> * Build with float on armhf (LP: #1125295)
> >>
> >>Can you share with us the results of the benchmarking you did that
> >>actually demonstrates this improves performance in speex, since I don't
> >>see any indication of that being done in the LP achievement you refer to?
> >
> >Sorry, i am just the patch forwarder. I am CCing David Henningssen who wrote
> >the patch. David, mind explaining how you came to doin what you did in the
> >patch to get what we wanted with speex?
>
> My problem is not performance related, and not PulseAudio related
> either. The problem is that using fixed point completely disables
> some libspeex functions, such as automatic gain control.
>
> Without working AGC, mumble do not function correctly. I e, in
> mumble there is no sound at all (and a long stream of error
> messages) if you build libspeex with fixed point.
So from what I could gather from the context around all of this, there
were basically 2 totally separate issues which people have tried to
fudge past with the same 'solution' submitted here:
1. People are using pulse wrongly/badly and having performance trouble.
2. There are some things in speexdsp that don't have a fixed-point
implementation yet.
Since the answer to 1 is "Don't do that", and the correct answer to 2
is send a patch upstream to implement them, and neither of them would
actually be solved just fudging this in the package build for one arch,
and since all the benchmarking I've seen says we currently are using
the best choice for armhf, I'm going to close this bug here and not
apply this patch.
Which isn't to say I don't think getting a fixed point implementation
for extra things some people use is not a wishlist issue, but the best
way to address that will certainly be to send a patch upstream, not
just wish for it in the Debian BTS - and that really is a separate
issue to what's been requested here.
Ron
--- End Message ---
