[FFmpeg-devel] [PATCH] make building swscale rgb template conditional

[Michael Niedermayer]

On Thu, Aug 26, 2010 at 06:24:37AM +0200, Luca Barbato wrote:
> On 08/25/2010 04:22 PM, Michael Niedermayer wrote:
> > On Wed, Aug 25, 2010 at 04:14:21PM +0200, Luca Barbato wrote:
> >> On 08/25/2010 10:18 AM, Michael Niedermayer wrote:
> >>> On Wed, Aug 25, 2010 at 12:34:01AM -0300, Ramiro Polla wrote:
> >>>> On Tue, Aug 17, 2010 at 7:39 AM, Michael Niedermayer <michaelni at gmx.at> wrote:
> >>>>> On Sun, Aug 15, 2010 at 10:34:47PM +0200, Luca Barbato wrote:
> >>>>>> $subj + why the sse2 variant is always disabled?
> >>>>>>
> >>>>>> lu
> >>>>>>
> >>>>>> --
> >>>>>>
> >>>>>> Luca Barbato
> >>>>>> Gentoo/linux
> >>>>>> http://dev.gentoo.org/~lu_zero
> >>>>>>
> >>>>>
> >>>>>>  rgb2rgb.c          |   78 ++++++++++++++++++------------
> >>>>>>  rgb2rgb_template.c |  134 ++++++++++++++++++++++++++---------------------------
> >>>>>>  2 files changed, 114 insertions(+), 98 deletions(-)
> >>>>>> 5c27bf038a4cebac04892e6434ff946149b89a22  swscale-make-rgb-templates-build-conditional.diff
> >>>>>> diff --git a/rgb2rgb.c b/rgb2rgb.c
> >>>>>> index f92e165..3a6395a 100644
> >>>>>> --- a/rgb2rgb.c
> >>>>>> +++ b/rgb2rgb.c
> >>>>>> @@ -148,41 +148,57 @@ DECLARE_ASM_CONST(8, uint64_t, blue_15mask)  = 0x0000001f0000001fULL;
> >>>>>>
> >>>>>>  //Note: We have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW + MMX2 one.
> >>>>>>  //plain C versions
> >>>>>> -#undef HAVE_MMX
> >>>>>> -#undef HAVE_MMX2
> >>>>>> -#undef HAVE_AMD3DNOW
> >>>>>> -#undef HAVE_SSE2
> >>>>>> -#define HAVE_MMX 0
> >>>>>> -#define HAVE_MMX2 0
> >>>>>> -#define HAVE_AMD3DNOW 0
> >>>>>> -#define HAVE_SSE2 0
> >>>>>> +
> >>>>>> +#define TEMPLATE_SSE2 0 // never enabled.
> >>>>>> +
> >>>>>> +#define TEMPLATE_MMX 0
> >>>>>> +#define TEMPLATE_MMX2 0
> >>>>>> +#define TEMPLATE_AMD3DNOW 0
> >>>>>>  #define RENAME(a) a ## _C
> >>>>>>  #include "rgb2rgb_template.c"
> >>>>>> +#undef RENAME
> >>>>>> +#undef TEMPLATE_AMD3DNOW
> >>>>>> +#undef TEMPLATE_MMX2
> >>>>>> +#undef TEMPLATE_MMX
> >>>>>>
> >>>>>>  #if ARCH_X86
> >>>>>>
> >>>>>>  //MMX versions
> >>>>>> -#undef RENAME
> >>>>>> -#undef HAVE_MMX
> >>>>>> -#define HAVE_MMX 1
> >>>>>> -#define RENAME(a) a ## _MMX
> >>>>>> -#include "rgb2rgb_template.c"
> >>>>>> -
> >>>>>> +#ifdef HAVE_MMX
> >>>>>> +#   define TEMPLATE_MMX 1
> >>>>>> +#   define TEMPLATE_MMX2 0
> >>>>>> +#   define TEMPLATE_AMD3DNOW 0
> >>>>>> +#   define RENAME(a) a ## _MMX
> >>>>>> +#   include "rgb2rgb_template.c"
> >>>>>> +#   undef RENAME
> >>>>>> +#   undef TEMPLATE_AMD3DNOW
> >>>>>> +#   undef TEMPLATE_MMX2
> >>>>>> +#   undef TEMPLATE_MMX
> >>>>>
> >>>>> inconsistent to the names used in swscale*
> >>>>
> >>>> No reply from Luca, so I send a new patchset that works like in
> >>>> swscale.c (and factors some code out).
> >>>
> >>>>  swscale.c |   40 ++++++++++++++++++++++++++++------------
> >>>>  1 file changed, 28 insertions(+), 12 deletions(-)
> >>>> 1c823b32dc37bc9e2e0de54e170f1e9f12d8e22f  0001-swscale-Always-define-COMPILE_xxx-to-either-0-or-1.patch
> >>>> From d24ed5ef306b8464439ba3c73fb07414e368f5da Mon Sep 17 00:00:00 2001
> >>>> From: Ramiro Polla <ramiro.polla at gmail.com>
> >>>> Date: Wed, 25 Aug 2010 00:12:46 -0300
> >>>> Subject: [PATCH] swscale: Always define COMPILE_xxx to either 0 or 1.
> >>>
> >>> ok
> >>> [...]
> >>>
> >>>
> >>>>  swscale.c |   48 +++++++++++++++++++++++++-----------------------
> >>>>  1 file changed, 25 insertions(+), 23 deletions(-)
> >>>> 97e2ce3157686bcd6c958d39046a1ba74aba845c  0002-swscale-simplify-selection-of-optimizations-to-comp.patch
> >>>> From a5a86307a1b2e0fc488174a4a878d81b8a3557b8 Mon Sep 17 00:00:00 2001
> >>>> From: Ramiro Polla <ramiro.polla at gmail.com>
> >>>> Date: Wed, 25 Aug 2010 00:13:33 -0300
> >>>> Subject: [PATCH] swscale: simplify selection of optimizations to compile.
> >>>
> >>> ok
> >>>
> >>> [...]
> >>>
> >>>
> >>>>  rgb2rgb.c          |   28 ++++-------
> >>>>  rgb2rgb_template.c |  134 ++++++++++++++++++++++++++---------------------------
> >>>>  2 files changed, 79 insertions(+), 83 deletions(-)
> >>>> cc67f107305b6399de3dcfcb90a29a7829df7855  0003-rgb2rgb-Do-not-misuse-HAVE_-defines.-Introduce-COMP.patch
> >>>> From 1197df87cd922bb2d9bc015f1429a5dc15d06bcb Mon Sep 17 00:00:00 2001
> >>>> From: Ramiro Polla <ramiro.polla at gmail.com>
> >>>> Date: Wed, 25 Aug 2010 00:14:03 -0300
> >>>> Subject: [PATCH] rgb2rgb: Do not misuse HAVE_ defines. Introduce COMPILE_TEMPLATE_ defines and
> >>>>  use them instead.
> >>>
> >>> SSE2 is always disabled after this patch
> >>> i dont think it was always disabled before, because the code likely was
> >>> tested when it was written
> >>
> >> Apparently not and that was the question I was waiting for an answer...
> > 
> > it did not work (in _mplayer_) when it was commited?
> 
> From a quick look, mplayer has HAVE_SSE2, ffmpeg has HAVE_SSE, so
> in mplayer interleaveBytes gets SSE2 instead of mmx no matter what, on
> ffmpeg the opposite...

hmm so it was working in mplayer (without runtime detect) but not ffmpeg ...
that should be fixed

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Dictatorship naturally arises out of democracy, and the most aggravated
form of tyranny and slavery out of the most extreme liberty. -- Plato
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100826/3fa227d6/attachment.pgp>