[oe-commits] org.oe.dev gst-ffmpeg 0.10.2 : Add armv5te optimizations
sirfred commit
openembedded-commits at lists.openembedded.org
Wed May 9 20:22:38 UTC 2007
gst-ffmpeg 0.10.2 : Add armv5te optimizations
* Fix IWMMXT and ARMV5TE autoconfiguration
* Add armv5te optimized IDCT algorithms from upstream ffmpeg
* Fix configure stage to autoreconfigure properly
* Bump up package release
Author: sirfred at openembedded.org
Branch: org.openembedded.dev
Revision: 93a022ce4bed6cb6364be32b60eeca3833859ebd
ViewMTN: http://monotone.openembedded.org/revision.psp?id=93a022ce4bed6cb6364be32b60eeca3833859ebd
Files:
1
packages/gstreamer/gst-ffmpeg
packages/gstreamer/gst-ffmpeg/armv5.patch
packages/gstreamer/gst-ffmpeg_0.10.2.bb
Diffs:
#
# mt diff -rf21d1e4f099c8aa5f4d101d7436b5f27bfbfc829 -r93a022ce4bed6cb6364be32b60eeca3833859ebd
#
#
#
# add_dir "packages/gstreamer/gst-ffmpeg"
#
# add_file "packages/gstreamer/gst-ffmpeg/armv5.patch"
# content [726ed5559e69e6ac717800729fd18219d1e9e2d1]
#
# patch "packages/gstreamer/gst-ffmpeg_0.10.2.bb"
# from [d727079fced7f772b8d488034de5c6b443a66590]
# to [e0e185738811f0607a57b4d5b4948e0b92301af6]
#
============================================================
--- packages/gstreamer/gst-ffmpeg/armv5.patch 726ed5559e69e6ac717800729fd18219d1e9e2d1
+++ packages/gstreamer/gst-ffmpeg/armv5.patch 726ed5559e69e6ac717800729fd18219d1e9e2d1
@@ -0,0 +1,1509 @@
+diff -Nur gst-ffmpeg-0.10.2/gst-libs/ext/ffmpeg/configure.ac gst-ffmpeg-0.10.2.new/gst-libs/ext/ffmpeg/configure.ac
+--- gst-ffmpeg-0.10.2/gst-libs/ext/ffmpeg/configure.ac 2006-09-23 15:35:21.000000000 +0200
++++ gst-ffmpeg-0.10.2.new/gst-libs/ext/ffmpeg/configure.ac 2007-05-01 12:23:39.000000000 +0200
+@@ -190,7 +190,7 @@
+ ARCH_X86=yes
+ ;;
+ # armv4l is a subset of armv5tel
+- armv4l|armv5tel)
++ arm|armv4l|armv5tel|armv5te)
+ TARGET_ARCH="armv4l"
+ ARCH_ARMV4L=yes
+ ;;
+@@ -364,11 +364,8 @@
+ AC_FF_ALLOW_DISABLE(HAVE_IWMMXT, iwmmxt, use ARM/IWMMXT optimizations,[
+ if test x$TARGET_ARCH = xarmv4l; then
+ AC_MSG_CHECKING(for support of IWMMXT optimizations)
+- AC_TRY_RUN([
+- int main () {
++ AC_TRY_COMPILE(,[
+ __asm__ __volatile__ ("wunpckelub wr6, wr4");
+- return 0;
+- }
+ ],[ HAVE_IWMMXT=yes && AC_MSG_RESULT(yes) ],[
+ HAVE_IWMMXT=no && AC_MSG_RESULT(no) ])
+ else
+@@ -376,6 +373,19 @@
+ fi
+ ])
+
++dnl ARMV5TE
++AC_FF_ALLOW_DISABLE(HAVE_ARMV5TE, armv5te, use ARMV5TE optimizations,[
++ if test x$TARGET_ARCH = xarmv4l; then
++ AC_MSG_CHECKING(for support of ARMV5TE specific instructions)
++ AC_TRY_COMPILE(, [
++ __asm__ __volatile__ ("smulbb a4,ip,a3");
++ ],[ HAVE_ARMV5TE=yes && AC_MSG_RESULT(yes) ],[
++ HAVE_ARMV5TE=no && AC_MSG_RESULT(no) ])
++ else
++ HAVE_ARMV5TE=no
++ fi
++])
++
+ dnl GProf (-p)
+ AC_FF_ALLOW_ENABLE(HAVE_GPROF, gprof, profiling with gprof,[
+ GPROF_FLAGS="-p"
+diff -Nur gst-ffmpeg-0.10.2/gst-libs/ext/ffmpeg/libavcodec/armv4l/dsputil_arm.c gst-ffmpeg-0.10.2.new/gst-libs/ext/ffmpeg/libavcodec/armv4l/dsputil_arm.c
+--- gst-ffmpeg-0.10.2/gst-libs/ext/ffmpeg/libavcodec/armv4l/dsputil_arm.c 2006-09-20 20:55:37.000000000 +0200
++++ gst-ffmpeg-0.10.2.new/gst-libs/ext/ffmpeg/libavcodec/armv4l/dsputil_arm.c 2007-05-01 12:23:40.000000000 +0200
+ * ARMv4L optimized DSP utils
+ * Copyright (c) 2001 Lionel Ulmer.
+ *
+- * This library is free software; you can redistribute it and/or
++ * This file is part of FFmpeg.
++ *
++ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+- * version 2 of the License, or (at your option) any later version.
++ * version 2.1 of the License, or (at your option) any later version.
+ *
+- * This library is distributed in the hope that it will be useful,
++ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+- * License along with this library; if not, write to the Free Software
++ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+-#include "dsputil.h"
++#include "../dsputil.h"
+ #ifdef HAVE_IPP
+ #include "ipp.h"
+ #endif
+@@ -27,6 +29,12 @@
+ extern void j_rev_dct_ARM(DCTELEM *data);
+ extern void simple_idct_ARM(DCTELEM *data);
+
++extern void simple_idct_armv5te(DCTELEM *data);
++extern void simple_idct_put_armv5te(uint8_t *dest, int line_size,
++ DCTELEM *data);
++extern void simple_idct_add_armv5te(uint8_t *dest, int line_size,
++ DCTELEM *data);
++
+ /* XXX: local hack */
+ static void (*ff_put_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size);
+ static void (*ff_add_pixels_clamped)(const DCTELEM *block, uint8_t *pixels, int line_size);
+@@ -196,8 +204,10 @@
+ ff_add_pixels_clamped = c->add_pixels_clamped;
+
+ if(idct_algo == FF_IDCT_AUTO){
+-#ifdef HAVE_IPP
++#if defined(HAVE_IPP)
+ idct_algo = FF_IDCT_IPP;
++#elif defined(HAVE_ARMV5TE)
++ idct_algo = FF_IDCT_SIMPLEARMV5TE;
+ #else
+ idct_algo = FF_IDCT_ARM;
+ #endif
+@@ -213,6 +223,13 @@
+ c->idct_add= simple_idct_ARM_add;
+ c->idct = simple_idct_ARM;
+ c->idct_permutation_type= FF_NO_IDCT_PERM;
++#ifdef HAVE_ARMV5TE
++ } else if (idct_algo==FF_IDCT_SIMPLEARMV5TE){
++ c->idct_put= simple_idct_put_armv5te;
++ c->idct_add= simple_idct_add_armv5te;
++ c->idct = simple_idct_armv5te;
++ c->idct_permutation_type = FF_NO_IDCT_PERM;
++#endif
+ #ifdef HAVE_IPP
+ } else if (idct_algo==FF_IDCT_IPP){
+ c->idct_put= simple_idct_ipp_put;
+--- gst-ffmpeg-0.10.2/gst-libs/ext/ffmpeg/libavcodec/armv4l/dsputil_arm_s.S 2006-02-19 00:04:59.000000000 +0100
++++ gst-ffmpeg-0.10.2.new/gst-libs/ext/ffmpeg/libavcodec/armv4l/dsputil_arm_s.S 2007-05-01 12:23:40.000000000 +0200
+@@ -2,20 +2,29 @@
+ @ ARMv4L optimized DSP utils
+ @ Copyright (c) 2004 AGAWA Koji <i (AT) atty (DOT) jp>
+ @
+-@ This library is free software; you can redistribute it and/or
++@ This file is part of FFmpeg.
++@
++@ FFmpeg is free software; you can redistribute it and/or
+ @ modify it under the terms of the GNU Lesser General Public
+ @ License as published by the Free Software Foundation; either
+-@ version 2 of the License, or (at your option) any later version.
++@ version 2.1 of the License, or (at your option) any later version.
+ @
+-@ This library is distributed in the hope that it will be useful,
++@ FFmpeg is distributed in the hope that it will be useful,
+ @ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ @ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ @ Lesser General Public License for more details.
+ @
+ @ You should have received a copy of the GNU Lesser General Public
+-@ License along with this library; if not, write to the Free Software
++@ License along with FFmpeg; if not, write to the Free Software
+ @ Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ @
++#if defined(__ARM_ARCH_5__) || \
++ defined(__ARM_ARCH_5T__) || \
++ defined(__ARM_ARCH_5TE__)
++#define PLD(code...) code
++#else
++#define PLD(code...)
++#endif
+
+ .macro ADJ_ALIGN_QUADWORD_D shift, Rd0, Rd1, Rd2, Rd3, Rn0, Rn1, Rn2, Rn3, Rn4
+ mov \Rd0, \Rn0, lsr #(\shift * 8)
+@@ -74,7 +83,7 @@
+ put_pixels16_arm:
+ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+ @ block = word aligned, pixles = unaligned
+- pld [r1]
++ PLD ( pld [r1] )
+ stmfd sp!, {r4-r11, lr} @ R14 is also called LR
+ adr r5, 5f
+ ands r4, r1, #3
+@@ -85,7 +94,7 @@
+ ldmia r1, {r4-r7}
+ add r1, r1, r2
+ stmia r0, {r4-r7}
+- pld [r1]
++ PLD ( pld [r1] )
+ subs r3, r3, #1
+ add r0, r0, r2
+ bne 1b
+@@ -95,7 +104,7 @@
+ ldmia r1, {r4-r8}
+ add r1, r1, r2
+ ADJ_ALIGN_QUADWORD_D 1, r9, r10, r11, r12, r4, r5, r6, r7, r8
+- pld [r1]
++ PLD ( pld [r1] )
+ subs r3, r3, #1
+ stmia r0, {r9-r12}
+ add r0, r0, r2
+@@ -106,7 +115,7 @@
+ ldmia r1, {r4-r8}
+ add r1, r1, r2
+ ADJ_ALIGN_QUADWORD_D 2, r9, r10, r11, r12, r4, r5, r6, r7, r8
+- pld [r1]
++ PLD ( pld [r1] )
+ subs r3, r3, #1
+ stmia r0, {r9-r12}
+ add r0, r0, r2
+@@ -117,7 +126,7 @@
+ ldmia r1, {r4-r8}
+ add r1, r1, r2
+ ADJ_ALIGN_QUADWORD_D 3, r9, r10, r11, r12, r4, r5, r6, r7, r8
+- pld [r1]
++ PLD ( pld [r1] )
+ subs r3, r3, #1
+ stmia r0, {r9-r12}
+ add r0, r0, r2
+@@ -136,7 +145,7 @@
+ put_pixels8_arm:
+ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+ @ block = word aligned, pixles = unaligned
+- pld [r1]
++ PLD ( pld [r1] )
+ stmfd sp!, {r4-r5,lr} @ R14 is also called LR
+ adr r5, 5f
+ ands r4, r1, #3
+@@ -147,7 +156,7 @@
+ ldmia r1, {r4-r5}
+ add r1, r1, r2
+ subs r3, r3, #1
+- pld [r1]
++ PLD ( pld [r1] )
+ stmia r0, {r4-r5}
+ add r0, r0, r2
+ bne 1b
+@@ -157,7 +166,7 @@
+ ldmia r1, {r4-r5, r12}
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD 1, r4, r5, r12
+- pld [r1]
++ PLD ( pld [r1] )
+ subs r3, r3, #1
+ stmia r0, {r4-r5}
+ add r0, r0, r2
+@@ -168,7 +177,7 @@
+ ldmia r1, {r4-r5, r12}
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD 2, r4, r5, r12
+- pld [r1]
++ PLD ( pld [r1] )
+ subs r3, r3, #1
+ stmia r0, {r4-r5}
+ add r0, r0, r2
+@@ -179,7 +188,7 @@
+ ldmia r1, {r4-r5, r12}
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD 3, r4, r5, r12
+- pld [r1]
++ PLD ( pld [r1] )
+ subs r3, r3, #1
+ stmia r0, {r4-r5}
+ add r0, r0, r2
+@@ -198,7 +207,7 @@
+ put_pixels8_x2_arm:
+ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+ @ block = word aligned, pixles = unaligned
+- pld [r1]
++ PLD ( pld [r1] )
+ stmfd sp!, {r4-r10,lr} @ R14 is also called LR
+ adr r5, 5f
+ ands r4, r1, #3
+@@ -210,7 +219,7 @@
+ ldmia r1, {r4-r5, r10}
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 1, r6, r7, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ RND_AVG32 r8, r9, r4, r5, r6, r7, r12
+ subs r3, r3, #1
+ stmia r0, {r8-r9}
+@@ -223,7 +232,7 @@
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 1, r6, r7, r4, r5, r10
+ ADJ_ALIGN_DOUBLEWORD_D 2, r8, r9, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ RND_AVG32 r4, r5, r6, r7, r8, r9, r12
+ subs r3, r3, #1
+ stmia r0, {r4-r5}
+@@ -236,7 +245,7 @@
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 2, r6, r7, r4, r5, r10
+ ADJ_ALIGN_DOUBLEWORD_D 3, r8, r9, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ RND_AVG32 r4, r5, r6, r7, r8, r9, r12
+ subs r3, r3, #1
+ stmia r0, {r4-r5}
+@@ -248,7 +257,7 @@
+ ldmia r1, {r4-r5, r10}
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 3, r6, r7, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ RND_AVG32 r8, r9, r6, r7, r5, r10, r12
+ subs r3, r3, #1
+ stmia r0, {r8-r9}
+@@ -267,7 +276,7 @@
+ put_no_rnd_pixels8_x2_arm:
+ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+ @ block = word aligned, pixles = unaligned
+- pld [r1]
++ PLD ( pld [r1] )
+ stmfd sp!, {r4-r10,lr} @ R14 is also called LR
+ adr r5, 5f
+ ands r4, r1, #3
+@@ -279,7 +288,7 @@
+ ldmia r1, {r4-r5, r10}
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 1, r6, r7, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ NO_RND_AVG32 r8, r9, r4, r5, r6, r7, r12
+ subs r3, r3, #1
+ stmia r0, {r8-r9}
+@@ -292,7 +301,7 @@
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 1, r6, r7, r4, r5, r10
+ ADJ_ALIGN_DOUBLEWORD_D 2, r8, r9, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ NO_RND_AVG32 r4, r5, r6, r7, r8, r9, r12
+ subs r3, r3, #1
+ stmia r0, {r4-r5}
+@@ -305,7 +314,7 @@
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 2, r6, r7, r4, r5, r10
+ ADJ_ALIGN_DOUBLEWORD_D 3, r8, r9, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ NO_RND_AVG32 r4, r5, r6, r7, r8, r9, r12
+ subs r3, r3, #1
+ stmia r0, {r4-r5}
+@@ -317,7 +326,7 @@
+ ldmia r1, {r4-r5, r10}
+ add r1, r1, r2
+ ADJ_ALIGN_DOUBLEWORD_D 3, r6, r7, r4, r5, r10
+- pld [r1]
++ PLD ( pld [r1] )
+ NO_RND_AVG32 r8, r9, r6, r7, r5, r10, r12
+ subs r3, r3, #1
+ stmia r0, {r8-r9}
+@@ -338,7 +347,7 @@
+ put_pixels8_y2_arm:
+ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+ @ block = word aligned, pixles = unaligned
+- pld [r1]
++ PLD ( pld [r1] )
+ stmfd sp!, {r4-r11,lr} @ R14 is also called LR
+ adr r5, 5f
+ ands r4, r1, #3
+@@ -352,13 +361,13 @@
+ add r1, r1, r2
+ 6: ldmia r1, {r6-r7}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ RND_AVG32 r8, r9, r4, r5, r6, r7, r12
+ ldmia r1, {r4-r5}
+ add r1, r1, r2
+ stmia r0, {r8-r9}
+ add r0, r0, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ RND_AVG32 r8, r9, r6, r7, r4, r5, r12
+ subs r3, r3, #1
+ stmia r0, {r8-r9}
+@@ -369,18 +378,18 @@
+ 2:
+ ldmia r1, {r4-r6}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 1, r4, r5, r6
+ 6: ldmia r1, {r7-r9}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 1, r7, r8, r9
+ RND_AVG32 r10, r11, r4, r5, r7, r8, r12
+ stmia r0, {r10-r11}
+ add r0, r0, r2
+ ldmia r1, {r4-r6}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 1, r4, r5, r6
+ subs r3, r3, #1
+ RND_AVG32 r10, r11, r7, r8, r4, r5, r12
+@@ -392,18 +401,18 @@
+ 3:
+ ldmia r1, {r4-r6}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 2, r4, r5, r6
+ 6: ldmia r1, {r7-r9}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 2, r7, r8, r9
+ RND_AVG32 r10, r11, r4, r5, r7, r8, r12
+ stmia r0, {r10-r11}
+ add r0, r0, r2
+ ldmia r1, {r4-r6}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 2, r4, r5, r6
+ subs r3, r3, #1
+ RND_AVG32 r10, r11, r7, r8, r4, r5, r12
+@@ -415,18 +424,18 @@
+ 4:
+ ldmia r1, {r4-r6}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 3, r4, r5, r6
+ 6: ldmia r1, {r7-r9}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 3, r7, r8, r9
+ RND_AVG32 r10, r11, r4, r5, r7, r8, r12
+ stmia r0, {r10-r11}
+ add r0, r0, r2
+ ldmia r1, {r4-r6}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 3, r4, r5, r6
+ subs r3, r3, #1
+ RND_AVG32 r10, r11, r7, r8, r4, r5, r12
+@@ -447,7 +456,7 @@
+ put_no_rnd_pixels8_y2_arm:
+ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h)
+ @ block = word aligned, pixles = unaligned
+- pld [r1]
++ PLD ( pld [r1] )
+ stmfd sp!, {r4-r11,lr} @ R14 is also called LR
+ adr r5, 5f
+ ands r4, r1, #3
+@@ -461,13 +470,13 @@
+ add r1, r1, r2
+ 6: ldmia r1, {r6-r7}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ NO_RND_AVG32 r8, r9, r4, r5, r6, r7, r12
+ ldmia r1, {r4-r5}
+ add r1, r1, r2
+ stmia r0, {r8-r9}
+ add r0, r0, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ NO_RND_AVG32 r8, r9, r6, r7, r4, r5, r12
+ subs r3, r3, #1
+ stmia r0, {r8-r9}
+@@ -478,18 +487,18 @@
+ 2:
+ ldmia r1, {r4-r6}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD 1, r4, r5, r6
+ 6: ldmia r1, {r7-r9}
+ add r1, r1, r2
+- pld [r1]
++ PLD ( pld [r1] )
+ ADJ_ALIGN_DOUBLEWORD %s
>>> DIFF TRUNCATED @ 16K
More information about the Openembedded-commits
mailing list