diff options
| -rw-r--r-- | AUTHORS | 3 | ||||
| -rw-r--r-- | ChangeLog | 1 | ||||
| -rw-r--r-- | acconfig.h | 6 | ||||
| -rw-r--r-- | configure.in | 9 | ||||
| -rw-r--r-- | src/libmpeg2/Makefile.am | 3 | ||||
| -rw-r--r-- | src/libmpeg2/idct.c | 8 | ||||
| -rw-r--r-- | src/libmpeg2/idct_altivec.c | 668 | ||||
| -rw-r--r-- | src/libmpeg2/motion_comp.c | 6 | ||||
| -rw-r--r-- | src/libmpeg2/motion_comp_altivec.c | 2024 | ||||
| -rw-r--r-- | src/libmpeg2/mpeg2_internal.h | 10 | ||||
| -rw-r--r-- | src/xine-utils/cpu_accel.c | 34 | ||||
| -rw-r--r-- | src/xine-utils/xineutils.h | 4 |
12 files changed, 2756 insertions, 20 deletions
@@ -170,4 +170,7 @@ Contributions Juan Manuel García Molina <juanmagm@mail.com> spanish internationalization files. + Jeffrey W. Baker <jwbaker@acm.org> + altivec support for libmpeg2 + (let us know if we've forgotten anyone) @@ -6,6 +6,7 @@ xine (0.9.7) unstable; urgency=low * OSD (On Screen Display) for rendering text and graphics into overlays * reworked spu and overlay manager (multiple overlays supported) * support for avi text subtitles (use something like xine stream.avi:foo.sub) + * altivec support -- Guenter Bartsch <guenter@users.sourceforge.net> Tue Nov 27 01:20:06 CET 2001 diff --git a/acconfig.h b/acconfig.h index e0f1b61cc..eaaea2831 100644 --- a/acconfig.h +++ b/acconfig.h @@ -18,6 +18,12 @@ /* Define this if you're running PowerPC architecture */ #undef __ppc__ +/* Define this if you're running PowerPC architecture */ +#undef ARCH_PPC + +/* Define this if you have the Motorola 74xx CPU */ +#undef ENABLE_ALTIVEC + /* Define this if you're running Sparc architecture */ #undef __sparc__ diff --git a/configure.in b/configure.in index d01c1b35f..31bb299f4 100644 --- a/configure.in +++ b/configure.in @@ -177,6 +177,9 @@ LIBMPEG2_CFLAGS="" dnl default include path removed, no more needed. LIBA52_CFLAGS="" LIBFFMPEG_CFLAGS="" +AC_ARG_ENABLE(altivec, [ --enable-altivec use assembly codes for Motorola 74xx CPUs], + enable_altivec=yes, enable_altivec=no) + if test x$enable_mlib = x; then AC_ARG_ENABLE(mlib, [ --disable-mlib make a version not using mediaLib], @@ -561,6 +564,12 @@ case "$host_or_hostalias" in GLOBAL_CFLAGS="$GLOBAL_CFLAGS -O3 -pipe -fomit-frame-pointer $m_wm $m_psb -fexpensive-optimizations $f_si $f_nsa -ffast-math -funroll-loops -funroll-all-loops -finline-functions" DEBUG_CFLAGS="$DEBUG_CFLAGS -O3" AC_DEFINE(FPM_PPC) + AC_DEFINE(ARCH_PPC) + + if test x$enable_altivec = xyes; then + AC_DEFINE(ENABLE_ALTIVEC) + GLOBAL_CFLAGS="$GLOBAL_CFLAGS -Wa,-m7400" + fi ;; sparc*-*-linux*) diff --git a/src/libmpeg2/Makefile.am b/src/libmpeg2/Makefile.am index cf093d802..eec705b15 100644 --- a/src/libmpeg2/Makefile.am +++ b/src/libmpeg2/Makefile.am @@ -14,7 +14,8 @@ lib_LTLIBRARIES = xineplug_decode_mpeg2.la xineplug_decode_mpeg2_la_SOURCES = slice.c header.c stats.c idct.c \ motion_comp.c decode.c idct_mmx.c motion_comp_mmx.c \ - idct_mlib.c motion_comp_mlib.c xine_decoder.c + idct_mlib.c motion_comp_mlib.c idct_altivec.c \ + motion_comp_altivec.c xine_decoder.c xineplug_decode_mpeg2_la_LDFLAGS = -avoid-version -module noinst_HEADERS = vlc.h mpeg2.h mpeg2_internal.h diff --git a/src/libmpeg2/idct.c b/src/libmpeg2/idct.c index 2aaf6eaf9..0b77a62ed 100644 --- a/src/libmpeg2/idct.c +++ b/src/libmpeg2/idct.c @@ -84,6 +84,14 @@ void idct_init (void) idct_block_add = idct_block_add_mlib; } else #endif +#ifdef ENABLE_ALTIVEC + if (config.flags & MM_ACCEL_PPC_ALTIVEC) { + fprintf (stderr, "Using altivec for IDCT transform\n"); + idct_block_copy = idct_block_copy_altivec; + idct_block_add = idct_block_add_altivec; + idct_altivec_init (); + } else +#endif { int i; diff --git a/src/libmpeg2/idct_altivec.c b/src/libmpeg2/idct_altivec.c new file mode 100644 index 000000000..96be2ebc3 --- /dev/null +++ b/src/libmpeg2/idct_altivec.c @@ -0,0 +1,668 @@ +/* + * idct_altivec.c + * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * mpeg2dec is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * mpeg2dec 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#ifndef __ALTIVEC__ + +#include "config.h" + +#ifdef ARCH_PPC +#ifdef ENABLE_ALTIVEC + +#include <inttypes.h> + +#include "mpeg2_internal.h" +#include "xineutils.h" + +static int16_t constants[5][8] ATTR_ALIGN(16) = { + {23170, 13573, 6518, 21895, -23170, -21895, 32, 31}, + {16384, 22725, 21407, 19266, 16384, 19266, 21407, 22725}, + {22725, 31521, 29692, 26722, 22725, 26722, 29692, 31521}, + {21407, 29692, 27969, 25172, 21407, 25172, 27969, 29692}, + {19266, 26722, 25172, 22654, 19266, 22654, 25172, 26722} +}; + +/* + * The asm code is generated with: + * + * gcc-2.95 -fvec -D__ALTIVEC__ -O9 -fomit-frame-pointer -mregnames -S + * idct_altivec.c + * + * awk '{args=""; len=split ($2, arg, ","); + * for (i=1; i<=len; i++) { a=arg[i]; if (i<len) a=a","; + * args = args sprintf ("%-6s", a) } + * printf ("\t\"\t%-16s%-24s\\n\"\n", $1, args) }' idct_altivec.s | + * unexpand -a + * + * I then do some simple trimming on the function prolog/trailers + */ + +void idct_block_copy_altivec (int16_t * block, uint8_t * dest, int stride) +{ + asm (" \n" + "# stwu %r1, -128(%r1) \n" + "# mflr %r0 \n" + "# stw %r0, 132(%r1) \n" + "# addi %r0, %r1, 128 \n" + "# bl _savev25 \n" + + " addi %r9, %r3, 112 \n" + " vspltish %v25, 4 \n" + " vxor %v13, %v13, %v13 \n" + " lis %r10, constants@ha \n" + " lvx %v1, 0, %r9 \n" + " la %r10, constants@l(%r10) \n" + " lvx %v5, 0, %r3 \n" + " addi %r9, %r3, 16 \n" + " lvx %v8, 0, %r10 \n" + " addi %r11, %r10, 32 \n" + " lvx %v12, 0, %r9 \n" + " lvx %v6, 0, %r11 \n" + " addi %r8, %r3, 48 \n" + " vslh %v1, %v1, %v25 \n" + " addi %r9, %r3, 80 \n" + " lvx %v11, 0, %r8 \n" + " vslh %v5, %v5, %v25 \n" + " lvx %v0, 0, %r9 \n" + " addi %r11, %r10, 64 \n" + " vsplth %v3, %v8, 2 \n" + " lvx %v7, 0, %r11 \n" + " addi %r9, %r3, 96 \n" + " vslh %v12, %v12, %v25 \n" + " vmhraddshs %v27, %v1, %v6, %v13 \n" + " addi %r8, %r3, 32 \n" + " vsplth %v2, %v8, 5 \n" + " lvx %v1, 0, %r9 \n" + " vslh %v11, %v11, %v25 \n" + " addi %r3, %r3, 64 \n" + " lvx %v9, 0, %r8 \n" + " addi %r9, %r10, 48 \n" + " vslh %v0, %v0, %v25 \n" + " lvx %v4, 0, %r9 \n" + " vmhraddshs %v31, %v12, %v6, %v13 \n" + " addi %r10, %r10, 16 \n" + " vmhraddshs %v30, %v0, %v7, %v13 \n" + " lvx %v10, 0, %r3 \n" + " vsplth %v19, %v8, 3 \n" + " vmhraddshs %v15, %v11, %v7, %v13 \n" + " lvx %v12, 0, %r10 \n" + " vsplth %v6, %v8, 4 \n" + " vslh %v1, %v1, %v25 \n" + " vsplth %v11, %v8, 1 \n" + " li %r9, 4 \n" + " vslh %v9, %v9, %v25 \n" + " vsplth %v7, %v8, 0 \n" + " vmhraddshs %v18, %v1, %v4, %v13 \n" + " vspltw %v8, %v8, 3 \n" + " vsubshs %v0, %v13, %v27 \n" + " vmhraddshs %v1, %v9, %v4, %v13 \n" + " vmhraddshs %v17, %v3, %v31, %v0 \n" + " vmhraddshs %v4, %v2, %v15, %v30 \n" + " vslh %v10, %v10, %v25 \n" + " vmhraddshs %v9, %v5, %v12, %v13 \n" + " vspltish %v25, 6 \n" + " vmhraddshs %v5, %v10, %v12, %v13 \n" + " vmhraddshs %v28, %v19, %v30, %v15 \n" + " vmhraddshs %v27, %v3, %v27, %v31 \n" + " vsubshs %v0, %v13, %v18 \n" + " vmhraddshs %v18, %v11, %v18, %v1 \n" + " vaddshs %v30, %v17, %v4 \n" + " vmhraddshs %v12, %v11, %v1, %v0 \n" + " vsubshs %v4, %v17, %v4 \n" + " vaddshs %v10, %v9, %v5 \n" + " vsubshs %v17, %v27, %v28 \n" + " vaddshs %v27, %v27, %v28 \n" + " vsubshs %v1, %v9, %v5 \n" + " vaddshs %v28, %v10, %v18 \n" + " vsubshs %v18, %v10, %v18 \n" + " vaddshs %v10, %v1, %v12 \n" + " vsubshs %v1, %v1, %v12 \n" + " vsubshs %v12, %v17, %v4 \n" + " vaddshs %v4, %v17, %v4 \n" + " vmhraddshs %v5, %v7, %v12, %v1 \n" + " vmhraddshs %v26, %v6, %v4, %v10 \n" + " vmhraddshs %v29, %v6, %v12, %v1 \n" + " vmhraddshs %v14, %v7, %v4, %v10 \n" + " vsubshs %v12, %v18, %v30 \n" + " vaddshs %v9, %v28, %v27 \n" + " vaddshs %v16, %v18, %v30 \n" + " vsubshs %v10, %v28, %v27 \n" + " vmrglh %v31, %v9, %v12 \n" + " vmrglh %v30, %v5, %v26 \n" + " vmrglh %v15, %v14, %v29 \n" + " vmrghh %v5, %v5, %v26 \n" + " vmrglh %v27, %v16, %v10 \n" + " vmrghh %v9, %v9, %v12 \n" + " vmrghh %v18, %v16, %v10 \n" + " vmrghh %v1, %v14, %v29 \n" + " vmrglh %v14, %v9, %v5 \n" + " vmrglh %v16, %v31, %v30 \n" + " vmrglh %v10, %v15, %v27 \n" + " vmrghh %v9, %v9, %v5 \n" + " vmrghh %v26, %v15, %v27 \n" + " vmrglh %v27, %v16, %v10 \n" + " vmrghh %v12, %v1, %v18 \n" + " vmrglh %v29, %v1, %v18 \n" + " vsubshs %v0, %v13, %v27 \n" + " vmrghh %v5, %v31, %v30 \n" + " vmrglh %v31, %v9, %v12 \n" + " vmrglh %v30, %v5, %v26 \n" + " vmrglh %v15, %v14, %v29 \n" + " vmhraddshs %v17, %v3, %v31, %v0 \n" + " vmrghh %v18, %v16, %v10 \n" + " vmhraddshs %v27, %v3, %v27, %v31 \n" + " vmhraddshs %v4, %v2, %v15, %v30 \n" + " vmrghh %v1, %v14, %v29 \n" + " vmhraddshs %v28, %v19, %v30, %v15 \n" + " vmrghh %v0, %v9, %v12 \n" + " vsubshs %v13, %v13, %v18 \n" + " vmrghh %v5, %v5, %v26 \n" + " vmhraddshs %v18, %v11, %v18, %v1 \n" + " vaddshs %v9, %v0, %v8 \n" + " vaddshs %v30, %v17, %v4 \n" + " vmhraddshs %v12, %v11, %v1, %v13 \n" + " vsubshs %v4, %v17, %v4 \n" + " vaddshs %v10, %v9, %v5 \n" + " vsubshs %v17, %v27, %v28 \n" + " vaddshs %v27, %v27, %v28 \n" + " vsubshs %v1, %v9, %v5 \n" + " vaddshs %v28, %v10, %v18 \n" + " vsubshs %v18, %v10, %v18 \n" + " vaddshs %v10, %v1, %v12 \n" + " vsubshs %v1, %v1, %v12 \n" + " vsubshs %v12, %v17, %v4 \n" + " vaddshs %v4, %v17, %v4 \n" + " vaddshs %v9, %v28, %v27 \n" + " vmhraddshs %v14, %v7, %v4, %v10 \n" + " vsrah %v9, %v9, %v25 \n" + " vmhraddshs %v5, %v7, %v12, %v1 \n" + " vpkshus %v0, %v9, %v9 \n" + " vmhraddshs %v29, %v6, %v12, %v1 \n" + " stvewx %v0, 0, %r4 \n" + " vaddshs %v16, %v18, %v30 \n" + " vsrah %v31, %v14, %v25 \n" + " stvewx %v0, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vsrah %v15, %v16, %v25 \n" + " vpkshus %v0, %v31, %v31 \n" + " vsrah %v1, %v5, %v25 \n" + " stvewx %v0, 0, %r4 \n" + " vsubshs %v12, %v18, %v30 \n" + " stvewx %v0, %r9, %r4 \n" + " vmhraddshs %v26, %v6, %v4, %v10 \n" + " vpkshus %v0, %v1, %v1 \n" + " add %r4, %r4, %r5 \n" + " vsrah %v5, %v12, %v25 \n" + " stvewx %v0, 0, %r4 \n" + " vsrah %v30, %v29, %v25 \n" + " stvewx %v0, %r9, %r4 \n" + " vsubshs %v10, %v28, %v27 \n" + " vpkshus %v0, %v15, %v15 \n" + " add %r4, %r4, %r5 \n" + " stvewx %v0, 0, %r4 \n" + " vsrah %v18, %v26, %v25 \n" + " stvewx %v0, %r9, %r4 \n" + " vsrah %v27, %v10, %v25 \n" + " vpkshus %v0, %v5, %v5 \n" + " add %r4, %r4, %r5 \n" + " stvewx %v0, 0, %r4 \n" + " stvewx %v0, %r9, %r4 \n" + " vpkshus %v0, %v30, %v30 \n" + " add %r4, %r4, %r5 \n" + " stvewx %v0, 0, %r4 \n" + " stvewx %v0, %r9, %r4 \n" + " vpkshus %v0, %v18, %v18 \n" + " add %r4, %r4, %r5 \n" + " stvewx %v0, 0, %r4 \n" + " stvewx %v0, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vpkshus %v0, %v27, %v27 \n" + " stvewx %v0, 0, %r4 \n" + " stvewx %v0, %r9, %r4 \n" + + "# addi %r0, %r1, 128 \n" + "# bl _restv25 \n" + "# lwz %r0, 132(%r1) \n" + "# mtlr %r0 \n" + "# la %r1, 128(%r1) \n" + ); +} + +void idct_block_add_altivec (int16_t * block, uint8_t * dest, int stride) +{ + asm (" \n" + "# stwu %r1, -192(%r1) \n" + "# mflr %r0 \n" + "# stw %r0, 196(%r1) \n" + "# addi %r0, %r1, 192 \n" + "# bl _savev21 \n" + + " addi %r9, %r3, 112 \n" + " vspltish %v21, 4 \n" + " vxor %v1, %v1, %v1 \n" + " lvx %v13, 0, %r9 \n" + " lis %r10, constants@ha \n" + " vspltisw %v3, -1 \n" + " la %r10, constants@l(%r10) \n" + " lvx %v5, 0, %r3 \n" + " addi %r9, %r3, 16 \n" + " lvx %v8, 0, %r10 \n" + " lvx %v12, 0, %r9 \n" + " addi %r11, %r10, 32 \n" + " lvx %v6, 0, %r11 \n" + " addi %r8, %r3, 48 \n" + " vslh %v13, %v13, %v21 \n" + " addi %r9, %r3, 80 \n" + " lvx %v11, 0, %r8 \n" + " vslh %v5, %v5, %v21 \n" + " lvx %v0, 0, %r9 \n" + " addi %r11, %r10, 64 \n" + " vsplth %v2, %v8, 2 \n" + " lvx %v7, 0, %r11 \n" + " vslh %v12, %v12, %v21 \n" + " addi %r9, %r3, 96 \n" + " vmhraddshs %v24, %v13, %v6, %v1 \n" + " addi %r8, %r3, 32 \n" + " vsplth %v17, %v8, 5 \n" + " lvx %v13, 0, %r9 \n" + " vslh %v11, %v11, %v21 \n" + " addi %r3, %r3, 64 \n" + " lvx %v10, 0, %r8 \n" + " vslh %v0, %v0, %v21 \n" + " addi %r9, %r10, 48 \n" + " vmhraddshs %v31, %v12, %v6, %v1 \n" + " lvx %v4, 0, %r9 \n" + " addi %r10, %r10, 16 \n" + " vmhraddshs %v26, %v0, %v7, %v1 \n" + " lvx %v9, 0, %r3 \n" + " vsplth %v16, %v8, 3 \n" + " vmhraddshs %v22, %v11, %v7, %v1 \n" + " lvx %v6, 0, %r10 \n" + " lvsl %v19, 0, %r4 \n" + " vsubshs %v12, %v1, %v24 \n" + " lvsl %v0, %r5, %r4 \n" + " vsplth %v11, %v8, 1 \n" + " vslh %v10, %v10, %v21 \n" + " vmrghb %v19, %v3, %v19 \n" + " lvx %v15, 0, %r4 \n" + " vslh %v13, %v13, %v21 \n" + " vmrghb %v3, %v3, %v0 \n" + " li %r9, 4 \n" + " vmhraddshs %v14, %v2, %v31, %v12 \n" + " vsplth %v7, %v8, 0 \n" + " vmhraddshs %v23, %v13, %v4, %v1 \n" + " vsplth %v18, %v8, 4 \n" + " vmhraddshs %v27, %v10, %v4, %v1 \n" + " vspltw %v8, %v8, 3 \n" + " vmhraddshs %v12, %v17, %v22, %v26 \n" + " vperm %v15, %v15, %v1, %v19 \n" + " vslh %v9, %v9, %v21 \n" + " vmhraddshs %v10, %v5, %v6, %v1 \n" + " vspltish %v21, 6 \n" + " vmhraddshs %v30, %v9, %v6, %v1 \n" + " vmhraddshs %v26, %v16, %v26, %v22 \n" + " vmhraddshs %v24, %v2, %v24, %v31 \n" + " vmhraddshs %v31, %v11, %v23, %v27 \n" + " vsubshs %v0, %v1, %v23 \n" + " vaddshs %v23, %v14, %v12 \n" + " vmhraddshs %v9, %v11, %v27, %v0 \n" + " vsubshs %v12, %v14, %v12 \n" + " vaddshs %v6, %v10, %v30 \n" + " vsubshs %v14, %v24, %v26 \n" + " vaddshs %v24, %v24, %v26 \n" + " vsubshs %v13, %v10, %v30 \n" + " vaddshs %v26, %v6, %v31 \n" + " vsubshs %v31, %v6, %v31 \n" + " vaddshs %v6, %v13, %v9 \n" + " vsubshs %v13, %v13, %v9 \n" + " vsubshs %v9, %v14, %v12 \n" + " vaddshs %v12, %v14, %v12 \n" + " vmhraddshs %v30, %v7, %v9, %v13 \n" + " vmhraddshs %v25, %v18, %v12, %v6 \n" + " vmhraddshs %v28, %v18, %v9, %v13 \n" + " vmhraddshs %v29, %v7, %v12, %v6 \n" + " vaddshs %v10, %v26, %v24 \n" + " vsubshs %v5, %v31, %v23 \n" + " vsubshs %v13, %v26, %v24 \n" + " vaddshs %v4, %v31, %v23 \n" + " vmrglh %v26, %v30, %v25 \n" + " vmrglh %v31, %v10, %v5 \n" + " vmrglh %v22, %v29, %v28 \n" + " vmrghh %v30, %v30, %v25 \n" + " vmrglh %v24, %v4, %v13 \n" + " vmrghh %v10, %v10, %v5 \n" + " vmrghh %v23, %v4, %v13 \n" + " vmrghh %v27, %v29, %v28 \n" + " vmrglh %v29, %v10, %v30 \n" + " vmrglh %v4, %v31, %v26 \n" + " vmrglh %v13, %v22, %v24 \n" + " vmrghh %v10, %v10, %v30 \n" + " vmrghh %v25, %v22, %v24 \n" + " vmrglh %v24, %v4, %v13 \n" + " vmrghh %v5, %v27, %v23 \n" + " vmrglh %v28, %v27, %v23 \n" + " vsubshs %v0, %v1, %v24 \n" + " vmrghh %v30, %v31, %v26 \n" + " vmrglh %v31, %v10, %v5 \n" + " vmrglh %v26, %v30, %v25 \n" + " vmrglh %v22, %v29, %v28 \n" + " vmhraddshs %v14, %v2, %v31, %v0 \n" + " vmrghh %v23, %v4, %v13 \n" + " vmhraddshs %v24, %v2, %v24, %v31 \n" + " vmhraddshs %v12, %v17, %v22, %v26 \n" + " vmrghh %v27, %v29, %v28 \n" + " vmhraddshs %v26, %v16, %v26, %v22 \n" + " vmrghh %v0, %v10, %v5 \n" + " vmhraddshs %v31, %v11, %v23, %v27 \n" + " vmrghh %v30, %v30, %v25 \n" + " vsubshs %v13, %v1, %v23 \n" + " vaddshs %v10, %v0, %v8 \n" + " vaddshs %v23, %v14, %v12 \n" + " vsubshs %v12, %v14, %v12 \n" + " vaddshs %v6, %v10, %v30 \n" + " vsubshs %v14, %v24, %v26 \n" + " vmhraddshs %v9, %v11, %v27, %v13 \n" + " vaddshs %v24, %v24, %v26 \n" + " vaddshs %v26, %v6, %v31 \n" + " vsubshs %v13, %v10, %v30 \n" + " vaddshs %v10, %v26, %v24 \n" + " vsubshs %v31, %v6, %v31 \n" + " vaddshs %v6, %v13, %v9 \n" + " vsrah %v10, %v10, %v21 \n" + " vsubshs %v13, %v13, %v9 \n" + " vaddshs %v0, %v15, %v10 \n" + " vsubshs %v9, %v14, %v12 \n" + " vaddshs %v12, %v14, %v12 \n" + " vpkshus %v15, %v0, %v0 \n" + " stvewx %v15, 0, %r4 \n" + " vaddshs %v4, %v31, %v23 \n" + " vmhraddshs %v29, %v7, %v12, %v6 \n" + " stvewx %v15, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vsubshs %v5, %v31, %v23 \n" + " lvx %v15, 0, %r4 \n" + " vmhraddshs %v30, %v7, %v9, %v13 \n" + " vsrah %v22, %v4, %v21 \n" + " vperm %v15, %v15, %v1, %v3 \n" + " vmhraddshs %v28, %v18, %v9, %v13 \n" + " vsrah %v31, %v29, %v21 \n" + " vsubshs %v13, %v26, %v24 \n" + " vaddshs %v0, %v15, %v31 \n" + " vsrah %v27, %v30, %v21 \n" + " vpkshus %v15, %v0, %v0 \n" + " vsrah %v30, %v5, %v21 \n" + " stvewx %v15, 0, %r4 \n" + " vsrah %v26, %v28, %v21 \n" + " stvewx %v15, %r9, %r4 \n" + " vmhraddshs %v25, %v18, %v12, %v6 \n" + " add %r4, %r4, %r5 \n" + " vsrah %v24, %v13, %v21 \n" + " lvx %v15, 0, %r4 \n" + " vperm %v15, %v15, %v1, %v19 \n" + " vsrah %v23, %v25, %v21 \n" + " vaddshs %v0, %v15, %v27 \n" + " vpkshus %v15, %v0, %v0 \n" + " stvewx %v15, 0, %r4 \n" + " stvewx %v15, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " lvx %v15, 0, %r4 \n" + " vperm %v15, %v15, %v1, %v3 \n" + " vaddshs %v0, %v15, %v22 \n" + " vpkshus %v15, %v0, %v0 \n" + " stvewx %v15, 0, %r4 \n" + " stvewx %v15, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " lvx %v15, 0, %r4 \n" + " vperm %v15, %v15, %v1, %v19 \n" + " vaddshs %v0, %v15, %v30 \n" + " vpkshus %v15, %v0, %v0 \n" + " stvewx %v15, 0, %r4 \n" + " stvewx %v15, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " lvx %v15, 0, %r4 \n" + " vperm %v15, %v15, %v1, %v3 \n" + " vaddshs %v0, %v15, %v26 \n" + " vpkshus %v15, %v0, %v0 \n" + " stvewx %v15, 0, %r4 \n" + " stvewx %v15, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " lvx %v15, 0, %r4 \n" + " vperm %v15, %v15, %v1, %v19 \n" + " vaddshs %v0, %v15, %v23 \n" + " vpkshus %v15, %v0, %v0 \n" + " stvewx %v15, 0, %r4 \n" + " stvewx %v15, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " lvx %v15, 0, %r4 \n" + " vperm %v15, %v15, %v1, %v3 \n" + " vaddshs %v0, %v15, %v24 \n" + " vpkshus %v15, %v0, %v0 \n" + " stvewx %v15, 0, %r4 \n" + " stvewx %v15, %r9, %r4 \n" + + "# addi %r0, %r1, 192 \n" + "# bl _restv21 \n" + "# lwz %r0, 196(%r1) \n" + "# mtlr %r0 \n" + "# la %r1, 192(%r1) \n" + ); +} + +void idct_altivec_init (void) +{ + extern uint8_t scan_norm[64]; + extern uint8_t scan_alt[64]; + int i, j; + + i = constants[0][0]; /* just pretending - keeps gcc happy */ + + /* the altivec idct uses a transposed input, so we patch scan tables */ + for (i = 0; i < 64; i++) { + j = scan_norm[i]; + scan_norm[i] = (j >> 3) | ((j & 7) << 3); + j = scan_alt[i]; + scan_alt[i] = (j >> 3) | ((j & 7) << 3); + } +} + +#endif /* ENABLE_ALTIVEC */ +#endif /* ARCH_PPC */ + +#else /* __ALTIVEC__ */ + +#define vector_s16_t vector signed short +#define vector_u16_t vector unsigned short +#define vector_s8_t vector signed char +#define vector_u8_t vector unsigned char +#define vector_s32_t vector signed int +#define vector_u32_t vector unsigned int + +#define IDCT_HALF \ + /* 1st stage */ \ + t1 = vec_mradds (a1, vx7, vx1 ); \ + t8 = vec_mradds (a1, vx1, vec_subs (zero, vx7)); \ + t7 = vec_mradds (a2, vx5, vx3); \ + t3 = vec_mradds (ma2, vx3, vx5); \ + \ + /* 2nd stage */ \ + t5 = vec_adds (vx0, vx4); \ + t0 = vec_subs (vx0, vx4); \ + t2 = vec_mradds (a0, vx6, vx2); \ + t4 = vec_mradds (a0, vx2, vec_subs (zero,vx6)); \ + t6 = vec_adds (t8, t3); \ + t3 = vec_subs (t8, t3); \ + t8 = vec_subs (t1, t7); \ + t1 = vec_adds (t1, t7); \ + \ + /* 3rd stage */ \ + t7 = vec_adds (t5, t2); \ + t2 = vec_subs (t5, t2); \ + t5 = vec_adds (t0, t4); \ + t0 = vec_subs (t0, t4); \ + t4 = vec_subs (t8, t3); \ + t3 = vec_adds (t8, t3); \ + \ + /* 4th stage */ \ + vy0 = vec_adds (t7, t1); \ + vy7 = vec_subs (t7, t1); \ + vy1 = vec_mradds (c4, t3, t5); \ + vy6 = vec_mradds (mc4, t3, t5); \ + vy2 = vec_mradds (c4, t4, t0); \ + vy5 = vec_mradds (mc4, t4, t0); \ + vy3 = vec_adds (t2, t6); \ + vy4 = vec_subs (t2, t6); + +#define IDCT \ + vector_s16_t vx0, vx1, vx2, vx3, vx4, vx5, vx6, vx7; \ + vector_s16_t vy0, vy1, vy2, vy3, vy4, vy5, vy6, vy7; \ + vector_s16_t a0, a1, a2, ma2, c4, mc4, zero, bias; \ + vector_s16_t t0, t1, t2, t3, t4, t5, t6, t7, t8; \ + vector_u16_t shift; \ + \ + c4 = vec_splat (constants[0], 0); \ + a0 = vec_splat (constants[0], 1); \ + a1 = vec_splat (constants[0], 2); \ + a2 = vec_splat (constants[0], 3); \ + mc4 = vec_splat (constants[0], 4); \ + ma2 = vec_splat (constants[0], 5); \ + bias = (vector_s16_t)vec_splat ((vector_s32_t)constants[0], 3); \ + \ + zero = vec_splat_s16 (0); \ + shift = vec_splat_u16 (4); \ + \ + vx0 = vec_mradds (vec_sl (block[0], shift), constants[1], zero); \ + vx1 = vec_mradds (vec_sl (block[1], shift), constants[2], zero); \ + vx2 = vec_mradds (vec_sl (block[2], shift), constants[3], zero); \ + vx3 = vec_mradds (vec_sl (block[3], shift), constants[4], zero); \ + vx4 = vec_mradds (vec_sl (block[4], shift), constants[1], zero); \ + vx5 = vec_mradds (vec_sl (block[5], shift), constants[4], zero); \ + vx6 = vec_mradds (vec_sl (block[6], shift), constants[3], zero); \ + vx7 = vec_mradds (vec_sl (block[7], shift), constants[2], zero); \ + \ + IDCT_HALF \ + \ + vx0 = vec_mergeh (vy0, vy4); \ + vx1 = vec_mergel (vy0, vy4); \ + vx2 = vec_mergeh (vy1, vy5); \ + vx3 = vec_mergel (vy1, vy5); \ + vx4 = vec_mergeh (vy2, vy6); \ + vx5 = vec_mergel (vy2, vy6); \ + vx6 = vec_mergeh (vy3, vy7); \ + vx7 = vec_mergel (vy3, vy7); \ + \ + vy0 = vec_mergeh (vx0, vx4); \ + vy1 = vec_mergel (vx0, vx4); \ + vy2 = vec_mergeh (vx1, vx5); \ + vy3 = vec_mergel (vx1, vx5); \ + vy4 = vec_mergeh (vx2, vx6); \ + vy5 = vec_mergel (vx2, vx6); \ + vy6 = vec_mergeh (vx3, vx7); \ + vy7 = vec_mergel (vx3, vx7); \ + \ + vx0 = vec_adds (vec_mergeh (vy0, vy4), bias); \ + vx1 = vec_mergel (vy0, vy4); \ + vx2 = vec_mergeh (vy1, vy5); \ + vx3 = vec_mergel (vy1, vy5); \ + vx4 = vec_mergeh (vy2, vy6); \ + vx5 = vec_mergel (vy2, vy6); \ + vx6 = vec_mergeh (vy3, vy7); \ + vx7 = vec_mergel (vy3, vy7); \ + \ + IDCT_HALF \ + \ + shift = vec_splat_u16 (6); \ + vx0 = vec_sra (vy0, shift); \ + vx1 = vec_sra (vy1, shift); \ + vx2 = vec_sra (vy2, shift); \ + vx3 = vec_sra (vy3, shift); \ + vx4 = vec_sra (vy4, shift); \ + vx5 = vec_sra (vy5, shift); \ + vx6 = vec_sra (vy6, shift); \ + vx7 = vec_sra (vy7, shift); + +static vector_s16_t constants[5] = { + (vector_s16_t)(23170, 13573, 6518, 21895, -23170, -21895, 32, 31), + (vector_s16_t)(16384, 22725, 21407, 19266, 16384, 19266, 21407, 22725), + (vector_s16_t)(22725, 31521, 29692, 26722, 22725, 26722, 29692, 31521), + (vector_s16_t)(21407, 29692, 27969, 25172, 21407, 25172, 27969, 29692), + (vector_s16_t)(19266, 26722, 25172, 22654, 19266, 22654, 25172, 26722) +}; + +void idct_block_copy_altivec (vector_s16_t * block, unsigned char * dest, + int stride) +{ + vector_u8_t tmp; + + IDCT + +#define COPY(dest,src) \ + tmp = vec_packsu (src, src); \ + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \ + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + + COPY (dest, vx0) dest += stride; + COPY (dest, vx1) dest += stride; + COPY (dest, vx2) dest += stride; + COPY (dest, vx3) dest += stride; + COPY (dest, vx4) dest += stride; + COPY (dest, vx5) dest += stride; + COPY (dest, vx6) dest += stride; + COPY (dest, vx7) +} + +void idct_block_add_altivec (vector_s16_t * block, unsigned char * dest, + int stride) +{ + vector_u8_t tmp; + vector_s16_t tmp2, tmp3; + vector_u8_t perm0; + vector_u8_t perm1; + vector_u8_t p0, p1, p; + + IDCT + + p0 = vec_lvsl (0, dest); + p1 = vec_lvsl (stride, dest); + p = vec_splat_u8 (-1); + perm0 = vec_mergeh (p, p0); + perm1 = vec_mergeh (p, p1); + +#define ADD(dest,src,perm) \ + /* *(uint64_t *)&tmp = *(uint64_t *)dest; */ \ + tmp = vec_ld (0, dest); \ + tmp2 = (vector_s16_t)vec_perm (tmp, (vector_u8_t)zero, perm); \ + tmp3 = vec_adds (tmp2, src); \ + tmp = vec_packsu (tmp3, tmp3); \ + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); \ + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + + ADD (dest, vx0, perm0) dest += stride; + ADD (dest, vx1, perm1) dest += stride; + ADD (dest, vx2, perm0) dest += stride; + ADD (dest, vx3, perm1) dest += stride; + ADD (dest, vx4, perm0) dest += stride; + ADD (dest, vx5, perm1) dest += stride; + ADD (dest, vx6, perm0) dest += stride; + ADD (dest, vx7, perm1) +} + +#endif /* __ALTIVEC__ */ diff --git a/src/libmpeg2/motion_comp.c b/src/libmpeg2/motion_comp.c index b8f308525..9ea10c532 100644 --- a/src/libmpeg2/motion_comp.c +++ b/src/libmpeg2/motion_comp.c @@ -50,6 +50,12 @@ void motion_comp_init (void) mc_functions = mc_functions_mlib; } else #endif +#ifdef ENABLE_ALTIVEC + if (config.flags & MM_ACCEL_PPC_ALTIVEC) { + fprintf (stderr, "Using altivec for motion compensation\n"); + mc_functions = mc_functions_altivec; + } else +#endif { fprintf (stderr, "No accelerated motion compensation found\n"); mc_functions = mc_functions_c; diff --git a/src/libmpeg2/motion_comp_altivec.c b/src/libmpeg2/motion_comp_altivec.c new file mode 100644 index 000000000..e4b3a234c --- /dev/null +++ b/src/libmpeg2/motion_comp_altivec.c @@ -0,0 +1,2024 @@ +/* + * motion_comp_altivec.c + * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org> + * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca> + * + * This file is part of mpeg2dec, a free MPEG-2 video stream decoder. + * See http://libmpeg2.sourceforge.net/ for updates. + * + * mpeg2dec is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * mpeg2dec 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + */ + +#ifndef __ALTIVEC__ + +#include "config.h" + +#ifdef ARCH_PPC +#ifdef ENABLE_ALTIVEC + +#include <inttypes.h> + +#include "mpeg2_internal.h" + +/* + * The asm code is generated with: + * + * gcc-2.95 -fvec -D__ALTIVEC__ -O9 -fomit-frame-pointer -mregnames -S + * motion_comp_altivec.c + * + * sed 's/.L/._L/g' motion_comp_altivec.s | + * awk '{args=""; len=split ($2, arg, ","); + * for (i=1; i<=len; i++) { a=arg[i]; if (i<len) a=a","; + * args = args sprintf ("%-6s", a) } + * printf ("\t\"\t%-16s%-24s\\n\"\n", $1, args) }' | + * unexpand -a + */ + +static void MC_put_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " srawi %r6, %r6, 1 \n" + " li %r9, 15 \n" + " addi %r6, %r6, -1 \n" + " lvsl %v12, 0, %r4 \n" + " mtctr %r6 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " add %r0, %r5, %r5 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + "._L6: \n" + " li %r9, 15 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " stvx %v13, %r5, %r3 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r0 \n" + " bdnz ._L6 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v13, %v1, %v0, %v12 \n" + " stvx %v13, %r5, %r3 \n" + ); +} + +static void MC_put_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v12, 0, %r4 \n" + " lvsl %v1, %r5, %r4 \n" + " vmrghb %v12, %v12, %v12 \n" + " srawi %r6, %r6, 1 \n" + " li %r9, 7 \n" + " vmrghb %v1, %v1, %v1 \n" + " addi %r6, %r6, -1 \n" + " vpkuhum %v10, %v12, %v12 \n" + " lvx %v13, 0, %r4 \n" + " mtctr %r6 \n" + " vpkuhum %v11, %v1, %v1 \n" + " lvx %v0, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v12, %v13, %v0, %v10 \n" + "._L11: \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " stvewx %v12, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v13, %v0, %v11 \n" + " stvewx %v12, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v1, 0, %r3 \n" + " vperm %v12, %v13, %v0, %v10 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L11 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " stvewx %v12, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v13, %v0, %v11 \n" + " stvewx %v12, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v1, 0, %r3 \n" + " stvewx %v1, %r9, %r3 \n" + ); +} + +static void MC_put_x16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v11, 0, %r4 \n" + " vspltisb %v0, 1 \n" + " li %r9, 16 \n" + " lvx %v12, 0, %r4 \n" + " vaddubm %v10, %v11, %v0 \n" + " lvx %v13, %r9, %r4 \n" + " srawi %r6, %r6, 1 \n" + " addi %r6, %r6, -1 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " mtctr %r6 \n" + " add %r0, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + "._L16: \n" + " li %r9, 16 \n" + " lvx %v12, 0, %r4 \n" + " lvx %v13, %r9, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " lvx %v12, 0, %r4 \n" + " lvx %v13, %r9, %r4 \n" + " vavgub %v0, %v0, %v1 \n" + " stvx %v0, %r5, %r3 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " add %r3, %r3, %r0 \n" + " vavgub %v0, %v0, %v1 \n" + " bdnz ._L16 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v1, %v12, %v13, %v10 \n" + " vperm %v0, %v12, %v13, %v11 \n" + " vavgub %v0, %v0, %v1 \n" + " stvx %v0, %r5, %r3 \n" + ); +} + +static void MC_put_x8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v0, 0, %r4 \n" + " vspltisb %v13, 1 \n" + " lvsl %v10, %r5, %r4 \n" + " vmrghb %v0, %v0, %v0 \n" + " li %r9, 8 \n" + " lvx %v11, 0, %r4 \n" + " vmrghb %v10, %v10, %v10 \n" + " vpkuhum %v8, %v0, %v0 \n" + " lvx %v12, %r9, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vpkuhum %v9, %v10, %v10 \n" + " vaddubm %v7, %v8, %v13 \n" + " addi %r6, %r6, -1 \n" + " vperm %v1, %v11, %v12, %v8 \n" + " mtctr %r6 \n" + " vaddubm %v13, %v9, %v13 \n" + " add %r4, %r4, %r5 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " vavgub %v0, %v1, %v0 \n" + "._L21: \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " stvewx %v0, %r9, %r3 \n" + " vperm %v0, %v11, %v12, %v9 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " vavgub %v10, %v0, %v1 \n" + " lvx %v11, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v10, 0, %r3 \n" + " vperm %v1, %v11, %v12, %v7 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " li %r9, 4 \n" + " stvewx %v10, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L21 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " stvewx %v0, %r9, %r3 \n" + " vperm %v0, %v11, %v12, %v9 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v10, %v0, %v1 \n" + " stvewx %v10, 0, %r3 \n" + " stvewx %v10, %r9, %r3 \n" + ); +} + +static void MC_put_y16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " li %r9, 15 \n" + " lvsl %v10, 0, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v12, %v13, %v1, %v10 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " addi %r6, %r6, -1 \n" + " vperm %v11, %v13, %v1, %v10 \n" + " mtctr %r6 \n" + " add %r0, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v12, %v11 \n" + "._L26: \n" + " li %r9, 15 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v12, %v13, %v1, %v10 \n" + " add %r4, %r4, %r5 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " vavgub %v0, %v12, %v11 \n" + " stvx %v0, %r5, %r3 \n" + " vperm %v11, %v13, %v1, %v10 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r0 \n" + " vavgub %v0, %v12, %v11 \n" + " bdnz ._L26 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " stvx %v0, 0, %r3 \n" + " vperm %v12, %v13, %v1, %v10 \n" + " vavgub %v0, %v12, %v11 \n" + " stvx %v0, %r5, %r3 \n" + ); +} + +static void MC_put_y8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v13, 0, %r4 \n" + " lvsl %v11, %r5, %r4 \n" + " vmrghb %v13, %v13, %v13 \n" + " li %r9, 7 \n" + " lvx %v12, 0, %r4 \n" + " vmrghb %v11, %v11, %v11 \n" + " lvx %v1, %r9, %r4 \n" + " vpkuhum %v9, %v13, %v13 \n" + " add %r4, %r4, %r5 \n" + " vpkuhum %v10, %v11, %v11 \n" + " vperm %v13, %v12, %v1, %v9 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v12, 0, %r4 \n" + " lvx %v1, %r9, %r4 \n" + " addi %r6, %r6, -1 \n" + " vperm %v11, %v12, %v1, %v10 \n" + " mtctr %r6 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v13, %v11 \n" + "._L31: \n" + " li %r9, 7 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v13, %v12, %v1, %v9 \n" + " stvewx %v0, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v13, %v11 \n" + " li %r9, 7 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v0, 0, %r3 \n" + " vperm %v11, %v12, %v1, %v10 \n" + " li %r9, 4 \n" + " stvewx %v0, %r9, %r3 \n" + " vavgub %v0, %v13, %v11 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L31 \n" + " li %r9, 7 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v12, 0, %r4 \n" + " stvewx %v0, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v13, %v12, %v1, %v9 \n" + " stvewx %v0, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v0, %v13, %v11 \n" + " stvewx %v0, 0, %r3 \n" + " stvewx %v0, %r9, %r3 \n" + ); +} + +static void MC_put_xy16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v5, 0, %r4 \n" + " vspltisb %v3, 1 \n" + " li %r9, 16 \n" + " lvx %v1, 0, %r4 \n" + " vaddubm %v4, %v5, %v3 \n" + " lvx %v0, %r9, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " srawi %r6, %r6, 1 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " addi %r6, %r6, -1 \n" + " lvx %v1, 0, %r4 \n" + " mtctr %r6 \n" + " lvx %v0, %r9, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r0, %r5, %r5 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vxor %v6, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + "._L36: \n" + " li %r9, 16 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " vavgub %v12, %v9, %v7 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vor %v13, %v8, %v6 \n" + " vxor %v0, %v9, %v7 \n" + " vxor %v6, %v11, %v10 \n" + " vand %v13, %v3, %v13 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v6 \n" + " vand %v13, %v13, %v0 \n" + " vxor %v0, %v9, %v7 \n" + " vand %v1, %v3, %v1 \n" + " vsububm %v13, %v12, %v13 \n" + " vand %v1, %v1, %v0 \n" + " stvx %v13, %r5, %r3 \n" + " vavgub %v0, %v9, %v7 \n" + " add %r3, %r3, %r0 \n" + " vsububm %v13, %v0, %v1 \n" + " bdnz ._L36 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvx %v13, 0, %r3 \n" + " vperm %v10, %v1, %v0, %v4 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vxor %v8, %v11, %v10 \n" + " vavgub %v9, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + " stvx %v13, %r5, %r3 \n" + ); +} + +static void MC_put_xy8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v4, 0, %r4 \n" + " vspltisb %v3, 1 \n" + " lvsl %v5, %r5, %r4 \n" + " vmrghb %v4, %v4, %v4 \n" + " li %r9, 16 \n" + " vmrghb %v5, %v5, %v5 \n" + " lvx %v1, 0, %r4 \n" + " vpkuhum %v4, %v4, %v4 \n" + " lvx %v0, %r9, %r4 \n" + " vpkuhum %v5, %v5, %v5 \n" + " add %r4, %r4, %r5 \n" + " vaddubm %v2, %v4, %v3 \n" + " vperm %v11, %v1, %v0, %v4 \n" + " srawi %r6, %r6, 1 \n" + " vaddubm %v19, %v5, %v3 \n" + " addi %r6, %r6, -1 \n" + " vperm %v10, %v1, %v0, %v2 \n" + " mtctr %r6 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r9, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v1, %v0, %v19 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vxor %v6, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + "._L41: \n" + " li %r9, 16 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvewx %v13, 0, %r3 \n" + " li %r9, 4 \n" + " vperm %v10, %v1, %v0, %v2 \n" + " stvewx %v13, %r9, %r3 \n" + " vperm %v11, %v1, %v0, %v4 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 16 \n" + " vavgub %v9, %v11, %v10 \n" + " lvx %v0, %r9, %r4 \n" + " vxor %v8, %v11, %v10 \n" + " lvx %v1, 0, %r4 \n" + " vavgub %v12, %v9, %v7 \n" + " vor %v13, %v8, %v6 \n" + " add %r3, %r3, %r5 \n" + " vperm %v10, %v1, %v0, %v19 \n" + " li %r9, 4 \n" + " vperm %v11, %v1, %v0, %v5 \n" + " vand %v13, %v3, %v13 \n" + " add %r4, %r4, %r5 \n" + " vxor %v0, %v9, %v7 \n" + " vxor %v6, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v6 \n" + " vand %v13, %v13, %v0 \n" + " vxor %v0, %v9, %v7 \n" + " vand %v1, %v3, %v1 \n" + " vsububm %v13, %v12, %v13 \n" + " vand %v1, %v1, %v0 \n" + " stvewx %v13, 0, %r3 \n" + " vavgub %v0, %v9, %v7 \n" + " stvewx %v13, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vsububm %v13, %v0, %v1 \n" + " bdnz ._L41 \n" + " li %r9, 16 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v1, 0, %r4 \n" + " stvewx %v13, 0, %r3 \n" + " vperm %v10, %v1, %v0, %v2 \n" + " li %r9, 4 \n" + " vperm %v11, %v1, %v0, %v4 \n" + " stvewx %v13, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vxor %v8, %v11, %v10 \n" + " vavgub %v9, %v11, %v10 \n" + " vor %v0, %v8, %v6 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v3, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v13, %v1, %v0 \n" + " stvewx %v13, 0, %r3 \n" + " stvewx %v13, %r9, %r3 \n" + ); +} + +static void MC_avg_16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " li %r9, 15 \n" + " lvx %v0, %r9, %r4 \n" + " lvsl %v11, 0, %r4 \n" + " lvx %v1, 0, %r4 \n" + " srawi %r6, %r6, 1 \n" + " addi %r6, %r6, -1 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " lvx %v13, 0, %r3 \n" + " mtctr %r6 \n" + " add %r9, %r5, %r5 \n" + " vavgub %v12, %v13, %v0 \n" + " add %r4, %r4, %r5 \n" + "._L46: \n" + " li %r11, 15 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r11, %r4 \n" + " lvx %v13, %r5, %r3 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " stvx %v12, 0, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v12, %v13, %v0 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v0, %r11, %r4 \n" + " lvx %v13, %r9, %r3 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " stvx %v12, %r5, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v12, %v13, %v0 \n" + " add %r3, %r3, %r9 \n" + " bdnz ._L46 \n" + " lvx %v0, %r11, %r4 \n" + " lvx %v1, 0, %r4 \n" + " lvx %v13, %r5, %r3 \n" + " vperm %v0, %v1, %v0, %v11 \n" + " stvx %v12, 0, %r3 \n" + " vavgub %v12, %v13, %v0 \n" + " stvx %v12, %r5, %r3 \n" + ); +} + +static void MC_avg_8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v12, 0, %r4 \n" + " li %r9, 7 \n" + " vmrghb %v12, %v12, %v12 \n" + " lvsl %v1, %r5, %r4 \n" + " lvx %v13, 0, %r4 \n" + " vpkuhum %v9, %v12, %v12 \n" + " lvx %v0, %r9, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vmrghb %v1, %v1, %v1 \n" + " addi %r6, %r6, -1 \n" + " vperm %v0, %v13, %v0, %v9 \n" + " lvx %v11, 0, %r3 \n" + " mtctr %r6 \n" + " vpkuhum %v10, %v1, %v1 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v12, %v11, %v0 \n" + "._L51: \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v11, %r5, %r3 \n" + " stvewx %v12, 0, %r3 \n" + " vperm %v0, %v13, %v0, %v10 \n" + " li %r9, 4 \n" + " stvewx %v12, %r9, %r3 \n" + " vavgub %v1, %v11, %v0 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " add %r3, %r3, %r5 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v11, %r5, %r3 \n" + " stvewx %v1, 0, %r3 \n" + " vperm %v0, %v13, %v0, %v9 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " vavgub %v12, %v11, %v0 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L51 \n" + " li %r9, 7 \n" + " lvx %v0, %r9, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v11, %r5, %r3 \n" + " stvewx %v12, 0, %r3 \n" + " vperm %v0, %v13, %v0, %v10 \n" + " li %r9, 4 \n" + " stvewx %v12, %r9, %r3 \n" + " vavgub %v1, %v11, %v0 \n" + " add %r3, %r3, %r5 \n" + " stvewx %v1, 0, %r3 \n" + " stvewx %v1, %r9, %r3 \n" + ); +} + +static void MC_avg_x16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v8, 0, %r4 \n" + " vspltisb %v0, 1 \n" + " li %r9, 16 \n" + " lvx %v12, %r9, %r4 \n" + " vaddubm %v7, %v8, %v0 \n" + " lvx %v11, 0, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vperm %v1, %v11, %v12, %v7 \n" + " addi %r6, %r6, -1 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " lvx %v9, 0, %r3 \n" + " mtctr %r6 \n" + " add %r9, %r5, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v10, %v9, %v0 \n" + "._L56: \n" + " li %r11, 16 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v12, %r11, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " stvx %v10, 0, %r3 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " add %r4, %r4, %r5 \n" + " vperm %v1, %v11, %v12, %v8 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v12, %r11, %r4 \n" + " vavgub %v1, %v1, %v0 \n" + " add %r4, %r4, %r5 \n" + " vperm %v13, %v11, %v12, %v7 \n" + " vavgub %v10, %v9, %v1 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " lvx %v9, %r9, %r3 \n" + " stvx %v10, %r5, %r3 \n" + " vavgub %v0, %v0, %v13 \n" + " add %r3, %r3, %r9 \n" + " vavgub %v10, %v9, %v0 \n" + " bdnz ._L56 \n" + " lvx %v12, %r11, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " vperm %v1, %v11, %v12, %v7 \n" + " stvx %v10, 0, %r3 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " vavgub %v0, %v0, %v1 \n" + " vavgub %v10, %v9, %v0 \n" + " stvx %v10, %r5, %r3 \n" + ); +} + +static void MC_avg_x8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v10, 0, %r4 \n" + " vspltisb %v13, 1 \n" + " li %r9, 8 \n" + " vmrghb %v10, %v10, %v10 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v12, %r9, %r4 \n" + " vpkuhum %v7, %v10, %v10 \n" + " srawi %r6, %r6, 1 \n" + " lvsl %v10, %r5, %r4 \n" + " vaddubm %v6, %v7, %v13 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " addi %r6, %r6, -1 \n" + " vmrghb %v10, %v10, %v10 \n" + " lvx %v9, 0, %r3 \n" + " mtctr %r6 \n" + " vperm %v1, %v11, %v12, %v6 \n" + " add %r4, %r4, %r5 \n" + " vpkuhum %v8, %v10, %v10 \n" + " vavgub %v0, %v0, %v1 \n" + " vaddubm %v13, %v8, %v13 \n" + " vavgub %v10, %v9, %v0 \n" + "._L61: \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " stvewx %v10, 0, %r3 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " li %r9, 4 \n" + " stvewx %v10, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v0, %v1 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " vavgub %v10, %v9, %v0 \n" + " lvx %v11, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " vperm %v1, %v11, %v12, %v6 \n" + " lvx %v9, %r5, %r3 \n" + " vperm %v0, %v11, %v12, %v7 \n" + " stvewx %v10, 0, %r3 \n" + " li %r9, 4 \n" + " vavgub %v0, %v0, %v1 \n" + " stvewx %v10, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v10, %v9, %v0 \n" + " bdnz ._L61 \n" + " li %r9, 8 \n" + " lvx %v12, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v9, %r5, %r3 \n" + " vperm %v1, %v11, %v12, %v13 \n" + " stvewx %v10, 0, %r3 \n" + " vperm %v0, %v11, %v12, %v8 \n" + " li %r9, 4 \n" + " stvewx %v10, %r9, %r3 \n" + " vavgub %v0, %v0, %v1 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v10, %v9, %v0 \n" + " stvewx %v10, 0, %r3 \n" + " stvewx %v10, %r9, %r3 \n" + ); +} + +static void MC_avg_y16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " li %r9, 15 \n" + " lvx %v1, %r9, %r4 \n" + " lvsl %v9, 0, %r4 \n" + " lvx %v13, 0, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v11, %v13, %v1, %v9 \n" + " li %r11, 15 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " srawi %r6, %r6, 1 \n" + " vperm %v10, %v13, %v1, %v9 \n" + " addi %r6, %r6, -1 \n" + " lvx %v12, 0, %r3 \n" + " mtctr %r6 \n" + " vavgub %v0, %v11, %v10 \n" + " add %r9, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v12, %v0 \n" + "._L66: \n" + " li %r11, 15 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v12, %r5, %r3 \n" + " vperm %v11, %v13, %v1, %v9 \n" + " stvx %v0, 0, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v11, %v10 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " vavgub %v0, %v12, %v0 \n" + " add %r4, %r4, %r5 \n" + " lvx %v12, %r9, %r3 \n" + " vperm %v10, %v13, %v1, %v9 \n" + " stvx %v0, %r5, %r3 \n" + " vavgub %v0, %v11, %v10 \n" + " add %r3, %r3, %r9 \n" + " vavgub %v0, %v12, %v0 \n" + " bdnz ._L66 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v12, %r5, %r3 \n" + " vperm %v11, %v13, %v1, %v9 \n" + " stvx %v0, 0, %r3 \n" + " vavgub %v0, %v11, %v10 \n" + " vavgub %v0, %v12, %v0 \n" + " stvx %v0, %r5, %r3 \n" + ); +} + +static void MC_avg_y8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v12, 0, %r4 \n" + " lvsl %v9, %r5, %r4 \n" + " vmrghb %v12, %v12, %v12 \n" + " li %r9, 7 \n" + " lvx %v11, 0, %r4 \n" + " vmrghb %v9, %v9, %v9 \n" + " lvx %v13, %r9, %r4 \n" + " vpkuhum %v7, %v12, %v12 \n" + " add %r4, %r4, %r5 \n" + " vpkuhum %v8, %v9, %v9 \n" + " vperm %v12, %v11, %v13, %v7 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v13, %r9, %r4 \n" + " addi %r6, %r6, -1 \n" + " vperm %v9, %v11, %v13, %v8 \n" + " lvx %v10, 0, %r3 \n" + " mtctr %r6 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v0, %v12, %v9 \n" + " vavgub %v1, %v10, %v0 \n" + "._L71: \n" + " li %r9, 7 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v10, %r5, %r3 \n" + " stvewx %v1, 0, %r3 \n" + " vperm %v12, %v11, %v13, %v7 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " vavgub %v0, %v12, %v9 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 7 \n" + " vavgub %v1, %v10, %v0 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " add %r3, %r3, %r5 \n" + " vperm %v9, %v11, %v13, %v8 \n" + " lvx %v10, %r5, %r3 \n" + " stvewx %v1, 0, %r3 \n" + " vavgub %v0, %v12, %v9 \n" + " li %r9, 4 \n" + " stvewx %v1, %r9, %r3 \n" + " add %r4, %r4, %r5 \n" + " vavgub %v1, %v10, %v0 \n" + " add %r3, %r3, %r5 \n" + " bdnz ._L71 \n" + " li %r9, 7 \n" + " lvx %v13, %r9, %r4 \n" + " lvx %v11, 0, %r4 \n" + " lvx %v10, %r5, %r3 \n" + " vperm %v12, %v11, %v13, %v7 \n" + " stvewx %v1, 0, %r3 \n" + " li %r9, 4 \n" + " vavgub %v0, %v12, %v9 \n" + " stvewx %v1, %r9, %r3 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v1, %v10, %v0 \n" + " stvewx %v1, 0, %r3 \n" + " stvewx %v1, %r9, %r3 \n" + ); +} + +static void MC_avg_xy16_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v4, 0, %r4 \n" + " vspltisb %v2, 1 \n" + " li %r9, 16 \n" + " lvx %v1, %r9, %r4 \n" + " vaddubm %v3, %v4, %v2 \n" + " lvx %v13, 0, %r4 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " li %r11, 16 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " srawi %r6, %r6, 1 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " addi %r6, %r6, -1 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " lvx %v6, 0, %r3 \n" + " mtctr %r6 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " add %r9, %r5, %r5 \n" + " add %r4, %r4, %r5 \n" + " vxor %v5, %v11, %v10 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v5 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v1, %v2, %v1 \n" + " vavgub %v0, %v9, %v7 \n" + " vand %v1, %v1, %v13 \n" + " vsububm %v0, %v0, %v1 \n" + " vavgub %v12, %v6, %v0 \n" + "._L76: \n" + " li %r11, 16 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v6, %r5, %r3 \n" + " stvx %v12, 0, %r3 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " add %r4, %r4, %r5 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v1, %r11, %r4 \n" + " vavgub %v9, %v11, %v10 \n" + " vxor %v8, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " vavgub %v12, %v9, %v7 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " vor %v0, %v8, %v5 \n" + " vxor %v13, %v9, %v7 \n" + " vxor %v5, %v11, %v10 \n" + " vand %v0, %v2, %v0 \n" + " vavgub %v7, %v11, %v10 \n" + " vor %v1, %v8, %v5 \n" + " vand %v0, %v0, %v13 \n" + " vand %v1, %v2, %v1 \n" + " vxor %v13, %v9, %v7 \n" + " vsububm %v12, %v12, %v0 \n" + " vand %v1, %v1, %v13 \n" + " vavgub %v0, %v9, %v7 \n" + " vavgub %v12, %v6, %v12 \n" + " lvx %v6, %r9, %r3 \n" + " vsububm %v0, %v0, %v1 \n" + " stvx %v12, %r5, %r3 \n" + " vavgub %v12, %v6, %v0 \n" + " add %r3, %r3, %r9 \n" + " bdnz ._L76 \n" + " lvx %v1, %r11, %r4 \n" + " lvx %v13, 0, %r4 \n" + " lvx %v6, %r5, %r3 \n" + " vperm %v10, %v13, %v1, %v3 \n" + " stvx %v12, 0, %r3 \n" + " vperm %v11, %v13, %v1, %v4 \n" + " vxor %v8, %v11, %v10 \n" + " vavgub %v9, %v11, %v10 \n" + " vor %v0, %v8, %v5 \n" + " vxor %v13, %v9, %v7 \n" + " vand %v0, %v2, %v0 \n" + " vavgub %v1, %v9, %v7 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v1, %v1, %v0 \n" + " vavgub %v12, %v6, %v1 \n" + " stvx %v12, %r5, %r3 \n" + ); +} + +static void MC_avg_xy8_altivec (uint8_t * dest, uint8_t * ref, + int stride, int height) +{ + asm (" \n" + " lvsl %v2, 0, %r4 \n" + " vspltisb %v19, 1 \n" + " lvsl %v3, %r5, %r4 \n" + " vmrghb %v2, %v2, %v2 \n" + " li %r9, 16 \n" + " vmrghb %v3, %v3, %v3 \n" + " lvx %v9, 0, %r4 \n" + " vpkuhum %v2, %v2, %v2 \n" + " lvx %v1, %r9, %r4 \n" + " vpkuhum %v3, %v3, %v3 \n" + " add %r4, %r4, %r5 \n" + " vaddubm %v18, %v2, %v19 \n" + " vperm %v11, %v9, %v1, %v2 \n" + " srawi %r6, %r6, 1 \n" + " vaddubm %v17, %v3, %v19 \n" + " addi %r6, %r6, -1 \n" + " vperm %v10, %v9, %v1, %v18 \n" + " lvx %v4, 0, %r3 \n" + " mtctr %r6 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v9, 0, %r4 \n" + " vavgub %v8, %v11, %v10 \n" + " vxor %v7, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " vperm %v10, %v9, %v1, %v17 \n" + " vperm %v11, %v9, %v1, %v3 \n" + " vxor %v5, %v11, %v10 \n" + " vavgub %v6, %v11, %v10 \n" + " vor %v1, %v7, %v5 \n" + " vxor %v13, %v8, %v6 \n" + " vand %v1, %v19, %v1 \n" + " vavgub %v0, %v8, %v6 \n" + " vand %v1, %v1, %v13 \n" + " vsububm %v0, %v0, %v1 \n" + " vavgub %v13, %v4, %v0 \n" + "._L81: \n" + " li %r9, 16 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v9, 0, %r4 \n" + " lvx %v4, %r5, %r3 \n" + " stvewx %v13, 0, %r3 \n" + " vperm %v10, %v9, %v1, %v18 \n" + " vperm %v11, %v9, %v1, %v2 \n" + " li %r9, 4 \n" + " stvewx %v13, %r9, %r3 \n" + " vxor %v7, %v11, %v10 \n" + " add %r4, %r4, %r5 \n" + " li %r9, 16 \n" + " vavgub %v8, %v11, %v10 \n" + " lvx %v1, %r9, %r4 \n" + " vor %v0, %v7, %v5 \n" + " lvx %v9, 0, %r4 \n" + " vxor %v12, %v8, %v6 \n" + " vand %v0, %v19, %v0 \n" + " add %r3, %r3, %r5 \n" + " vperm %v10, %v9, %v1, %v17 \n" + " vavgub %v13, %v8, %v6 \n" + " li %r9, 4 \n" + " vperm %v11, %v9, %v1, %v3 \n" + " vand %v0, %v0, %v12 \n" + " add %r4, %r4, %r5 \n" + " vxor %v5, %v11, %v10 \n" + " vavgub %v6, %v11, %v10 \n" + " vor %v1, %v7, %v5 \n" + " vsububm %v13, %v13, %v0 \n" + " vxor %v0, %v8, %v6 \n" + " vand %v1, %v19, %v1 \n" + " vavgub %v13, %v4, %v13 \n" + " vand %v1, %v1, %v0 \n" + " lvx %v4, %r5, %r3 \n" + " vavgub %v0, %v8, %v6 \n" + " stvewx %v13, 0, %r3 \n" + " stvewx %v13, %r9, %r3 \n" + " vsububm %v0, %v0, %v1 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v13, %v4, %v0 \n" + " bdnz ._L81 \n" + " li %r9, 16 \n" + " lvx %v1, %r9, %r4 \n" + " lvx %v9, 0, %r4 \n" + " lvx %v4, %r5, %r3 \n" + " vperm %v10, %v9, %v1, %v18 \n" + " stvewx %v13, 0, %r3 \n" + " vperm %v11, %v9, %v1, %v2 \n" + " li %r9, 4 \n" + " stvewx %v13, %r9, %r3 \n" + " vxor %v7, %v11, %v10 \n" + " add %r3, %r3, %r5 \n" + " vavgub %v8, %v11, %v10 \n" + " vor %v0, %v7, %v5 \n" + " vxor %v13, %v8, %v6 \n" + " vand %v0, %v19, %v0 \n" + " vavgub %v1, %v8, %v6 \n" + " vand %v0, %v0, %v13 \n" + " vsububm %v1, %v1, %v0 \n" + " vavgub %v13, %v4, %v1 \n" + " stvewx %v13, 0, %r3 \n" + " stvewx %v13, %r9, %r3 \n" + ); +} + +MOTION_COMP_EXTERN (altivec) + +#endif /* ENABLE_ALTIVEC */ +#endif /* ARCH_PPC */ + +#else /* __ALTIVEC__ */ + +#define vector_s16_t vector signed short +#define vector_u16_t vector unsigned short +#define vector_s8_t vector signed char +#define vector_u8_t vector unsigned char +#define vector_s32_t vector signed int +#define vector_u32_t vector unsigned int + +void MC_put_o_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp = vec_perm (ref0, ref1, perm); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, 0, dest); + tmp = vec_perm (ref0, ref1, perm); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_perm (ref0, ref1, perm); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + vec_st (tmp, 0, dest); + tmp = vec_perm (ref0, ref1, perm); + vec_st (tmp, stride, dest); +} + +void MC_put_o_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, ref0, ref1; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_put_x_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, tmp; + + permA = vec_lvsl (0, ref); + permB = vec_add (permA, vec_splat_u8 (1)); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, 0, dest); + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + vec_st (tmp, 0, dest); + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + vec_st (tmp, stride, dest); +} + +void MC_put_x_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ones, tmp0, tmp1, ref0, ref1; + + ones = vec_splat_u8 (1); + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0A = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + perm0B = vec_add (perm0A, ones); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1A = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + tmp0 = vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B)); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_put_y_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp0, tmp1, tmp; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (tmp0, tmp1); + vec_st (tmp, stride, dest); +} + +void MC_put_y_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, tmp, ref0, ref1; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (tmp0, tmp1); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (tmp0, tmp1); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (tmp0, tmp1); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (tmp0, tmp1); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +void MC_put_xy_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, A, B, avg0, avg1, xor0, xor1, tmp; + vector_u8_t ones; + + ones = vec_splat_u8 (1); + permA = vec_lvsl (0, ref); + permB = vec_add (permA, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_st (tmp, stride, dest); + dest += 2*stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + vec_st (tmp, stride, dest); +} + +void MC_put_xy_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ref0, ref1, A, B; + vector_u8_t avg0, avg1, xor0, xor1, tmp, ones; + + ones = vec_splat_u8 (1); + perm0A = vec_lvsl (0, ref); + perm0A = vec_mergeh (perm0A, perm0A); + perm0A = vec_pack ((vector_u16_t)perm0A, (vector_u16_t)perm0A); + perm0B = vec_add (perm0A, ones); + perm1A = vec_lvsl (stride, ref); + perm1A = vec_mergeh (perm1A, perm1A); + perm1A = vec_pack ((vector_u16_t)perm1A, (vector_u16_t)perm1A); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1))); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +#if 0 +void MC_put_xy_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, A, B, C, D, tmp, zero, ones; + vector_u16_t splat2, temp; + + ones = vec_splat_u8 (1); + permA = vec_lvsl (0, ref); + permB = vec_add (permA, ones); + + zero = vec_splat_u8 (0); + splat2 = vec_splat_u16 (2); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + C = vec_perm (ref0, ref1, permA); + D = vec_perm (ref0, ref1, permB); + + temp = vec_add (vec_add ((vector_u16_t)vec_mergeh (zero, A), + (vector_u16_t)vec_mergeh (zero, B)), + vec_add ((vector_u16_t)vec_mergeh (zero, C), + (vector_u16_t)vec_mergeh (zero, D))); + temp = vec_sr (vec_add (temp, splat2), splat2); + tmp = vec_pack (temp, temp); + + vec_st (tmp, 0, dest); + dest += stride; + tmp = vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB)); + } while (--height); +} +#endif + +void MC_avg_o_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp, prev; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_perm (ref0, ref1, perm)); + vec_st (tmp, stride, dest); +} + +void MC_avg_o_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, ref0, ref1, prev; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp0 = vec_avg (prev, vec_perm (ref0, ref1, perm0)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_perm (ref0, ref1, perm1)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_avg (prev, vec_perm (ref0, ref1, perm0)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_perm (ref0, ref1, perm1)); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_avg_x_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, tmp, prev; + + permA = vec_lvsl (0, ref); + permB = vec_add (permA, vec_splat_u8 (1)); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (0, dest); + ref += stride; + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, permA), + vec_perm (ref0, ref1, permB))); + vec_st (tmp, stride, dest); +} + +void MC_avg_x_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ones, tmp0, tmp1, ref0, ref1; + vector_u8_t prev; + + ones = vec_splat_u8 (1); + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0A = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + perm0B = vec_add (perm0A, ones); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1A = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + prev = vec_ld (0, dest); + ref += stride; + tmp0 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B))); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm0A), + vec_perm (ref0, ref1, perm0B))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (8, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp0, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp0, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_avg (prev, vec_avg (vec_perm (ref0, ref1, perm1A), + vec_perm (ref0, ref1, perm1B))); + vec_ste ((vector_u32_t)tmp1, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp1, 4, (unsigned int *)dest); +} + +void MC_avg_y_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm, ref0, ref1, tmp0, tmp1, tmp, prev; + + perm = vec_lvsl (0, ref); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + tmp1 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (15, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + tmp0 = vec_perm (ref0, ref1, perm); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + vec_st (tmp, stride, dest); +} + +void MC_avg_y_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0, perm1, tmp0, tmp1, tmp, ref0, ref1, prev; + + tmp0 = vec_lvsl (0, ref); + tmp0 = vec_mergeh (tmp0, tmp0); + perm0 = vec_pack ((vector_u16_t)tmp0, (vector_u16_t)tmp0); + tmp1 = vec_lvsl (stride, ref); + tmp1 = vec_mergeh (tmp1, tmp1); + perm1 = vec_pack ((vector_u16_t)tmp1, (vector_u16_t)tmp1); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (0, dest); + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp1 = vec_perm (ref0, ref1, perm1); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (7, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + tmp0 = vec_perm (ref0, ref1, perm0); + tmp = vec_avg (prev, vec_avg (tmp0, tmp1)); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +void MC_avg_xy_16_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t permA, permB, ref0, ref1, A, B, avg0, avg1, xor0, xor1, tmp; + vector_u8_t ones, prev; + + ones = vec_splat_u8 (1); + permA = vec_lvsl (0, ref); + permB = vec_add (permA, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (2*stride, dest); + vec_st (tmp, stride, dest); + dest += 2*stride; + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (stride, dest); + vec_st (tmp, 0, dest); + A = vec_perm (ref0, ref1, permA); + B = vec_perm (ref0, ref1, permB); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + vec_st (tmp, stride, dest); +} + +void MC_avg_xy_8_altivec (unsigned char * dest, unsigned char * ref, + int stride, int height) +{ + vector_u8_t perm0A, perm0B, perm1A, perm1B, ref0, ref1, A, B; + vector_u8_t avg0, avg1, xor0, xor1, tmp, ones, prev; + + ones = vec_splat_u8 (1); + perm0A = vec_lvsl (0, ref); + perm0A = vec_mergeh (perm0A, perm0A); + perm0A = vec_pack ((vector_u16_t)perm0A, (vector_u16_t)perm0A); + perm0B = vec_add (perm0A, ones); + perm1A = vec_lvsl (stride, ref); + perm1A = vec_mergeh (perm1A, perm1A); + perm1A = vec_pack ((vector_u16_t)perm1A, (vector_u16_t)perm1A); + perm1B = vec_add (perm1A, ones); + + height = (height >> 1) - 1; + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (0, dest); + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + + do { + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + ref += stride; + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm1A); + B = vec_perm (ref0, ref1, perm1B); + avg1 = vec_avg (A, B); + xor1 = vec_xor (A, B); + tmp = vec_avg (prev, + vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + } while (--height); + + ref0 = vec_ld (0, ref); + ref1 = vec_ld (16, ref); + prev = vec_ld (stride, dest); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); + dest += stride; + A = vec_perm (ref0, ref1, perm0A); + B = vec_perm (ref0, ref1, perm0B); + avg0 = vec_avg (A, B); + xor0 = vec_xor (A, B); + tmp = vec_avg (prev, vec_sub (vec_avg (avg0, avg1), + vec_and (vec_and (ones, vec_or (xor0, xor1)), + vec_xor (avg0, avg1)))); + vec_ste ((vector_u32_t)tmp, 0, (unsigned int *)dest); + vec_ste ((vector_u32_t)tmp, 4, (unsigned int *)dest); +} + +#endif /* __ALTIVEC__ */ diff --git a/src/libmpeg2/mpeg2_internal.h b/src/libmpeg2/mpeg2_internal.h index 8e7ce0d11..d9bc71c62 100644 --- a/src/libmpeg2/mpeg2_internal.h +++ b/src/libmpeg2/mpeg2_internal.h @@ -170,6 +170,13 @@ void idct_block_copy_mmx (int16_t *block, uint8_t * dest, int stride); void idct_block_add_mmx (int16_t *block, uint8_t * dest, int stride); void idct_mmx_init (void); +#ifdef ENABLE_ALTIVEC +/* idct_altivec.c */ +void idct_block_copy_altivec (int16_t * block, uint8_t * dest, int stride); +void idct_block_add_altivec (int16_t * block, uint8_t * dest, int stride); +void idct_altivec_init (void); +#endif + /* motion_comp.c */ void motion_comp_init (void); @@ -191,6 +198,9 @@ extern mc_functions_t mc_functions_c; extern mc_functions_t mc_functions_mmx; extern mc_functions_t mc_functions_mmxext; extern mc_functions_t mc_functions_3dnow; +#ifdef ENABLE_ALTIVEC +extern mc_functions_t mc_functions_altivec; +#endif extern mc_functions_t mc_functions_mlib; /* slice.c */ diff --git a/src/xine-utils/cpu_accel.c b/src/xine-utils/cpu_accel.c index 5fa5a45ed..41baa5d5e 100644 --- a/src/xine-utils/cpu_accel.c +++ b/src/xine-utils/cpu_accel.c @@ -121,33 +121,31 @@ static void sigill_handler (int n) { uint32_t xine_mm_accel (void) { #ifdef ARCH_X86 - static int got_accel = 0; static uint32_t accel; - if (!got_accel) { - got_accel = 1; - - accel = x86_accel (); - - /* test OS support for SSE */ - if( accel & MM_ACCEL_X86_SSE ) { - if (setjmp(sigill_return)) { - accel &= ~(MM_ACCEL_X86_SSE|MM_ACCEL_X86_SSE2); - } else { - signal (SIGILL, sigill_handler); - __asm __volatile ("xorps %xmm0, %xmm0"); - signal (SIGILL, SIG_DFL); - } + accel = x86_accel (); + + /* test OS support for SSE */ + if( accel & MM_ACCEL_X86_SSE ) { + if (setjmp(sigill_return)) { + accel &= ~(MM_ACCEL_X86_SSE|MM_ACCEL_X86_SSE2); + } else { + signal (SIGILL, sigill_handler); + __asm __volatile ("xorps %xmm0, %xmm0"); + signal (SIGILL, SIG_DFL); } } return accel; -#else +#endif #ifdef HAVE_MLIB return MM_ACCEL_MLIB; -#else - return 0; +#endif +#ifdef ARCH_PPC +#ifdef ENABLE_ALTIVEC + return MM_ACCEL_PPC_ALTIVEC; #endif #endif + return 0; } diff --git a/src/xine-utils/xineutils.h b/src/xine-utils/xineutils.h index a4a689f4a..634cf0c74 100644 --- a/src/xine-utils/xineutils.h +++ b/src/xine-utils/xineutils.h @@ -17,7 +17,7 @@ * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA * - * $Id: xineutils.h,v 1.4 2001/11/30 00:53:51 f1rmb Exp $ + * $Id: xineutils.h,v 1.5 2001/12/05 15:12:03 guenter Exp $ * */ #ifndef XINEUTILS_H @@ -49,6 +49,8 @@ extern "C" { #define MM_ACCEL_X86_MMXEXT 0x20000000 #define MM_ACCEL_X86_SSE 0x10000000 #define MM_ACCEL_X86_SSE2 0x08000000 +/* powerpc accelerations */ +#define MM_ACCEL_PPC_ALTIVEC 0x04000000 /* x86 compat defines */ #define MM_MMX MM_ACCEL_X86_MMX #define MM_3DNOW MM_ACCEL_X86_3DNOW |