diff options
author | Justin Ruggles <justin.ruggles@gmail.com> | 2012-05-21 16:24:42 -0400 |
---|---|---|
committer | Justin Ruggles <justin.ruggles@gmail.com> | 2012-06-08 13:14:38 -0400 |
commit | 98db4e2a4e35ccc2406004216270ceaa1c6a7d00 (patch) | |
tree | ba767ab77a685b73e5593f36dcfd53edc2cac3a3 /libavutil | |
parent | 94d2b0d2fdfec0882d072ffa089c5bf32d1b8642 (diff) | |
download | ffmpeg-98db4e2a4e35ccc2406004216270ceaa1c6a7d00.tar.gz |
PPC: Move types_altivec.h and util_altivec.h from libavcodec to libavutil
This will allow for easier implementation of Altivec functions in libraries
other than libavcodec.
Diffstat (limited to 'libavutil')
-rw-r--r-- | libavutil/ppc/types_altivec.h | 47 | ||||
-rw-r--r-- | libavutil/ppc/util_altivec.h | 118 |
2 files changed, 165 insertions, 0 deletions
diff --git a/libavutil/ppc/types_altivec.h b/libavutil/ppc/types_altivec.h new file mode 100644 index 0000000000..0a4eaf885f --- /dev/null +++ b/libavutil/ppc/types_altivec.h @@ -0,0 +1,47 @@ +/* + * Copyright (c) 2006 Guillaume Poirier <gpoirier@mplayerhq.hu> + * + * This file is part of Libav. + * + * Libav 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 2.1 of the License, or (at your option) any later version. + * + * Libav 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 Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +#ifndef AVUTIL_PPC_TYPES_ALTIVEC_H +#define AVUTIL_PPC_TYPES_ALTIVEC_H + +/*********************************************************************** + * Vector types + **********************************************************************/ +#define vec_u8 vector unsigned char +#define vec_s8 vector signed char +#define vec_u16 vector unsigned short +#define vec_s16 vector signed short +#define vec_u32 vector unsigned int +#define vec_s32 vector signed int +#define vec_f vector float + +/*********************************************************************** + * Null vector + **********************************************************************/ +#define LOAD_ZERO const vec_u8 zerov = vec_splat_u8( 0 ) + +#define zero_u8v (vec_u8) zerov +#define zero_s8v (vec_s8) zerov +#define zero_u16v (vec_u16) zerov +#define zero_s16v (vec_s16) zerov +#define zero_u32v (vec_u32) zerov +#define zero_s32v (vec_s32) zerov + +#endif /* AVUTIL_PPC_TYPES_ALTIVEC_H */ diff --git a/libavutil/ppc/util_altivec.h b/libavutil/ppc/util_altivec.h new file mode 100644 index 0000000000..bdbf86219d --- /dev/null +++ b/libavutil/ppc/util_altivec.h @@ -0,0 +1,118 @@ +/* + * This file is part of Libav. + * + * Libav 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 2.1 of the License, or (at your option) any later version. + * + * Libav 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 Libav; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/** + * @file + * Contains misc utility macros and inline functions + */ + +#ifndef AVUTIL_PPC_UTIL_ALTIVEC_H +#define AVUTIL_PPC_UTIL_ALTIVEC_H + +#include <stdint.h> + +#include "config.h" + +#if HAVE_ALTIVEC_H +#include <altivec.h> +#endif + +#include "types_altivec.h" + +// used to build registers permutation vectors (vcprm) +// the 's' are for words in the _s_econd vector +#define WORD_0 0x00,0x01,0x02,0x03 +#define WORD_1 0x04,0x05,0x06,0x07 +#define WORD_2 0x08,0x09,0x0a,0x0b +#define WORD_3 0x0c,0x0d,0x0e,0x0f +#define WORD_s0 0x10,0x11,0x12,0x13 +#define WORD_s1 0x14,0x15,0x16,0x17 +#define WORD_s2 0x18,0x19,0x1a,0x1b +#define WORD_s3 0x1c,0x1d,0x1e,0x1f + +#define vcprm(a,b,c,d) (const vector unsigned char){WORD_ ## a, WORD_ ## b, WORD_ ## c, WORD_ ## d} +#define vcii(a,b,c,d) (const vector float){FLOAT_ ## a, FLOAT_ ## b, FLOAT_ ## c, FLOAT_ ## d} + +// vcprmle is used to keep the same index as in the SSE version. +// it's the same as vcprm, with the index inversed +// ('le' is Little Endian) +#define vcprmle(a,b,c,d) vcprm(d,c,b,a) + +// used to build inverse/identity vectors (vcii) +// n is _n_egative, p is _p_ositive +#define FLOAT_n -1. +#define FLOAT_p 1. + + +// Transpose 8x8 matrix of 16-bit elements (in-place) +#define TRANSPOSE8(a,b,c,d,e,f,g,h) \ +do { \ + vector signed short A1, B1, C1, D1, E1, F1, G1, H1; \ + vector signed short A2, B2, C2, D2, E2, F2, G2, H2; \ + \ + A1 = vec_mergeh (a, e); \ + B1 = vec_mergel (a, e); \ + C1 = vec_mergeh (b, f); \ + D1 = vec_mergel (b, f); \ + E1 = vec_mergeh (c, g); \ + F1 = vec_mergel (c, g); \ + G1 = vec_mergeh (d, h); \ + H1 = vec_mergel (d, h); \ + \ + A2 = vec_mergeh (A1, E1); \ + B2 = vec_mergel (A1, E1); \ + C2 = vec_mergeh (B1, F1); \ + D2 = vec_mergel (B1, F1); \ + E2 = vec_mergeh (C1, G1); \ + F2 = vec_mergel (C1, G1); \ + G2 = vec_mergeh (D1, H1); \ + H2 = vec_mergel (D1, H1); \ + \ + a = vec_mergeh (A2, E2); \ + b = vec_mergel (A2, E2); \ + c = vec_mergeh (B2, F2); \ + d = vec_mergel (B2, F2); \ + e = vec_mergeh (C2, G2); \ + f = vec_mergel (C2, G2); \ + g = vec_mergeh (D2, H2); \ + h = vec_mergel (D2, H2); \ +} while (0) + + +/** @brief loads unaligned vector @a *src with offset @a offset + and returns it */ +static inline vector unsigned char unaligned_load(int offset, uint8_t *src) +{ + register vector unsigned char first = vec_ld(offset, src); + register vector unsigned char second = vec_ld(offset+15, src); + register vector unsigned char mask = vec_lvsl(offset, src); + return vec_perm(first, second, mask); +} + +/** + * loads vector known misalignment + * @param perm_vec the align permute vector to combine the two loads from lvsl + */ +static inline vec_u8 load_with_perm_vec(int offset, uint8_t *src, vec_u8 perm_vec) +{ + vec_u8 a = vec_ld(offset, src); + vec_u8 b = vec_ld(offset+15, src); + return vec_perm(a, b, perm_vec); +} + +#endif /* AVUTIL_PPC_UTIL_ALTIVEC_H */ |