diff options
| -rw-r--r-- | liba52/a52.h | 3 | ||||
| -rw-r--r-- | liba52/liba52_changes.diff | 92 |
2 files changed, 50 insertions, 45 deletions
diff --git a/liba52/a52.h b/liba52/a52.h index 78a92ed16a..3547ad1d36 100644 --- a/liba52/a52.h +++ b/liba52/a52.h @@ -28,6 +28,9 @@ #ifndef A52_H #define A52_H +#include <stdint.h> +#include "mm_accel.h" + #ifndef LIBA52_DOUBLE typedef float sample_t; #else diff --git a/liba52/liba52_changes.diff b/liba52/liba52_changes.diff index 0e9e71c53f..206e84c213 100644 --- a/liba52/liba52_changes.diff +++ b/liba52/liba52_changes.diff @@ -1,6 +1,6 @@ --- include/a52.h 2006-06-12 15:04:57.000000000 +0200 +++ liba52/a52.h 2006-06-05 02:23:02.000000000 +0200 -@@ -63,4 +63,9 @@ +@@ -59,4 +66,9 @@ int a52_block (a52_state_t * state); void a52_free (a52_state_t * state); @@ -12,11 +12,11 @@ #endif /* A52_H */ --- liba52/a52_internal.h 2006-06-12 15:05:07.000000000 +0200 +++ liba52/a52_internal.h 2006-06-05 02:23:02.000000000 +0200 -@@ -107,18 +107,34 @@ +@@ -103,18 +107,34 @@ #define DELTA_BIT_NONE (2) #define DELTA_BIT_RESERVED (3) -+#ifdef ARCH_X86_64 ++#if ARCH_X86_64 +# define REG_a "rax" +# define REG_d "rdx" +# define REG_S "rsi" @@ -52,7 +52,7 @@ +void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias); --- liba52/bitstream.c 2006-06-12 15:05:07.000000000 +0200 +++ liba52/bitstream.c 2006-06-05 02:23:02.000000000 +0200 -@@ -35,6 +35,10 @@ +@@ -31,6 +35,10 @@ #define BUFFER_SIZE 4096 @@ -63,7 +63,7 @@ void a52_bitstream_set_ptr (a52_state_t * state, uint8_t * buf) { int align; -@@ -42,6 +46,9 @@ +@@ -38,6 +46,9 @@ align = (long)buf & 3; state->buffer_start = (uint32_t *) (buf - align); state->bits_left = 0; @@ -75,7 +75,7 @@ --- liba52/bitstream.h 2006-06-12 15:05:07.000000000 +0200 +++ liba52/bitstream.h 2006-06-05 02:23:02.000000000 +0200 -@@ -25,6 +25,42 @@ +@@ -21,6 +25,42 @@ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ @@ -91,7 +91,7 @@ +#define ALT_BITSTREAM_READER + +/* used to avoid misaligned exceptions on some archs (alpha, ...) */ -+#if defined (ARCH_X86) || defined(HAVE_ARMV6) ++#if ARCH_X86 || HAVE_ARMV6 +# define unaligned32(a) (*(uint32_t*)(a)) +#else +# ifdef __GNUC__ @@ -118,7 +118,7 @@ /* (stolen from the kernel) */ #ifdef WORDS_BIGENDIAN -@@ -32,7 +74,7 @@ +@@ -28,7 +68,7 @@ #else @@ -127,7 +127,7 @@ # define swab32(x) __i386_swab32(x) static inline const uint32_t __i386_swab32(uint32_t x) -@@ -43,19 +85,34 @@ +@@ -39,19 +79,34 @@ # else @@ -166,7 +166,7 @@ uint32_t result; if (num_bits < state->bits_left) { -@@ -65,10 +122,29 @@ +@@ -61,10 +116,29 @@ } return a52_bitstream_get_bh (state, num_bits); @@ -196,7 +196,7 @@ int32_t result; if (num_bits < state->bits_left) { -@@ -78,4 +154,5 @@ +@@ -74,4 +148,5 @@ } return a52_bitstream_get_bh_2 (state, num_bits); @@ -204,7 +204,7 @@ } --- liba52/downmix.c 2006-06-12 15:17:53.000000000 +0200 +++ liba52/downmix.c 2006-06-05 02:23:02.000000000 +0200 -@@ -23,18 +23,46 @@ +@@ -19,18 +23,46 @@ * 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 @@ -241,7 +241,7 @@ +{ + a52_upmix= upmix_C; + a52_downmix= downmix_C; -+#if defined(ARCH_X86) || defined(ARCH_X86_64) ++#if ARCH_X86 || ARCH_X86_64 + if(mm_accel & MM_ACCEL_X86_MMX) a52_upmix= upmix_MMX; + if(mm_accel & MM_ACCEL_X86_SSE) a52_downmix= downmix_SSE; + if(mm_accel & MM_ACCEL_X86_3DNOW) a52_downmix= downmix_3dnow; @@ -251,7 +251,7 @@ int a52_downmix_init (int input, int flags, sample_t * level, sample_t clev, sample_t slev) { -@@ -451,7 +479,7 @@ +@@ -447,7 +479,7 @@ samples[i] = 0; } @@ -260,7 +260,7 @@ sample_t clev, sample_t slev) { switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) { -@@ -563,7 +591,7 @@ +@@ -559,7 +591,7 @@ break; case CONVERT (A52_3F2R, A52_2F1R): @@ -269,7 +269,7 @@ move2to1 (samples + 768, samples + 512, bias); break; -@@ -587,12 +615,12 @@ +@@ -583,12 +615,12 @@ break; case CONVERT (A52_3F1R, A52_3F2R): @@ -284,12 +284,12 @@ { switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) { -@@ -657,3 +685,1104 @@ +@@ -653,3 +685,1104 @@ goto mix_31to21; } } + -+#if defined(ARCH_X86) || defined(ARCH_X86_64) ++#if ARCH_X86 || ARCH_X86_64 +static void mix2to1_SSE (sample_t * dest, sample_t * src, sample_t bias) +{ + __asm__ volatile( @@ -1391,7 +1391,7 @@ +#endif // ARCH_X86 || ARCH_X86_64 --- liba52/imdct.c 2008-02-19 00:18:33.000000000 +0100 +++ liba52/imdct.c 2008-02-19 00:16:40.000000000 +0100 -@@ -26,6 +26,11 @@ +@@ -22,6 +26,11 @@ * 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 @@ -1403,7 +1403,7 @@ */ #include "config.h" -@@ -43,12 +48,49 @@ +@@ -39,12 +48,50 @@ #include "a52.h" #include "a52_internal.h" #include "mm_accel.h" @@ -1413,6 +1413,7 @@ + +#ifdef RUNTIME_CPUDETECT +#undef HAVE_AMD3DNOWEXT ++#define HAVE_AMD3DNOWEXT 0 +#endif typedef struct complex_s { @@ -1453,7 +1454,7 @@ static uint8_t fftorder[] = { 0,128, 64,192, 32,160,224, 96, 16,144, 80,208,240,112, 48,176, 8,136, 72,200, 40,168,232,104,248,120, 56,184, 24,152,216, 88, -@@ -60,6 +102,40 @@ +@@ -56,6 +103,40 @@ 6,134, 70,198, 38,166,230,102,246,118, 54,182, 22,150,214, 86 }; @@ -1473,7 +1474,7 @@ +static sample_t __attribute__((aligned(16))) xcos1[128]; +static sample_t __attribute__((aligned(16))) xsin1[128]; + -+#if defined(ARCH_X86) || defined(ARCH_X86_64) ++#if ARCH_X86 || ARCH_X86_64 +// NOTE: SSE needs 16byte alignment or it will segfault +// +static float __attribute__((aligned(16))) sseSinCos1c[256]; @@ -1494,7 +1495,7 @@ /* Root values for IFFT */ static sample_t roots16[3]; static sample_t roots32[7]; -@@ -245,7 +321,7 @@ +@@ -241,7 +322,7 @@ ifft_pass (buf, roots128 - 32, 32); } @@ -1503,11 +1504,11 @@ { int i, k; sample_t t_r, t_i, a_r, a_i, b_r, b_i, w_1, w_2; -@@ -289,6 +365,701 @@ +@@ -285,6 +366,702 @@ } } -+#ifdef HAVE_ALTIVEC ++#if HAVE_ALTIVEC + +#ifdef HAVE_ALTIVEC_H +#include <altivec.h> @@ -1852,10 +1853,9 @@ + +// Stuff below this line is borrowed from libac3 +#include "srfftp.h" -+#if defined(ARCH_X86) || defined(ARCH_X86_64) -+#ifndef HAVE_AMD3DNOW ++#if ARCH_X86 || ARCH_X86_64 ++#undef HAVE_AMD3DNOW +#define HAVE_AMD3DNOW 1 -+#endif +#include "srfftp_3dnow.h" + +const i_cmplx_t x_plus_minus_3dnow __attribute__ ((aligned (8))) = {{ 0x00000000UL, 0x80000000UL }}; @@ -1863,8 +1863,10 @@ +const complex_t HSQRT2_3DNOW __attribute__ ((aligned (8))) = { 0.707106781188, 0.707106781188 }; + +#undef HAVE_AMD3DNOWEXT ++#define HAVE_AMD3DNOWEXT 0 +#include "imdct_3dnow.h" -+#define HAVE_AMD3DNOWEXT ++#undef HAVE_AMD3DNOWEXT ++#define HAVE_AMD3DNOWEXT 1 +#include "imdct_3dnow.h" + +void @@ -2205,7 +2207,7 @@ void a52_imdct_256(sample_t * data, sample_t * delay, sample_t bias) { int i, k; -@@ -368,7 +1145,7 @@ +@@ -364,7 +1141,7 @@ void a52_imdct_init (uint32_t mm_accel) { @@ -2214,7 +2216,7 @@ double sum; /* compute imdct window - kaiser-bessel derived window, alpha = 5.0 */ -@@ -420,6 +1197,99 @@ +@@ -416,6 +1193,99 @@ post2[i].real = cos ((M_PI / 128) * (i + 0.5)); post2[i].imag = sin ((M_PI / 128) * (i + 0.5)); } @@ -2229,7 +2231,7 @@ + w[i][k].imag = sin (-M_PI * k / j); + } + } -+#if defined(ARCH_X86) || defined(ARCH_X86_64) ++#if ARCH_X86 || ARCH_X86_64 + for (i = 0; i < 128; i++) { + sseSinCos1c[2*i+0]= xcos1[i]; + sseSinCos1c[2*i+1]= -xcos1[i]; @@ -2283,7 +2285,7 @@ + ifft128 = ifft128_c; + ifft64 = ifft64_c; + -+#if defined(ARCH_X86) || defined(ARCH_X86_64) ++#if ARCH_X86 || ARCH_X86_64 + if(mm_accel & MM_ACCEL_X86_SSE) + { + fprintf (stderr, "Using SSE optimized IMDCT transform\n"); @@ -2303,7 +2305,7 @@ + } + else +#endif // ARCH_X86 || ARCH_X86_64 -+#ifdef HAVE_ALTIVEC ++#if HAVE_ALTIVEC + if (mm_accel & MM_ACCEL_PPC_ALTIVEC) + { + fprintf(stderr, "Using AltiVec optimized IMDCT transform\n"); @@ -2314,7 +2316,7 @@ #ifdef LIBA52_DJBFFT if (mm_accel & MM_ACCEL_DJBFFT) { -@@ -430,7 +1300,5 @@ +@@ -426,7 +1296,5 @@ #endif { fprintf (stderr, "No accelerated IMDCT transform found\n"); @@ -2324,7 +2326,7 @@ } --- include/mm_accel.h 2006-06-12 15:05:00.000000000 +0200 +++ liba52/mm_accel.h 2006-06-05 02:23:04.000000000 +0200 -@@ -34,7 +34,12 @@ +@@ -30,7 +34,12 @@ /* x86 accelerations */ #define MM_ACCEL_X86_MMX 0x80000000 #define MM_ACCEL_X86_3DNOW 0x40000000 @@ -2339,7 +2341,7 @@ --- liba52/parse.c 2006-12-05 08:08:01.000000000 +0100 +++ liba52/parse.c 2006-12-05 08:08:44.000000000 +0100 -@@ -28,6 +28,7 @@ +@@ -24,6 +28,7 @@ #include "config.h" #include <stdlib.h> @@ -2347,7 +2349,7 @@ #include <string.h> #include <inttypes.h> -@@ -35,13 +36,12 @@ +@@ -31,13 +36,12 @@ #include "a52_internal.h" #include "bitstream.h" #include "tables.h" @@ -2364,7 +2366,7 @@ #endif typedef struct { -@@ -64,7 +64,16 @@ +@@ -60,7 +64,16 @@ if (state == NULL) return NULL; @@ -2381,7 +2383,7 @@ if (state->samples == NULL) { free (state); return NULL; -@@ -78,6 +87,7 @@ +@@ -74,6 +87,7 @@ state->lfsr_state = 1; a52_imdct_init (mm_accel); @@ -2389,7 +2391,7 @@ return state; } -@@ -145,7 +155,7 @@ +@@ -141,7 +155,7 @@ state->acmod = acmod = buf[6] >> 5; a52_bitstream_set_ptr (state, buf + 6); @@ -2398,7 +2400,7 @@ if ((acmod == 2) && (bitstream_get (state, 2) == 2)) /* dsurmod */ acmod = A52_DOLBY; -@@ -176,28 +186,28 @@ +@@ -172,28 +186,28 @@ chaninfo = !acmod; do { @@ -2435,7 +2437,7 @@ } while (addbsil--); } -@@ -684,7 +694,7 @@ +@@ -680,7 +694,7 @@ state->fbw_expbap[i].exp[0], state->fbw_expbap[i].exp + 1)) return 1; @@ -2444,7 +2446,7 @@ } if (lfeexpstr != EXP_REUSE) { do_bit_alloc |= 32; -@@ -759,7 +769,7 @@ +@@ -755,7 +769,7 @@ if (bitstream_get (state, 1)) { /* skiple */ i = bitstream_get (state, 9); /* skipl */ while (i--) @@ -2453,7 +2455,7 @@ } samples = state->samples; -@@ -900,6 +910,10 @@ +@@ -896,6 +910,10 @@ void a52_free (a52_state_t * state) { |