Mercurial > mplayer.hg
diff liba52/liba52_changes.diff @ 18720:4bad7f00556e
sync with liba52 0.7.4, patch by Emanuele Giaquinta >emanuele.giaquinta ! gmail * com<
part 1: functional changes
| author | rathann |
|---|---|
| date | Thu, 15 Jun 2006 22:58:06 +0000 |
| parents | 9de84a73f6d0 |
| children | 722ac20fac5f |
line wrap: on
line diff
--- a/liba52/liba52_changes.diff Thu Jun 15 08:43:41 2006 +0000 +++ b/liba52/liba52_changes.diff Thu Jun 15 22:58:06 2006 +0000 @@ -1,71 +1,81 @@ ---- include/a52.h 2005-03-22 19:58:53.000000000 +0100 -+++ a52.h 2004-03-19 01:15:49.000000000 +0100 -@@ -19,6 +25,9 @@ - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA - */ +--- liba52-0.7.4/a52.h 2006-06-12 15:04:57.000000000 +0200 ++++ liba52/a52.h 2006-06-05 02:23:02.000000000 +0200 +@@ -59,4 +63,9 @@ + int a52_block (a52_state_t * state); + void a52_free (a52_state_t * state); -+#ifndef A52_H -+#define A52_H -+ - #ifndef LIBA52_DOUBLE - typedef float sample_t; - #else -@@ -113,3 +122,10 @@ - void a52_dynrng (a52_state_t * state, - sample_t (* call) (sample_t, void *), void * data); - int a52_block (a52_state_t * state, sample_t * samples); -+ +void* a52_resample_init(uint32_t mm_accel,int flags,int chans); +extern int (* a52_resample) (float * _f, int16_t * s16); + +uint16_t crc16_block(uint8_t *data,uint32_t num_bytes); + -+#endif /* A52_H */ ---- liba52/a52_internal.h 2005-03-22 19:59:35.000000000 +0100 -+++ a52_internal.h 2004-03-19 01:15:49.000000000 +0100 -@@ -41,11 +43,12 @@ + #endif /* A52_H */ +--- liba52-0.7.4/a52_internal.h 2006-06-12 15:05:07.000000000 +0200 ++++ liba52/a52_internal.h 2006-06-05 02:23:02.000000000 +0200 +@@ -103,18 +107,34 @@ + #define DELTA_BIT_NONE (2) + #define DELTA_BIT_RESERVED (3) - int downmix_init (int input, int flags, sample_t * level, ++#ifdef ARCH_X86_64 ++# define REG_a "rax" ++# define REG_d "rdx" ++# define REG_S "rsi" ++# define REG_D "rdi" ++# define REG_BP "rbp" ++#else ++# define REG_a "eax" ++# define REG_d "edx" ++# define REG_S "esi" ++# define REG_D "edi" ++# define REG_BP "ebp" ++#endif ++ + void a52_bit_allocate (a52_state_t * state, ba_t * ba, int bndstart, + int start, int end, int fastleak, int slowleak, + expbap_t * expbap); + + int a52_downmix_init (int input, int flags, sample_t * level, sample_t clev, sample_t slev); +void downmix_accel_init(uint32_t mm_accel); - int downmix_coeff (sample_t * coeff, int acmod, int output, sample_t level, + int a52_downmix_coeff (sample_t * coeff, int acmod, int output, sample_t level, sample_t clev, sample_t slev); --void downmix (sample_t * samples, int acmod, int output, sample_t bias, -+extern void (*downmix) (sample_t * samples, int acmod, int output, sample_t bias, +-void a52_downmix (sample_t * samples, int acmod, int output, sample_t bias, ++extern void (*a52_downmix) (sample_t * samples, int acmod, int output, sample_t bias, sample_t clev, sample_t slev); --void upmix (sample_t * samples, int acmod, int output); -+extern void (*upmix) (sample_t * samples, int acmod, int output); +-void a52_upmix (sample_t * samples, int acmod, int output); ++extern void (*a52_upmix) (sample_t * samples, int acmod, int output); - void imdct_init (uint32_t mm_accel); - extern void (* imdct_256) (sample_t * data, sample_t * delay, sample_t bias); ---- liba52/bitstream.c 2005-03-22 19:59:35.000000000 +0100 -+++ bitstream.c 2004-03-19 01:15:49.000000000 +0100 -@@ -29,7 +35,12 @@ + void a52_imdct_init (uint32_t mm_accel); + void a52_imdct_256 (sample_t * data, sample_t * delay, sample_t bias); +-void a52_imdct_512 (sample_t * data, sample_t * delay, sample_t bias); ++extern void (*a52_imdct_512) (sample_t * data, sample_t * delay, sample_t bias); ++void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias); +--- liba52-0.7.4/bitstream.c 2006-06-12 15:05:07.000000000 +0200 ++++ liba52/bitstream.c 2006-06-05 02:23:02.000000000 +0200 +@@ -31,6 +35,10 @@ #define BUFFER_SIZE 4096 +#ifdef ALT_BITSTREAM_READER +int indx=0; -+uint32_t * buffer_start; -+#else - static uint32_t * buffer_start; +#endif - - uint32_t bits_left; - uint32_t current_word; -@@ -41,6 +52,9 @@ - align = (int)buf & 3; - buffer_start = (uint32_t *) (buf - align); - bits_left = 0; ++ + void a52_bitstream_set_ptr (a52_state_t * state, uint8_t * buf) + { + int align; +@@ -38,6 +46,9 @@ + align = (long)buf & 3; + state->buffer_start = (uint32_t *) (buf - align); + state->bits_left = 0; +#ifdef ALT_BITSTREAM_READER + indx=0; +#endif - bitstream_get (align * 8); + bitstream_get (state, align * 8); } ---- liba52/bitstream.h 2005-03-22 19:59:35.000000000 +0100 -+++ bitstream.h 2004-03-19 01:15:49.000000000 +0100 -@@ -19,6 +25,48 @@ +--- liba52-0.7.4/bitstream.h 2006-06-12 15:05:07.000000000 +0200 ++++ liba52/bitstream.h 2006-06-05 02:23:02.000000000 +0200 +@@ -21,6 +25,48 @@ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA */ @@ -114,16 +124,16 @@ /* (stolen from the kernel) */ #ifdef WORDS_BIGENDIAN -@@ -29,7 +77,7 @@ - # if defined (__i386__) +@@ -28,7 +74,7 @@ + + #else + +-# if 0 && defined (__i386__) ++# if defined (__i386__) # define swab32(x) __i386_swab32(x) -- static inline const uint32_t __i386_swab32(uint32_t x) -+ static always_inline const uint32_t __i386_swab32(uint32_t x) - { - __asm__("bswap %0" : "=r" (x) : "0" (x)); - return x; -@@ -37,25 +85,42 @@ + static inline const uint32_t __i386_swab32(uint32_t x) +@@ -39,19 +85,34 @@ # else @@ -141,24 +151,17 @@ #endif +#ifdef ALT_BITSTREAM_READER -+extern uint32_t *buffer_start; +extern int indx; -+#else - extern uint32_t bits_left; - extern uint32_t current_word; +#endif ++ + void a52_bitstream_set_ptr (a52_state_t * state, uint8_t * buf); + uint32_t a52_bitstream_get_bh (a52_state_t * state, uint32_t num_bits); + int32_t a52_bitstream_get_bh_2 (a52_state_t * state, uint32_t num_bits); - void bitstream_set_ptr (uint8_t * buf); - uint32_t bitstream_get_bh(uint32_t num_bits); - int32_t bitstream_get_bh_2(uint32_t num_bits); - -+ - static inline uint32_t --bitstream_get(uint32_t num_bits) -+bitstream_get(uint32_t num_bits) // note num_bits is practically a constant due to inlineing + static inline uint32_t bitstream_get (a52_state_t * state, uint32_t num_bits) { +#ifdef ALT_BITSTREAM_READER -+ uint32_t result= swab32( unaligned32(((uint8_t *)buffer_start)+(indx>>3)) ); ++ uint32_t result= swab32( unaligned32(((uint8_t *)state->buffer_start)+(indx>>3)) ); + + result<<= (indx&0x07); + result>>= 32 - num_bits; @@ -167,32 +170,28 @@ + return result; +#else uint32_t result; -- -+ - if(num_bits < bits_left) { - result = (current_word << (32 - bits_left)) >> (32 - num_bits); - bits_left -= num_bits; -@@ -63,11 +128,30 @@ + + if (num_bits < state->bits_left) { +@@ -61,10 +122,29 @@ } - return bitstream_get_bh(num_bits); + return a52_bitstream_get_bh (state, num_bits); +#endif +} + -+static inline void bitstream_skip(int num_bits) ++static inline void bitstream_skip(a52_state_t * state, int num_bits) +{ +#ifdef ALT_BITSTREAM_READER + indx+= num_bits; +#else -+ bitstream_get(num_bits); ++ bitstream_get(state, num_bits); +#endif } - static inline int32_t - bitstream_get_2(uint32_t num_bits) + static inline int32_t bitstream_get_2 (a52_state_t * state, uint32_t num_bits) { +#ifdef ALT_BITSTREAM_READER -+ int32_t result= swab32( unaligned32(((uint8_t *)buffer_start)+(indx>>3)) ); ++ int32_t result= swab32( unaligned32(((uint8_t *)state->buffer_start)+(indx>>3)) ); + + result<<= (indx&0x07); + result>>= 32 - num_bits; @@ -202,16 +201,16 @@ +#else int32_t result; - if(num_bits < bits_left) { -@@ -77,4 +161,5 @@ + if (num_bits < state->bits_left) { +@@ -74,4 +154,5 @@ } - return bitstream_get_bh_2(num_bits); + return a52_bitstream_get_bh_2 (state, num_bits); +#endif } ---- liba52/downmix.c 2005-03-22 19:59:35.000000000 +0100 -+++ downmix.c 2004-04-12 18:42:14.000000000 +0200 -@@ -17,18 +23,46 @@ +--- liba52-0.7.4/downmix.c 2006-06-12 15:17:53.000000000 +0200 ++++ liba52/downmix.c 2006-06-05 02:23:02.000000000 +0200 +@@ -23,18 +23,47 @@ * 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 @@ -220,10 +219,10 @@ */ #include "config.h" ++#include "asmalign.h" --#include <inttypes.h> #include <string.h> -+#include <inttypes.h> + #include <inttypes.h> #include "a52.h" #include "a52_internal.h" @@ -232,9 +231,9 @@ #define CONVERT(acmod,output) (((output) << 3) + (acmod)) + -+void (*downmix)(sample_t * samples, int acmod, int output, sample_t bias, ++void (*a52_downmix)(sample_t * samples, int acmod, int output, sample_t bias, + sample_t clev, sample_t slev)= NULL; -+void (*upmix)(sample_t * samples, int acmod, int output)= NULL; ++void (*a52_upmix)(sample_t * samples, int acmod, int output)= NULL; + +static void downmix_SSE (sample_t * samples, int acmod, int output, sample_t bias, + sample_t clev, sample_t slev); @@ -247,50 +246,28 @@ + +void downmix_accel_init(uint32_t mm_accel) +{ -+ upmix= upmix_C; -+ downmix= downmix_C; -+#ifdef ARCH_X86 -+ if(mm_accel & MM_ACCEL_X86_MMX) upmix= upmix_MMX; -+ if(mm_accel & MM_ACCEL_X86_SSE) downmix= downmix_SSE; -+ if(mm_accel & MM_ACCEL_X86_3DNOW) downmix= downmix_3dnow; ++ a52_upmix= upmix_C; ++ a52_downmix= downmix_C; ++#if defined(ARCH_X86) || defined(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; +#endif +} + - int downmix_init (int input, int flags, sample_t * level, + int a52_downmix_init (int input, int flags, sample_t * level, sample_t clev, sample_t slev) { -@@ -61,7 +95,7 @@ - output = flags & A52_CHANNEL_MASK; - if (output > A52_DOLBY) - return -1; -- -+ - output = table[output][input & 7]; - - if ((output == A52_STEREO) && -@@ -145,7 +179,6 @@ - *level *= 1 / (1 + 3 * LEVEL_3DB); - break; - } -- - return output; - } - -@@ -440,12 +473,11 @@ - static void zero (sample_t * samples) - { - int i; -- - for (i = 0; i < 256; i++) +@@ -451,7 +480,7 @@ samples[i] = 0; } --void downmix (sample_t * samples, int acmod, int output, sample_t bias, -+static void downmix_C (sample_t * samples, int acmod, int output, sample_t bias, +-void a52_downmix (sample_t * samples, int acmod, int output, sample_t bias, ++void downmix_C (sample_t * samples, int acmod, int output, sample_t bias, sample_t clev, sample_t slev) { switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) { -@@ -557,7 +589,7 @@ +@@ -563,7 +592,7 @@ break; case CONVERT (A52_3F2R, A52_2F1R): @@ -299,7 +276,7 @@ move2to1 (samples + 768, samples + 512, bias); break; -@@ -581,12 +613,12 @@ +@@ -587,12 +616,12 @@ break; case CONVERT (A52_3F1R, A52_3F2R): @@ -309,37 +286,37 @@ } } --void upmix (sample_t * samples, int acmod, int output) -+static void upmix_C (sample_t * samples, int acmod, int output) +-void a52_upmix (sample_t * samples, int acmod, int output) ++void upmix_C (sample_t * samples, int acmod, int output) { switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) { -@@ -651,3 +683,1137 @@ +@@ -657,3 +686,1137 @@ goto mix_31to21; } } + -+#ifdef ARCH_X86 ++#if defined(ARCH_X86) || defined(ARCH_X86_64) +static void mix2to1_SSE (sample_t * dest, sample_t * src, sample_t bias) +{ + asm volatile( + "movlps %2, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps (%0, %%esi), %%xmm0 \n\t" -+ "movaps 16(%0, %%esi), %%xmm1 \n\t" -+ "addps (%1, %%esi), %%xmm0 \n\t" -+ "addps 16(%1, %%esi), %%xmm1 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm0 \n\t" ++ "movaps 16(%0, %%"REG_S"), %%xmm1\n\t" ++ "addps (%1, %%"REG_S"), %%xmm0 \n\t" ++ "addps 16(%1, %%"REG_S"), %%xmm1\n\t" + "addps %%xmm7, %%xmm0 \n\t" + "addps %%xmm7, %%xmm1 \n\t" -+ "movaps %%xmm0, (%1, %%esi) \n\t" -+ "movaps %%xmm1, 16(%1, %%esi) \n\t" -+ "addl $32, %%esi \n\t" ++ "movaps %%xmm0, (%1, %%"REG_S") \n\t" ++ "movaps %%xmm1, 16(%1, %%"REG_S")\n\t" ++ "add $32, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (src+256), "r" (dest+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -348,19 +325,19 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps (%0, %%esi), %%xmm0 \n\t" -+ "movaps 1024(%0, %%esi), %%xmm1 \n\t" -+ "addps 2048(%0, %%esi), %%xmm0 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm0 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm1\n\t" ++ "addps 2048(%0, %%"REG_S"), %%xmm0\n\t" + "addps %%xmm7, %%xmm1 \n\t" + "addps %%xmm1, %%xmm0 \n\t" -+ "movaps %%xmm0, (%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm0, (%0, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -369,20 +346,20 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps (%0, %%esi), %%xmm0 \n\t" -+ "movaps 1024(%0, %%esi), %%xmm1 \n\t" -+ "addps 2048(%0, %%esi), %%xmm0 \n\t" -+ "addps 3072(%0, %%esi), %%xmm1 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm0 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm1\n\t" ++ "addps 2048(%0, %%"REG_S"), %%xmm0\n\t" ++ "addps 3072(%0, %%"REG_S"), %%xmm1\n\t" + "addps %%xmm7, %%xmm0 \n\t" + "addps %%xmm1, %%xmm0 \n\t" -+ "movaps %%xmm0, (%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm0, (%0, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -391,21 +368,21 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps (%0, %%esi), %%xmm0 \n\t" -+ "movaps 1024(%0, %%esi), %%xmm1 \n\t" -+ "addps 2048(%0, %%esi), %%xmm0 \n\t" -+ "addps 3072(%0, %%esi), %%xmm1 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm0 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm1\n\t" ++ "addps 2048(%0, %%"REG_S"), %%xmm0\n\t" ++ "addps 3072(%0, %%"REG_S"), %%xmm1\n\t" + "addps %%xmm7, %%xmm0 \n\t" -+ "addps 4096(%0, %%esi), %%xmm1 \n\t" ++ "addps 4096(%0, %%"REG_S"), %%xmm1\n\t" + "addps %%xmm1, %%xmm0 \n\t" -+ "movaps %%xmm0, (%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm0, (%0, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -414,21 +391,21 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 1024(%0, %%esi), %%xmm0 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t" + "addps %%xmm7, %%xmm0 \n\t" //common -+ "movaps (%0, %%esi), %%xmm1 \n\t" -+ "movaps 2048(%0, %%esi), %%xmm2 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" ++ "movaps 2048(%0, %%"REG_S"), %%xmm2\n\t" + "addps %%xmm0, %%xmm1 \n\t" + "addps %%xmm0, %%xmm2 \n\t" -+ "movaps %%xmm1, (%0, %%esi) \n\t" -+ "movaps %%xmm2, 1024(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm1, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -437,21 +414,21 @@ + asm volatile( + "movlps %2, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 1024(%1, %%esi), %%xmm0 \n\t" ++ "movaps 1024(%1, %%"REG_S"), %%xmm0\n\t" + "addps %%xmm7, %%xmm0 \n\t" //common -+ "movaps (%0, %%esi), %%xmm1 \n\t" -+ "movaps (%1, %%esi), %%xmm2 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" ++ "movaps (%1, %%"REG_S"), %%xmm2 \n\t" + "addps %%xmm0, %%xmm1 \n\t" + "addps %%xmm0, %%xmm2 \n\t" -+ "movaps %%xmm1, (%0, %%esi) \n\t" -+ "movaps %%xmm2, (%1, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm1, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm2, (%1, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (left+256), "r" (right+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -460,22 +437,22 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 2048(%0, %%esi), %%xmm0 \n\t" // surround -+ "movaps (%0, %%esi), %%xmm1 \n\t" -+ "movaps 1024(%0, %%esi), %%xmm2 \n\t" ++ "movaps 2048(%0, %%"REG_S"), %%xmm0\n\t" // surround ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm2\n\t" + "addps %%xmm7, %%xmm1 \n\t" + "addps %%xmm7, %%xmm2 \n\t" + "subps %%xmm0, %%xmm1 \n\t" + "addps %%xmm0, %%xmm2 \n\t" -+ "movaps %%xmm1, (%0, %%esi) \n\t" -+ "movaps %%xmm2, 1024(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm1, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -484,22 +461,22 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 1024(%0, %%esi), %%xmm0 \n\t" -+ "addps 3072(%0, %%esi), %%xmm0 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t" ++ "addps 3072(%0, %%"REG_S"), %%xmm0\n\t" + "addps %%xmm7, %%xmm0 \n\t" // common -+ "movaps (%0, %%esi), %%xmm1 \n\t" -+ "movaps 2048(%0, %%esi), %%xmm2 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" ++ "movaps 2048(%0, %%"REG_S"), %%xmm2\n\t" + "addps %%xmm0, %%xmm1 \n\t" + "addps %%xmm0, %%xmm2 \n\t" -+ "movaps %%xmm1, (%0, %%esi) \n\t" -+ "movaps %%xmm2, 1024(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm1, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -508,24 +485,24 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 1024(%0, %%esi), %%xmm0 \n\t" -+ "movaps 3072(%0, %%esi), %%xmm3 \n\t" // surround ++ "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t" ++ "movaps 3072(%0, %%"REG_S"), %%xmm3\n\t" // surround + "addps %%xmm7, %%xmm0 \n\t" // common -+ "movaps (%0, %%esi), %%xmm1 \n\t" -+ "movaps 2048(%0, %%esi), %%xmm2 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" ++ "movaps 2048(%0, %%"REG_S"), %%xmm2\n\t" + "addps %%xmm0, %%xmm1 \n\t" + "addps %%xmm0, %%xmm2 \n\t" + "subps %%xmm3, %%xmm1 \n\t" + "addps %%xmm3, %%xmm2 \n\t" -+ "movaps %%xmm1, (%0, %%esi) \n\t" -+ "movaps %%xmm2, 1024(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm1, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -534,23 +511,23 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 2048(%0, %%esi), %%xmm0 \n\t" -+ "addps 3072(%0, %%esi), %%xmm0 \n\t" // surround -+ "movaps (%0, %%esi), %%xmm1 \n\t" -+ "movaps 1024(%0, %%esi), %%xmm2 \n\t" ++ "movaps 2048(%0, %%"REG_S"), %%xmm0\n\t" ++ "addps 3072(%0, %%"REG_S"), %%xmm0\n\t" // surround ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm2\n\t" + "addps %%xmm7, %%xmm1 \n\t" + "addps %%xmm7, %%xmm2 \n\t" + "subps %%xmm0, %%xmm1 \n\t" + "addps %%xmm0, %%xmm2 \n\t" -+ "movaps %%xmm1, (%0, %%esi) \n\t" -+ "movaps %%xmm2, 1024(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm1, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm2, 1024(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -559,22 +536,22 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 1024(%0, %%esi), %%xmm0 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t" + "addps %%xmm7, %%xmm0 \n\t" // common + "movaps %%xmm0, %%xmm1 \n\t" // common -+ "addps (%0, %%esi), %%xmm0 \n\t" -+ "addps 2048(%0, %%esi), %%xmm1 \n\t" -+ "addps 3072(%0, %%esi), %%xmm0 \n\t" -+ "addps 4096(%0, %%esi), %%xmm1 \n\t" -+ "movaps %%xmm0, (%0, %%esi) \n\t" -+ "movaps %%xmm1, 1024(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "addps (%0, %%"REG_S"), %%xmm0 \n\t" ++ "addps 2048(%0, %%"REG_S"), %%xmm1\n\t" ++ "addps 3072(%0, %%"REG_S"), %%xmm0\n\t" ++ "addps 4096(%0, %%"REG_S"), %%xmm1\n\t" ++ "movaps %%xmm0, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm1, 1024(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -583,25 +560,25 @@ + asm volatile( + "movlps %1, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 1024(%0, %%esi), %%xmm0 \n\t" -+ "movaps 3072(%0, %%esi), %%xmm2 \n\t" ++ "movaps 1024(%0, %%"REG_S"), %%xmm0\n\t" ++ "movaps 3072(%0, %%"REG_S"), %%xmm2\n\t" + "addps %%xmm7, %%xmm0 \n\t" // common -+ "addps 4096(%0, %%esi), %%xmm2 \n\t" // surround -+ "movaps (%0, %%esi), %%xmm1 \n\t" -+ "movaps 2048(%0, %%esi), %%xmm3 \n\t" ++ "addps 4096(%0, %%"REG_S"), %%xmm2\n\t" // surround ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" ++ "movaps 2048(%0, %%"REG_S"), %%xmm3\n\t" + "subps %%xmm2, %%xmm1 \n\t" + "addps %%xmm2, %%xmm3 \n\t" + "addps %%xmm0, %%xmm1 \n\t" + "addps %%xmm0, %%xmm3 \n\t" -+ "movaps %%xmm1, (%0, %%esi) \n\t" -+ "movaps %%xmm3, 1024(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm1, (%0, %%"REG_S") \n\t" ++ "movaps %%xmm3, 1024(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -610,40 +587,40 @@ + asm volatile( + "movlps %2, %%xmm7 \n\t" + "shufps $0x00, %%xmm7, %%xmm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps (%0, %%esi), %%xmm0 \n\t" -+ "movaps 16(%0, %%esi), %%xmm1 \n\t" -+ "addps 1024(%0, %%esi), %%xmm0 \n\t" -+ "addps 1040(%0, %%esi), %%xmm1 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm0 \n\t" ++ "movaps 16(%0, %%"REG_S"), %%xmm1\n\t" ++ "addps 1024(%0, %%"REG_S"), %%xmm0\n\t" ++ "addps 1040(%0, %%"REG_S"), %%xmm1\n\t" + "addps %%xmm7, %%xmm0 \n\t" + "addps %%xmm7, %%xmm1 \n\t" -+ "movaps %%xmm0, (%1, %%esi) \n\t" -+ "movaps %%xmm1, 16(%1, %%esi) \n\t" -+ "addl $32, %%esi \n\t" ++ "movaps %%xmm0, (%1, %%"REG_S") \n\t" ++ "movaps %%xmm1, 16(%1, %%"REG_S")\n\t" ++ "add $32, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (src+256), "r" (dest+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + +static void zero_MMX(sample_t * samples) +{ + asm volatile( -+ "movl $-1024, %%esi \n\t" ++ "mov $-1024, %%"REG_S" \n\t" + "pxor %%mm0, %%mm0 \n\t" -+ ".balign 16\n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq %%mm0, (%0, %%esi) \n\t" -+ "movq %%mm0, 8(%0, %%esi) \n\t" -+ "movq %%mm0, 16(%0, %%esi) \n\t" -+ "movq %%mm0, 24(%0, %%esi) \n\t" -+ "addl $32, %%esi \n\t" ++ "movq %%mm0, (%0, %%"REG_S") \n\t" ++ "movq %%mm0, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm0, 16(%0, %%"REG_S") \n\t" ++ "movq %%mm0, 24(%0, %%"REG_S") \n\t" ++ "add $32, %%"REG_S" \n\t" + " jnz 1b \n\t" + "emms" + :: "r" (samples+256) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -892,29 +869,29 @@ + asm volatile( + "movd %2, %%mm7 \n\t" + "punpckldq %2, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq (%0, %%esi), %%mm0 \n\t" -+ "movq 8(%0, %%esi), %%mm1 \n\t" -+ "movq 16(%0, %%esi), %%mm2 \n\t" -+ "movq 24(%0, %%esi), %%mm3 \n\t" -+ "pfadd (%1, %%esi), %%mm0 \n\t" -+ "pfadd 8(%1, %%esi), %%mm1 \n\t" -+ "pfadd 16(%1, %%esi), %%mm2 \n\t" -+ "pfadd 24(%1, %%esi), %%mm3 \n\t" ++ "movq (%0, %%"REG_S"), %%mm0 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm1 \n\t" ++ "movq 16(%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 24(%0, %%"REG_S"), %%mm3 \n\t" ++ "pfadd (%1, %%"REG_S"), %%mm0 \n\t" ++ "pfadd 8(%1, %%"REG_S"), %%mm1 \n\t" ++ "pfadd 16(%1, %%"REG_S"), %%mm2 \n\t" ++ "pfadd 24(%1, %%"REG_S"), %%mm3 \n\t" + "pfadd %%mm7, %%mm0 \n\t" + "pfadd %%mm7, %%mm1 \n\t" + "pfadd %%mm7, %%mm2 \n\t" + "pfadd %%mm7, %%mm3 \n\t" -+ "movq %%mm0, (%1, %%esi) \n\t" -+ "movq %%mm1, 8(%1, %%esi) \n\t" -+ "movq %%mm2, 16(%1, %%esi) \n\t" -+ "movq %%mm3, 24(%1, %%esi) \n\t" -+ "addl $32, %%esi \n\t" ++ "movq %%mm0, (%1, %%"REG_S") \n\t" ++ "movq %%mm1, 8(%1, %%"REG_S") \n\t" ++ "movq %%mm2, 16(%1, %%"REG_S") \n\t" ++ "movq %%mm3, 24(%1, %%"REG_S") \n\t" ++ "add $32, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (src+256), "r" (dest+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -923,25 +900,25 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq (%0, %%esi), %%mm0 \n\t" -+ "movq 8(%0, %%esi), %%mm1 \n\t" -+ "movq 1024(%0, %%esi), %%mm2 \n\t" -+ "movq 1032(%0, %%esi), %%mm3 \n\t" -+ "pfadd 2048(%0, %%esi), %%mm0 \n\t" -+ "pfadd 2056(%0, %%esi), %%mm1 \n\t" ++ "movq (%0, %%"REG_S"), %%mm0 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm1 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm2\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm3\n\t" ++ "pfadd 2048(%0, %%"REG_S"), %%mm0\n\t" ++ "pfadd 2056(%0, %%"REG_S"), %%mm1\n\t" + "pfadd %%mm7, %%mm0 \n\t" + "pfadd %%mm7, %%mm1 \n\t" + "pfadd %%mm2, %%mm0 \n\t" + "pfadd %%mm3, %%mm1 \n\t" -+ "movq %%mm0, (%0, %%esi) \n\t" -+ "movq %%mm1, 8(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm0, (%0, %%"REG_S") \n\t" ++ "movq %%mm1, 8(%0, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -950,27 +927,27 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq (%0, %%esi), %%mm0 \n\t" -+ "movq 8(%0, %%esi), %%mm1 \n\t" -+ "movq 1024(%0, %%esi), %%mm2 \n\t" -+ "movq 1032(%0, %%esi), %%mm3 \n\t" -+ "pfadd 2048(%0, %%esi), %%mm0 \n\t" -+ "pfadd 2056(%0, %%esi), %%mm1 \n\t" -+ "pfadd 3072(%0, %%esi), %%mm2 \n\t" -+ "pfadd 3080(%0, %%esi), %%mm3 \n\t" ++ "movq (%0, %%"REG_S"), %%mm0 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm1 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm2\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm3\n\t" ++ "pfadd 2048(%0, %%"REG_S"), %%mm0\n\t" ++ "pfadd 2056(%0, %%"REG_S"), %%mm1\n\t" ++ "pfadd 3072(%0, %%"REG_S"), %%mm2\n\t" ++ "pfadd 3080(%0, %%"REG_S"), %%mm3\n\t" + "pfadd %%mm7, %%mm0 \n\t" + "pfadd %%mm7, %%mm1 \n\t" + "pfadd %%mm2, %%mm0 \n\t" + "pfadd %%mm3, %%mm1 \n\t" -+ "movq %%mm0, (%0, %%esi) \n\t" -+ "movq %%mm1, 8(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm0, (%0, %%"REG_S") \n\t" ++ "movq %%mm1, 8(%0, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -979,29 +956,29 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq (%0, %%esi), %%mm0 \n\t" -+ "movq 8(%0, %%esi), %%mm1 \n\t" -+ "movq 1024(%0, %%esi), %%mm2 \n\t" -+ "movq 1032(%0, %%esi), %%mm3 \n\t" -+ "pfadd 2048(%0, %%esi), %%mm0 \n\t" -+ "pfadd 2056(%0, %%esi), %%mm1 \n\t" -+ "pfadd 3072(%0, %%esi), %%mm2 \n\t" -+ "pfadd 3080(%0, %%esi), %%mm3 \n\t" ++ "movq (%0, %%"REG_S"), %%mm0 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm1 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm2\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm3\n\t" ++ "pfadd 2048(%0, %%"REG_S"), %%mm0\n\t" ++ "pfadd 2056(%0, %%"REG_S"), %%mm1\n\t" ++ "pfadd 3072(%0, %%"REG_S"), %%mm2\n\t" ++ "pfadd 3080(%0, %%"REG_S"), %%mm3\n\t" + "pfadd %%mm7, %%mm0 \n\t" + "pfadd %%mm7, %%mm1 \n\t" -+ "pfadd 4096(%0, %%esi), %%mm2 \n\t" -+ "pfadd 4104(%0, %%esi), %%mm3 \n\t" ++ "pfadd 4096(%0, %%"REG_S"), %%mm2\n\t" ++ "pfadd 4104(%0, %%"REG_S"), %%mm3\n\t" + "pfadd %%mm2, %%mm0 \n\t" + "pfadd %%mm3, %%mm1 \n\t" -+ "movq %%mm0, (%0, %%esi) \n\t" -+ "movq %%mm1, 8(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm0, (%0, %%"REG_S") \n\t" ++ "movq %%mm1, 8(%0, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1010,29 +987,29 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq 1024(%0, %%esi), %%mm0 \n\t" -+ "movq 1032(%0, %%esi), %%mm1 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm0\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm1\n\t" + "pfadd %%mm7, %%mm0 \n\t" //common + "pfadd %%mm7, %%mm1 \n\t" //common -+ "movq (%0, %%esi), %%mm2 \n\t" -+ "movq 8(%0, %%esi), %%mm3 \n\t" -+ "movq 2048(%0, %%esi), %%mm4 \n\t" -+ "movq 2056(%0, %%esi), %%mm5 \n\t" ++ "movq (%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm3 \n\t" ++ "movq 2048(%0, %%"REG_S"), %%mm4\n\t" ++ "movq 2056(%0, %%"REG_S"), %%mm5\n\t" + "pfadd %%mm0, %%mm2 \n\t" + "pfadd %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm4 \n\t" + "pfadd %%mm1, %%mm5 \n\t" -+ "movq %%mm2, (%0, %%esi) \n\t" -+ "movq %%mm3, 8(%0, %%esi) \n\t" -+ "movq %%mm4, 1024(%0, %%esi) \n\t" -+ "movq %%mm5, 1032(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm2, (%0, %%"REG_S") \n\t" ++ "movq %%mm3, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm4, 1024(%0, %%"REG_S")\n\t" ++ "movq %%mm5, 1032(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1041,29 +1018,29 @@ + asm volatile( + "movd %2, %%mm7 \n\t" + "punpckldq %2, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq 1024(%1, %%esi), %%mm0 \n\t" -+ "movq 1032(%1, %%esi), %%mm1 \n\t" ++ "movq 1024(%1, %%"REG_S"), %%mm0\n\t" ++ "movq 1032(%1, %%"REG_S"), %%mm1\n\t" + "pfadd %%mm7, %%mm0 \n\t" //common + "pfadd %%mm7, %%mm1 \n\t" //common -+ "movq (%0, %%esi), %%mm2 \n\t" -+ "movq 8(%0, %%esi), %%mm3 \n\t" -+ "movq (%1, %%esi), %%mm4 \n\t" -+ "movq 8(%1, %%esi), %%mm5 \n\t" ++ "movq (%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm3 \n\t" ++ "movq (%1, %%"REG_S"), %%mm4 \n\t" ++ "movq 8(%1, %%"REG_S"), %%mm5 \n\t" + "pfadd %%mm0, %%mm2 \n\t" + "pfadd %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm4 \n\t" + "pfadd %%mm1, %%mm5 \n\t" -+ "movq %%mm2, (%0, %%esi) \n\t" -+ "movq %%mm3, 8(%0, %%esi) \n\t" -+ "movq %%mm4, (%1, %%esi) \n\t" -+ "movq %%mm5, 8(%1, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm2, (%0, %%"REG_S") \n\t" ++ "movq %%mm3, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm4, (%1, %%"REG_S") \n\t" ++ "movq %%mm5, 8(%1, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (left+256), "r" (right+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1072,15 +1049,15 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq 2048(%0, %%esi), %%mm0 \n\t" // surround -+ "movq 2056(%0, %%esi), %%mm1 \n\t" // surround -+ "movq (%0, %%esi), %%mm2 \n\t" -+ "movq 8(%0, %%esi), %%mm3 \n\t" -+ "movq 1024(%0, %%esi), %%mm4 \n\t" -+ "movq 1032(%0, %%esi), %%mm5 \n\t" ++ "movq 2048(%0, %%"REG_S"), %%mm0\n\t" // surround ++ "movq 2056(%0, %%"REG_S"), %%mm1\n\t" // surround ++ "movq (%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm3 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm4\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm5\n\t" + "pfadd %%mm7, %%mm2 \n\t" + "pfadd %%mm7, %%mm3 \n\t" + "pfadd %%mm7, %%mm4 \n\t" @@ -1089,14 +1066,14 @@ + "pfsub %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm4 \n\t" + "pfadd %%mm1, %%mm5 \n\t" -+ "movq %%mm2, (%0, %%esi) \n\t" -+ "movq %%mm3, 8(%0, %%esi) \n\t" -+ "movq %%mm4, 1024(%0, %%esi) \n\t" -+ "movq %%mm5, 1032(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm2, (%0, %%"REG_S") \n\t" ++ "movq %%mm3, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm4, 1024(%0, %%"REG_S")\n\t" ++ "movq %%mm5, 1032(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1105,31 +1082,31 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq 1024(%0, %%esi), %%mm0 \n\t" -+ "movq 1032(%0, %%esi), %%mm1 \n\t" -+ "pfadd 3072(%0, %%esi), %%mm0 \n\t" -+ "pfadd 3080(%0, %%esi), %%mm1 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm0\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm1\n\t" ++ "pfadd 3072(%0, %%"REG_S"), %%mm0\n\t" ++ "pfadd 3080(%0, %%"REG_S"), %%mm1\n\t" + "pfadd %%mm7, %%mm0 \n\t" // common + "pfadd %%mm7, %%mm1 \n\t" // common -+ "movq (%0, %%esi), %%mm2 \n\t" -+ "movq 8(%0, %%esi), %%mm3 \n\t" -+ "movq 2048(%0, %%esi), %%mm4 \n\t" -+ "movq 2056(%0, %%esi), %%mm5 \n\t" ++ "movq (%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm3 \n\t" ++ "movq 2048(%0, %%"REG_S"), %%mm4\n\t" ++ "movq 2056(%0, %%"REG_S"), %%mm5\n\t" + "pfadd %%mm0, %%mm2 \n\t" + "pfadd %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm4 \n\t" + "pfadd %%mm1, %%mm5 \n\t" -+ "movq %%mm2, (%0, %%esi) \n\t" -+ "movq %%mm3, 8(%0, %%esi) \n\t" -+ "movq %%mm4, 1024(%0, %%esi) \n\t" -+ "movq %%mm5, 1032(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm2, (%0, %%"REG_S") \n\t" ++ "movq %%mm3, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm4, 1024(%0, %%"REG_S")\n\t" ++ "movq %%mm5, 1032(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1138,35 +1115,35 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq 1024(%0, %%esi), %%mm0 \n\t" -+ "movq 1032(%0, %%esi), %%mm1 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm0\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm1\n\t" + "pfadd %%mm7, %%mm0 \n\t" // common + "pfadd %%mm7, %%mm1 \n\t" // common -+ "movq (%0, %%esi), %%mm2 \n\t" -+ "movq 8(%0, %%esi), %%mm3 \n\t" -+ "movq 2048(%0, %%esi), %%mm4 \n\t" -+ "movq 2056(%0, %%esi), %%mm5 \n\t" ++ "movq (%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm3 \n\t" ++ "movq 2048(%0, %%"REG_S"), %%mm4\n\t" ++ "movq 2056(%0, %%"REG_S"), %%mm5\n\t" + "pfadd %%mm0, %%mm2 \n\t" + "pfadd %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm4 \n\t" + "pfadd %%mm1, %%mm5 \n\t" -+ "movq 3072(%0, %%esi), %%mm0 \n\t" // surround -+ "movq 3080(%0, %%esi), %%mm1 \n\t" // surround ++ "movq 3072(%0, %%"REG_S"), %%mm0\n\t" // surround ++ "movq 3080(%0, %%"REG_S"), %%mm1\n\t" // surround + "pfsub %%mm0, %%mm2 \n\t" + "pfsub %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm4 \n\t" + "pfadd %%mm1, %%mm5 \n\t" -+ "movq %%mm2, (%0, %%esi) \n\t" -+ "movq %%mm3, 8(%0, %%esi) \n\t" -+ "movq %%mm4, 1024(%0, %%esi) \n\t" -+ "movq %%mm5, 1032(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm2, (%0, %%"REG_S") \n\t" ++ "movq %%mm3, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm4, 1024(%0, %%"REG_S")\n\t" ++ "movq %%mm5, 1032(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1175,17 +1152,17 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq 2048(%0, %%esi), %%mm0 \n\t" -+ "movq 2056(%0, %%esi), %%mm1 \n\t" -+ "pfadd 3072(%0, %%esi), %%mm0 \n\t" // surround -+ "pfadd 3080(%0, %%esi), %%mm1 \n\t" // surround -+ "movq (%0, %%esi), %%mm2 \n\t" -+ "movq 8(%0, %%esi), %%mm3 \n\t" -+ "movq 1024(%0, %%esi), %%mm4 \n\t" -+ "movq 1032(%0, %%esi), %%mm5 \n\t" ++ "movq 2048(%0, %%"REG_S"), %%mm0\n\t" ++ "movq 2056(%0, %%"REG_S"), %%mm1\n\t" ++ "pfadd 3072(%0, %%"REG_S"), %%mm0\n\t" // surround ++ "pfadd 3080(%0, %%"REG_S"), %%mm1\n\t" // surround ++ "movq (%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm3 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm4\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm5\n\t" + "pfadd %%mm7, %%mm2 \n\t" + "pfadd %%mm7, %%mm3 \n\t" + "pfadd %%mm7, %%mm4 \n\t" @@ -1194,14 +1171,14 @@ + "pfsub %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm4 \n\t" + "pfadd %%mm1, %%mm5 \n\t" -+ "movq %%mm2, (%0, %%esi) \n\t" -+ "movq %%mm3, 8(%0, %%esi) \n\t" -+ "movq %%mm4, 1024(%0, %%esi) \n\t" -+ "movq %%mm5, 1032(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm2, (%0, %%"REG_S") \n\t" ++ "movq %%mm3, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm4, 1024(%0, %%"REG_S")\n\t" ++ "movq %%mm5, 1032(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1210,31 +1187,31 @@ + asm volatile( + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq 1024(%0, %%esi), %%mm0 \n\t" -+ "movq 1032(%0, %%esi), %%mm1 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm0\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm1\n\t" + "pfadd %%mm7, %%mm0 \n\t" // common + "pfadd %%mm7, %%mm1 \n\t" // common + "movq %%mm0, %%mm2 \n\t" // common + "movq %%mm1, %%mm3 \n\t" // common -+ "pfadd (%0, %%esi), %%mm0 \n\t" -+ "pfadd 8(%0, %%esi), %%mm1 \n\t" -+ "pfadd 2048(%0, %%esi), %%mm2 \n\t" -+ "pfadd 2056(%0, %%esi), %%mm3 \n\t" -+ "pfadd 3072(%0, %%esi), %%mm0 \n\t" -+ "pfadd 3080(%0, %%esi), %%mm1 \n\t" -+ "pfadd 4096(%0, %%esi), %%mm2 \n\t" -+ "pfadd 4104(%0, %%esi), %%mm3 \n\t" -+ "movq %%mm0, (%0, %%esi) \n\t" -+ "movq %%mm1, 8(%0, %%esi) \n\t" -+ "movq %%mm2, 1024(%0, %%esi) \n\t" -+ "movq %%mm3, 1032(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "pfadd (%0, %%"REG_S"), %%mm0 \n\t" ++ "pfadd 8(%0, %%"REG_S"), %%mm1 \n\t" ++ "pfadd 2048(%0, %%"REG_S"), %%mm2\n\t" ++ "pfadd 2056(%0, %%"REG_S"), %%mm3\n\t" ++ "pfadd 3072(%0, %%"REG_S"), %%mm0\n\t" ++ "pfadd 3080(%0, %%"REG_S"), %%mm1\n\t" ++ "pfadd 4096(%0, %%"REG_S"), %%mm2\n\t" ++ "pfadd 4104(%0, %%"REG_S"), %%mm3\n\t" ++ "movq %%mm0, (%0, %%"REG_S") \n\t" ++ "movq %%mm1, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm2, 1024(%0, %%"REG_S")\n\t" ++ "movq %%mm3, 1032(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1242,23 +1219,23 @@ +static void mix32toS_3dnow (sample_t * samples, sample_t bias) +{ + asm volatile( -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" + "movd %1, %%mm7 \n\t" + "punpckldq %1, %%mm7 \n\t" -+ "movq 1024(%0, %%esi), %%mm0 \n\t" -+ "movq 1032(%0, %%esi), %%mm1 \n\t" -+ "movq 3072(%0, %%esi), %%mm4 \n\t" -+ "movq 3080(%0, %%esi), %%mm5 \n\t" ++ "movq 1024(%0, %%"REG_S"), %%mm0\n\t" ++ "movq 1032(%0, %%"REG_S"), %%mm1\n\t" ++ "movq 3072(%0, %%"REG_S"), %%mm4\n\t" ++ "movq 3080(%0, %%"REG_S"), %%mm5\n\t" + "pfadd %%mm7, %%mm0 \n\t" // common + "pfadd %%mm7, %%mm1 \n\t" // common -+ "pfadd 4096(%0, %%esi), %%mm4 \n\t" // surround -+ "pfadd 4104(%0, %%esi), %%mm5 \n\t" // surround -+ "movq (%0, %%esi), %%mm2 \n\t" -+ "movq 8(%0, %%esi), %%mm3 \n\t" -+ "movq 2048(%0, %%esi), %%mm6 \n\t" -+ "movq 2056(%0, %%esi), %%mm7 \n\t" ++ "pfadd 4096(%0, %%"REG_S"), %%mm4\n\t" // surround ++ "pfadd 4104(%0, %%"REG_S"), %%mm5\n\t" // surround ++ "movq (%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm3 \n\t" ++ "movq 2048(%0, %%"REG_S"), %%mm6\n\t" ++ "movq 2056(%0, %%"REG_S"), %%mm7\n\t" + "pfsub %%mm4, %%mm2 \n\t" + "pfsub %%mm5, %%mm3 \n\t" + "pfadd %%mm4, %%mm6 \n\t" @@ -1267,14 +1244,14 @@ + "pfadd %%mm1, %%mm3 \n\t" + "pfadd %%mm0, %%mm6 \n\t" + "pfadd %%mm1, %%mm7 \n\t" -+ "movq %%mm2, (%0, %%esi) \n\t" -+ "movq %%mm3, 8(%0, %%esi) \n\t" -+ "movq %%mm6, 1024(%0, %%esi) \n\t" -+ "movq %%mm7, 1032(%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movq %%mm2, (%0, %%"REG_S") \n\t" ++ "movq %%mm3, 8(%0, %%"REG_S") \n\t" ++ "movq %%mm6, 1024(%0, %%"REG_S")\n\t" ++ "movq %%mm7, 1032(%0, %%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (samples+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1283,29 +1260,29 @@ + asm volatile( + "movd %2, %%mm7 \n\t" + "punpckldq %2, %%mm7 \n\t" -+ "movl $-1024, %%esi \n\t" -+ ".balign 16\n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movq (%0, %%esi), %%mm0 \n\t" -+ "movq 8(%0, %%esi), %%mm1 \n\t" -+ "movq 16(%0, %%esi), %%mm2 \n\t" -+ "movq 24(%0, %%esi), %%mm3 \n\t" -+ "pfadd 1024(%0, %%esi), %%mm0 \n\t" -+ "pfadd 1032(%0, %%esi), %%mm1 \n\t" -+ "pfadd 1040(%0, %%esi), %%mm2 \n\t" -+ "pfadd 1048(%0, %%esi), %%mm3 \n\t" ++ "movq (%0, %%"REG_S"), %%mm0 \n\t" ++ "movq 8(%0, %%"REG_S"), %%mm1 \n\t" ++ "movq 16(%0, %%"REG_S"), %%mm2 \n\t" ++ "movq 24(%0, %%"REG_S"), %%mm3 \n\t" ++ "pfadd 1024(%0, %%"REG_S"), %%mm0\n\t" ++ "pfadd 1032(%0, %%"REG_S"), %%mm1\n\t" ++ "pfadd 1040(%0, %%"REG_S"), %%mm2\n\t" ++ "pfadd 1048(%0, %%"REG_S"), %%mm3\n\t" + "pfadd %%mm7, %%mm0 \n\t" + "pfadd %%mm7, %%mm1 \n\t" + "pfadd %%mm7, %%mm2 \n\t" + "pfadd %%mm7, %%mm3 \n\t" -+ "movq %%mm0, (%1, %%esi) \n\t" -+ "movq %%mm1, 8(%1, %%esi) \n\t" -+ "movq %%mm2, 16(%1, %%esi) \n\t" -+ "movq %%mm3, 24(%1, %%esi) \n\t" -+ "addl $32, %%esi \n\t" ++ "movq %%mm0, (%1, %%"REG_S") \n\t" ++ "movq %%mm1, 8(%1, %%"REG_S") \n\t" ++ "movq %%mm2, 16(%1, %%"REG_S") \n\t" ++ "movq %%mm3, 24(%1, %%"REG_S") \n\t" ++ "add $32, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (src+256), "r" (dest+256), "m" (bias) -+ : "%esi" ++ : "%"REG_S + ); +} + @@ -1451,10 +1428,10 @@ + __asm __volatile("femms":::"memory"); +} + -+#endif //ARCH_X86 ---- liba52/imdct.c 2005-03-22 19:59:35.000000000 +0100 -+++ imdct.c 2004-04-26 22:00:57.000000000 +0200 -@@ -17,17 +23,32 @@ ++#endif // ARCH_X86 || ARCH_X86_64 +--- liba52-0.7.4/imdct.c 2006-06-12 15:18:27.000000000 +0200 ++++ liba52/imdct.c 2006-06-12 19:18:39.000000000 +0200 +@@ -26,9 +26,15 @@ * 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 @@ -1466,34 +1443,27 @@ */ #include "config.h" ++#include "asmalign.h" --#include <inttypes.h> #include <math.h> #include <stdio.h> -+#ifndef M_PI -+#define M_PI 3.1415926535897932384626433832795029 -+#endif -+#include <inttypes.h> - +@@ -43,12 +49,49 @@ #include "a52.h" #include "a52_internal.h" #include "mm_accel.h" +#include "mangle.h" + ++void (*a52_imdct_512) (sample_t * data, sample_t * delay, sample_t bias); ++ +#ifdef RUNTIME_CPUDETECT +#undef HAVE_3DNOWEX +#endif -+ -+#define USE_AC3_C - void (* imdct_256) (sample_t data[], sample_t delay[], sample_t bias); - void (* imdct_512) (sample_t data[], sample_t delay[], sample_t bias); -@@ -37,9 +58,22 @@ + typedef struct complex_s { + sample_t real; sample_t imag; } complex_t; -+static void fft_128p(complex_t *a); -+ +static const int pm128[128] attribute_used __attribute__((aligned(16))) = +{ + 0, 16, 32, 48, 64, 80, 96, 112, 8, 40, 72, 104, 24, 56, 88, 120, @@ -1505,22 +1475,51 @@ + 3, 19, 35, 51, 67, 83, 99, 115, 11, 43, 75, 107, 27, 59, 91, 123, + 7, 23, 39, 55, 71, 87, 103, 119, 15, 31, 47, 63, 79, 95, 111, 127 +}; - - /* 128 point bit-reverse LUT */ --static uint8_t bit_reverse_512[] = { ++ +static uint8_t attribute_used bit_reverse_512[] = { - 0x00, 0x40, 0x20, 0x60, 0x10, 0x50, 0x30, 0x70, - 0x08, 0x48, 0x28, 0x68, 0x18, 0x58, 0x38, 0x78, - 0x04, 0x44, 0x24, 0x64, 0x14, 0x54, 0x34, 0x74, -@@ -67,23 +101,42 @@ - 0x03, 0x23, 0x13, 0x33, 0x0b, 0x2b, 0x1b, 0x3b, - 0x07, 0x27, 0x17, 0x37, 0x0f, 0x2f, 0x1f, 0x3f}; ++ 0x00, 0x40, 0x20, 0x60, 0x10, 0x50, 0x30, 0x70, ++ 0x08, 0x48, 0x28, 0x68, 0x18, 0x58, 0x38, 0x78, ++ 0x04, 0x44, 0x24, 0x64, 0x14, 0x54, 0x34, 0x74, ++ 0x0c, 0x4c, 0x2c, 0x6c, 0x1c, 0x5c, 0x3c, 0x7c, ++ 0x02, 0x42, 0x22, 0x62, 0x12, 0x52, 0x32, 0x72, ++ 0x0a, 0x4a, 0x2a, 0x6a, 0x1a, 0x5a, 0x3a, 0x7a, ++ 0x06, 0x46, 0x26, 0x66, 0x16, 0x56, 0x36, 0x76, ++ 0x0e, 0x4e, 0x2e, 0x6e, 0x1e, 0x5e, 0x3e, 0x7e, ++ 0x01, 0x41, 0x21, 0x61, 0x11, 0x51, 0x31, 0x71, ++ 0x09, 0x49, 0x29, 0x69, 0x19, 0x59, 0x39, 0x79, ++ 0x05, 0x45, 0x25, 0x65, 0x15, 0x55, 0x35, 0x75, ++ 0x0d, 0x4d, 0x2d, 0x6d, 0x1d, 0x5d, 0x3d, 0x7d, ++ 0x03, 0x43, 0x23, 0x63, 0x13, 0x53, 0x33, 0x73, ++ 0x0b, 0x4b, 0x2b, 0x6b, 0x1b, 0x5b, 0x3b, 0x7b, ++ 0x07, 0x47, 0x27, 0x67, 0x17, 0x57, 0x37, 0x77, ++ 0x0f, 0x4f, 0x2f, 0x6f, 0x1f, 0x5f, 0x3f, 0x7f}; ++ + 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 +103,40 @@ + 6,134, 70,198, 38,166,230,102,246,118, 54,182, 22,150,214, 86 + }; --static complex_t buf[128]; -+#ifdef ARCH_X86 ++static complex_t __attribute__((aligned(16))) buf[128]; ++ ++/* Twiddle factor LUT */ ++static complex_t __attribute__((aligned(16))) w_1[1]; ++static complex_t __attribute__((aligned(16))) w_2[2]; ++static complex_t __attribute__((aligned(16))) w_4[4]; ++static complex_t __attribute__((aligned(16))) w_8[8]; ++static complex_t __attribute__((aligned(16))) w_16[16]; ++static complex_t __attribute__((aligned(16))) w_32[32]; ++static complex_t __attribute__((aligned(16))) w_64[64]; ++static complex_t __attribute__((aligned(16))) * w[7] = {w_1, w_2, w_4, w_8, w_16, w_32, w_64}; ++ ++/* Twiddle factors for IMDCT */ ++static sample_t __attribute__((aligned(16))) xcos1[128]; ++static sample_t __attribute__((aligned(16))) xsin1[128]; ++ ++#if defined(ARCH_X86) || defined(ARCH_X86_64) +// NOTE: SSE needs 16byte alignment or it will segfault +// -+static complex_t __attribute__((aligned(16))) buf[128]; +static float __attribute__((aligned(16))) sseSinCos1c[256]; +static float __attribute__((aligned(16))) sseSinCos1d[256]; +static float attribute_used __attribute__((aligned(16))) ps111_1[4]={1,1,1,-1}; @@ -1534,234 +1533,21 @@ +static float __attribute__((aligned(16))) *sseW[7]= + {NULL /*sseW0*/,sseW1,sseW2,sseW3,sseW4,sseW5,sseW6}; +static float __attribute__((aligned(16))) sseWindow[512]; -+#else -+static complex_t __attribute__((aligned(16))) buf[128]; -+#endif - - /* Twiddle factor LUT */ --static complex_t w_1[1]; --static complex_t w_2[2]; --static complex_t w_4[4]; --static complex_t w_8[8]; --static complex_t w_16[16]; --static complex_t w_32[32]; --static complex_t w_64[64]; --static complex_t * w[7] = {w_1, w_2, w_4, w_8, w_16, w_32, w_64}; -+static complex_t __attribute__((aligned(16))) w_1[1]; -+static complex_t __attribute__((aligned(16))) w_2[2]; -+static complex_t __attribute__((aligned(16))) w_4[4]; -+static complex_t __attribute__((aligned(16))) w_8[8]; -+static complex_t __attribute__((aligned(16))) w_16[16]; -+static complex_t __attribute__((aligned(16))) w_32[32]; -+static complex_t __attribute__((aligned(16))) w_64[64]; -+static complex_t __attribute__((aligned(16))) * w[7] = {w_1, w_2, w_4, w_8, w_16, w_32, w_64}; - - /* Twiddle factors for IMDCT */ --static sample_t xcos1[128]; --static sample_t xsin1[128]; --static sample_t xcos2[64]; --static sample_t xsin2[64]; -+static sample_t __attribute__((aligned(16))) xcos1[128]; -+static sample_t __attribute__((aligned(16))) xsin1[128]; -+static sample_t __attribute__((aligned(16))) xcos2[64]; -+static sample_t __attribute__((aligned(16))) xsin2[64]; - - /* Windowing function for Modified DCT - Thank you acroread */ - sample_t imdct_window[] = { -@@ -145,16 +198,19 @@ - void - imdct_do_512(sample_t data[],sample_t delay[], sample_t bias) - { -- int i,k; -+ int i; -+#ifndef USE_AC3_C -+ int k; - int p,q; - int m; - int two_m; - int two_m_plus_one; - -- sample_t tmp_a_i; -- sample_t tmp_a_r; - sample_t tmp_b_i; - sample_t tmp_b_r; -+#endif -+ sample_t tmp_a_i; -+ sample_t tmp_a_r; - - sample_t *data_ptr; - sample_t *delay_ptr; -@@ -162,22 +218,21 @@ - - /* 512 IMDCT with source and dest data in 'data' */ - -- /* Pre IFFT complex multiply plus IFFT cmplx conjugate */ -+ /* Pre IFFT complex multiply plus IFFT cmplx conjugate & reordering*/ - for( i=0; i < 128; i++) { - /* z[i] = (X[256-2*i-1] + j * X[2*i]) * (xcos1[i] + j * xsin1[i]) ; */ -- buf[i].real = (data[256-2*i-1] * xcos1[i]) - (data[2*i] * xsin1[i]); -- buf[i].imag = -1.0 * ((data[2*i] * xcos1[i]) + (data[256-2*i-1] * xsin1[i])); -- } -- -- /* Bit reversed shuffling */ -- for(i=0; i<128; i++) { -- k = bit_reverse_512[i]; -- if (k < i) -- swap_cmplx(&buf[i],&buf[k]); -+#ifdef USE_AC3_C -+ int j= pm128[i]; -+#else -+ int j= bit_reverse_512[i]; -+#endif -+ buf[i].real = (data[256-2*j-1] * xcos1[j]) - (data[2*j] * xsin1[j]); -+ buf[i].imag = -1.0 * ((data[2*j] * xcos1[j]) + (data[256-2*j-1] * xsin1[j])); ++#endif ++ + /* Root values for IFFT */ + static sample_t roots16[3]; + static sample_t roots32[7]; +@@ -245,7 +322,7 @@ + ifft_pass (buf, roots128 - 32, 32); } - /* FFT Merge */ -- for (m=0; m < 7; m++) { -+/* unoptimized variant -+ for (m=1; m < 7; m++) { - if(m) - two_m = (1 << m); - else -@@ -185,8 +240,8 @@ - - two_m_plus_one = (1 << (m+1)); - -- for(k = 0; k < two_m; k++) { -- for(i = 0; i < 128; i += two_m_plus_one) { -+ for(i = 0; i < 128; i += two_m_plus_one) { -+ for(k = 0; k < two_m; k++) { - p = k + i; - q = p + two_m; - tmp_a_r = buf[p].real; -@@ -200,7 +255,102 @@ - } - } - } -+*/ -+#ifdef USE_AC3_C -+ fft_128p (&buf[0]); -+#else -+ -+ /* 1. iteration */ -+ for(i = 0; i < 128; i += 2) { -+ tmp_a_r = buf[i].real; -+ tmp_a_i = buf[i].imag; -+ tmp_b_r = buf[i+1].real; -+ tmp_b_i = buf[i+1].imag; -+ buf[i].real = tmp_a_r + tmp_b_r; -+ buf[i].imag = tmp_a_i + tmp_b_i; -+ buf[i+1].real = tmp_a_r - tmp_b_r; -+ buf[i+1].imag = tmp_a_i - tmp_b_i; -+ } -+ -+ /* 2. iteration */ -+ // Note w[1]={{1,0}, {0,-1}} -+ for(i = 0; i < 128; i += 4) { -+ tmp_a_r = buf[i].real; -+ tmp_a_i = buf[i].imag; -+ tmp_b_r = buf[i+2].real; -+ tmp_b_i = buf[i+2].imag; -+ buf[i].real = tmp_a_r + tmp_b_r; -+ buf[i].imag = tmp_a_i + tmp_b_i; -+ buf[i+2].real = tmp_a_r - tmp_b_r; -+ buf[i+2].imag = tmp_a_i - tmp_b_i; -+ tmp_a_r = buf[i+1].real; -+ tmp_a_i = buf[i+1].imag; -+ tmp_b_r = buf[i+3].imag; -+ tmp_b_i = buf[i+3].real; -+ buf[i+1].real = tmp_a_r + tmp_b_r; -+ buf[i+1].imag = tmp_a_i - tmp_b_i; -+ buf[i+3].real = tmp_a_r - tmp_b_r; -+ buf[i+3].imag = tmp_a_i + tmp_b_i; -+ } - -+ /* 3. iteration */ -+ for(i = 0; i < 128; i += 8) { -+ tmp_a_r = buf[i].real; -+ tmp_a_i = buf[i].imag; -+ tmp_b_r = buf[i+4].real; -+ tmp_b_i = buf[i+4].imag; -+ buf[i].real = tmp_a_r + tmp_b_r; -+ buf[i].imag = tmp_a_i + tmp_b_i; -+ buf[i+4].real = tmp_a_r - tmp_b_r; -+ buf[i+4].imag = tmp_a_i - tmp_b_i; -+ tmp_a_r = buf[1+i].real; -+ tmp_a_i = buf[1+i].imag; -+ tmp_b_r = (buf[i+5].real + buf[i+5].imag) * w[2][1].real; -+ tmp_b_i = (buf[i+5].imag - buf[i+5].real) * w[2][1].real; -+ buf[1+i].real = tmp_a_r + tmp_b_r; -+ buf[1+i].imag = tmp_a_i + tmp_b_i; -+ buf[i+5].real = tmp_a_r - tmp_b_r; -+ buf[i+5].imag = tmp_a_i - tmp_b_i; -+ tmp_a_r = buf[i+2].real; -+ tmp_a_i = buf[i+2].imag; -+ tmp_b_r = buf[i+6].imag; -+ tmp_b_i = - buf[i+6].real; -+ buf[i+2].real = tmp_a_r + tmp_b_r; -+ buf[i+2].imag = tmp_a_i + tmp_b_i; -+ buf[i+6].real = tmp_a_r - tmp_b_r; -+ buf[i+6].imag = tmp_a_i - tmp_b_i; -+ tmp_a_r = buf[i+3].real; -+ tmp_a_i = buf[i+3].imag; -+ tmp_b_r = (buf[i+7].real - buf[i+7].imag) * w[2][3].imag; -+ tmp_b_i = (buf[i+7].imag + buf[i+7].real) * w[2][3].imag; -+ buf[i+3].real = tmp_a_r + tmp_b_r; -+ buf[i+3].imag = tmp_a_i + tmp_b_i; -+ buf[i+7].real = tmp_a_r - tmp_b_r; -+ buf[i+7].imag = tmp_a_i - tmp_b_i; -+ } -+ -+ /* 4-7. iterations */ -+ for (m=3; m < 7; m++) { -+ two_m = (1 << m); -+ -+ two_m_plus_one = two_m<<1; -+ -+ for(i = 0; i < 128; i += two_m_plus_one) { -+ for(k = 0; k < two_m; k++) { -+ int p = k + i; -+ int q = p + two_m; -+ tmp_a_r = buf[p].real; -+ tmp_a_i = buf[p].imag; -+ tmp_b_r = buf[q].real * w[m][k].real - buf[q].imag * w[m][k].imag; -+ tmp_b_i = buf[q].imag * w[m][k].real + buf[q].real * w[m][k].imag; -+ buf[p].real = tmp_a_r + tmp_b_r; -+ buf[p].imag = tmp_a_i + tmp_b_i; -+ buf[q].real = tmp_a_r - tmp_b_r; -+ buf[q].imag = tmp_a_i - tmp_b_i; -+ } -+ } -+ } -+#endif - /* Post IFFT complex multiply plus IFFT complex conjugate*/ - for( i=0; i < 128; i++) { - /* y[n] = z[n] * (xcos1[n] + j * xsin1[n]) ; */ -@@ -219,12 +369,12 @@ - *data_ptr++ = -buf[64+i].imag * *window_ptr++ + *delay_ptr++ + bias; - *data_ptr++ = buf[64-i-1].real * *window_ptr++ + *delay_ptr++ + bias; - } -- -+ - for(i=0; i< 64; i++) { - *data_ptr++ = -buf[i].real * *window_ptr++ + *delay_ptr++ + bias; - *data_ptr++ = buf[128-i-1].imag * *window_ptr++ + *delay_ptr++ + bias; - } -- -+ - /* The trailing edge of the window goes into the delay line */ - delay_ptr = delay; - -@@ -232,13 +382,717 @@ - *delay_ptr++ = -buf[64+i].real * *--window_ptr; - *delay_ptr++ = buf[64-i-1].imag * *--window_ptr; - } -- -+ - for(i=0; i<64; i++) { - *delay_ptr++ = buf[i].imag * *--window_ptr; - *delay_ptr++ = -buf[128-i-1].real * *--window_ptr; +-void a52_imdct_512 (sample_t * data, sample_t * delay, sample_t bias) ++void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias) + { + int i, k; + sample_t t_r, t_i, a_r, a_i, b_r, b_i, w_1, w_2; +@@ -289,6 +366,714 @@ } } @@ -1818,8 +1604,8 @@ + int k; + int p,q; + int m; -+ int two_m; -+ int two_m_plus_one; ++ long two_m; ++ long two_m_plus_one; + + sample_t tmp_b_i; + sample_t tmp_b_r; @@ -2092,7 +1878,7 @@ + + data_ptr = data; + delay_ptr = delay; -+ window_ptr = imdct_window; ++ window_ptr = a52_imdct_window; + + /* Window and convert to real valued signal */ + for(i=0; i< 64; i++) { @@ -2123,7 +1909,7 @@ + +// Stuff below this line is borrowed from libac3 +#include "srfftp.h" -+#ifdef ARCH_X86 ++#if defined(ARCH_X86) || defined(ARCH_X86_64) +#ifndef HAVE_3DNOW +#define HAVE_3DNOW 1 +#endif @@ -2144,8 +1930,10 @@ +/* int i,k; + int p,q;*/ + int m; -+ int two_m; -+ int two_m_plus_one; ++ long two_m; ++ long two_m_plus_one; ++ long two_m_plus_one_shl3; ++ complex_t *buf_offset; + +/* sample_t tmp_a_i; + sample_t tmp_a_r; @@ -2162,33 +1950,33 @@ + /* Pre IFFT complex multiply plus IFFT cmplx conjugate */ + /* Bit reversed shuffling */ + asm volatile( -+ "xorl %%esi, %%esi \n\t" -+ "leal "MANGLE(bit_reverse_512)", %%eax \n\t" -+ "movl $1008, %%edi \n\t" -+ "pushl %%ebp \n\t" //use ebp without telling gcc -+ ".balign 16 \n\t" ++ "xor %%"REG_S", %%"REG_S" \n\t" ++ "lea "MANGLE(bit_reverse_512)", %%"REG_a"\n\t" ++ "mov $1008, %%"REG_D" \n\t" ++ "push %%"REG_BP" \n\t" //use ebp without telling gcc ++ ASMALIGN16 + "1: \n\t" -+ "movlps (%0, %%esi), %%xmm0 \n\t" // XXXI -+ "movhps 8(%0, %%edi), %%xmm0 \n\t" // RXXI -+ "movlps 8(%0, %%esi), %%xmm1 \n\t" // XXXi -+ "movhps (%0, %%edi), %%xmm1 \n\t" // rXXi ++ "movlps (%0, %%"REG_S"), %%xmm0 \n\t" // XXXI ++ "movhps 8(%0, %%"REG_D"), %%xmm0 \n\t" // RXXI ++ "movlps 8(%0, %%"REG_S"), %%xmm1 \n\t" // XXXi ++ "movhps (%0, %%"REG_D"), %%xmm1 \n\t" // rXXi + "shufps $0x33, %%xmm1, %%xmm0 \n\t" // irIR -+ "movaps "MANGLE(sseSinCos1c)"(%%esi), %%xmm2\n\t" ++ "movaps "MANGLE(sseSinCos1c)"(%%"REG_S"), %%xmm2\n\t" + "mulps %%xmm0, %%xmm2 \n\t" + "shufps $0xB1, %%xmm0, %%xmm0 \n\t" // riRI -+ "mulps "MANGLE(sseSinCos1d)"(%%esi), %%xmm0\n\t" ++ "mulps "MANGLE(sseSinCos1d)"(%%"REG_S"), %%xmm0\n\t" + "subps %%xmm0, %%xmm2 \n\t" -+ "movzbl (%%eax), %%edx \n\t" -+ "movzbl 1(%%eax), %%ebp \n\t" -+ "movlps %%xmm2, (%1, %%edx,8) \n\t" -+ "movhps %%xmm2, (%1, %%ebp,8) \n\t" -+ "addl $16, %%esi \n\t" -+ "addl $2, %%eax \n\t" // avoid complex addressing for P4 crap -+ "subl $16, %%edi \n\t" ++ "movzb (%%"REG_a"), %%"REG_d" \n\t" ++ "movzb 1(%%"REG_a"), %%"REG_BP" \n\t" ++ "movlps %%xmm2, (%1, %%"REG_d", 8) \n\t" ++ "movhps %%xmm2, (%1, %%"REG_BP", 8) \n\t" ++ "add $16, %%"REG_S" \n\t" ++ "add $2, %%"REG_a" \n\t" // avoid complex addressing for P4 crap ++ "sub $16, %%"REG_D" \n\t" + " jnc 1b \n\t" -+ "popl %%ebp \n\t"//no we didnt touch ebp *g* ++ "pop %%"REG_BP" \n\t"//no we didnt touch ebp *g* + :: "b" (data), "c" (buf) -+ : "%esi", "%edi", "%eax", "%edx" ++ : "%"REG_S, "%"REG_D, "%"REG_a, "%"REG_d + ); + + @@ -2224,44 +2012,44 @@ + asm volatile( + "xorps %%xmm1, %%xmm1 \n\t" + "xorps %%xmm2, %%xmm2 \n\t" -+ "movl %0, %%esi \n\t" -+ ".balign 16 \n\t" ++ "mov %0, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movlps (%%esi), %%xmm0 \n\t" //buf[p] -+ "movlps 8(%%esi), %%xmm1\n\t" //buf[q] -+ "movhps (%%esi), %%xmm0 \n\t" //buf[p] -+ "movhps 8(%%esi), %%xmm2\n\t" //buf[q] ++ "movlps (%%"REG_S"), %%xmm0\n\t" //buf[p] ++ "movlps 8(%%"REG_S"), %%xmm1\n\t" //buf[q] ++ "movhps (%%"REG_S"), %%xmm0\n\t" //buf[p] ++ "movhps 8(%%"REG_S"), %%xmm2\n\t" //buf[q] + "addps %%xmm1, %%xmm0 \n\t" + "subps %%xmm2, %%xmm0 \n\t" -+ "movaps %%xmm0, (%%esi) \n\t" -+ "addl $16, %%esi \n\t" -+ "cmpl %1, %%esi \n\t" ++ "movaps %%xmm0, (%%"REG_S")\n\t" ++ "add $16, %%"REG_S" \n\t" ++ "cmp %1, %%"REG_S" \n\t" + " jb 1b \n\t" + :: "g" (buf), "r" (buf + 128) -+ : "%esi" ++ : "%"REG_S + ); + + /* 2. iteration */ + // Note w[1]={{1,0}, {0,-1}} + asm volatile( + "movaps "MANGLE(ps111_1)", %%xmm7\n\t" // 1,1,1,-1 -+ "movl %0, %%esi \n\t" -+ ".balign 16 \n\t" ++ "mov %0, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 16(%%esi), %%xmm2 \n\t" //r2,i2,r3,i3 ++ "movaps 16(%%"REG_S"), %%xmm2 \n\t" //r2,i2,r3,i3 + "shufps $0xB4, %%xmm2, %%xmm2 \n\t" //r2,i2,i3,r3 + "mulps %%xmm7, %%xmm2 \n\t" //r2,i2,i3,-r3 -+ "movaps (%%esi), %%xmm0 \n\t" //r0,i0,r1,i1 -+ "movaps (%%esi), %%xmm1 \n\t" //r0,i0,r1,i1 ++ "movaps (%%"REG_S"), %%xmm0 \n\t" //r0,i0,r1,i1 ++ "movaps (%%"REG_S"), %%xmm1 \n\t" //r0,i0,r1,i1 + "addps %%xmm2, %%xmm0 \n\t" + "subps %%xmm2, %%xmm1 \n\t" -+ "movaps %%xmm0, (%%esi) \n\t" -+ "movaps %%xmm1, 16(%%esi) \n\t" -+ "addl $32, %%esi \n\t" -+ "cmpl %1, %%esi \n\t" ++ "movaps %%xmm0, (%%"REG_S") \n\t" ++ "movaps %%xmm1, 16(%%"REG_S") \n\t" ++ "add $32, %%"REG_S" \n\t" ++ "cmp %1, %%"REG_S" \n\t" + " jb 1b \n\t" + :: "g" (buf), "r" (buf + 128) -+ : "%esi" ++ : "%"REG_S + ); + + /* 3. iteration */ @@ -2276,11 +2064,11 @@ + "movaps 16+"MANGLE(sseW2)", %%xmm7\n\t" + "xorps %%xmm5, %%xmm5 \n\t" + "xorps %%xmm2, %%xmm2 \n\t" -+ "movl %0, %%esi \n\t" -+ ".balign 16 \n\t" ++ "mov %0, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps 32(%%esi), %%xmm2 \n\t" //r4,i4,r5,i5 -+ "movaps 48(%%esi), %%xmm3 \n\t" //r6,i6,r7,i7 ++ "movaps 32(%%"REG_S"), %%xmm2 \n\t" //r4,i4,r5,i5 ++ "movaps 48(%%"REG_S"), %%xmm3 \n\t" //r6,i6,r7,i7 + "movaps "MANGLE(sseW2)", %%xmm4 \n\t" //r4,i4,r5,i5 + "movaps 32+"MANGLE(sseW2)", %%xmm5\n\t" //r6,i6,r7,i7 + "mulps %%xmm2, %%xmm4 \n\t" @@ -2289,8 +2077,8 @@ + "shufps $0xB1, %%xmm3, %%xmm3 \n\t" //i6,r6,i7,r7 + "mulps %%xmm6, %%xmm3 \n\t" + "mulps %%xmm7, %%xmm2 \n\t" -+ "movaps (%%esi), %%xmm0 \n\t" //r0,i0,r1,i1 -+ "movaps 16(%%esi), %%xmm1 \n\t" //r2,i2,r3,i3 ++ "movaps (%%"REG_S"), %%xmm0 \n\t" //r0,i0,r1,i1 ++ "movaps 16(%%"REG_S"), %%xmm1 \n\t" //r2,i2,r3,i3 + "addps %%xmm4, %%xmm2 \n\t" + "addps %%xmm5, %%xmm3 \n\t" + "movaps %%xmm2, %%xmm4 \n\t" @@ -2299,125 +2087,127 @@ + "addps %%xmm1, %%xmm3 \n\t" + "subps %%xmm4, %%xmm0 \n\t" + "subps %%xmm5, %%xmm1 \n\t" -+ "movaps %%xmm2, (%%esi) \n\t" -+ "movaps %%xmm3, 16(%%esi) \n\t" -+ "movaps %%xmm0, 32(%%esi) \n\t" -+ "movaps %%xmm1, 48(%%esi) \n\t" -+ "addl $64, %%esi \n\t" -+ "cmpl %1, %%esi \n\t" ++ "movaps %%xmm2, (%%"REG_S") \n\t" ++ "movaps %%xmm3, 16(%%"REG_S") \n\t" ++ "movaps %%xmm0, 32(%%"REG_S") \n\t" ++ "movaps %%xmm1, 48(%%"REG_S") \n\t" ++ "add $64, %%"REG_S" \n\t" ++ "cmp %1, %%"REG_S" \n\t" + " jb 1b \n\t" + :: "g" (buf), "r" (buf + 128) -+ : "%esi" ++ : "%"REG_S + ); + + /* 4-7. iterations */ + for (m=3; m < 7; m++) { + two_m = (1 << m); + two_m_plus_one = two_m<<1; ++ two_m_plus_one_shl3 = (two_m_plus_one<<3); ++ buf_offset = buf+128; + asm volatile( -+ "movl %0, %%esi \n\t" -+ ".balign 16 \n\t" ++ "mov %0, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "xorl %%edi, %%edi \n\t" // k -+ "leal (%%esi, %3), %%edx \n\t" ++ "xor %%"REG_D", %%"REG_D" \n\t" // k ++ "lea (%%"REG_S", %3), %%"REG_d" \n\t" + "2: \n\t" -+ "movaps (%%edx, %%edi), %%xmm1 \n\t" -+ "movaps (%4, %%edi, 2), %%xmm2 \n\t" ++ "movaps (%%"REG_d", %%"REG_D"), %%xmm1 \n\t" ++ "movaps (%4, %%"REG_D", 2), %%xmm2 \n\t" + "mulps %%xmm1, %%xmm2 \n\t" + "shufps $0xB1, %%xmm1, %%xmm1 \n\t" -+ "mulps 16(%4, %%edi, 2), %%xmm1 \n\t" -+ "movaps (%%esi, %%edi), %%xmm0 \n\t" ++ "mulps 16(%4, %%"REG_D", 2), %%xmm1 \n\t" ++ "movaps (%%"REG_S", %%"REG_D"), %%xmm0 \n\t" + "addps %%xmm2, %%xmm1 \n\t" + "movaps %%xmm1, %%xmm2 \n\t" + "addps %%xmm0, %%xmm1 \n\t" + "subps %%xmm2, %%xmm0 \n\t" -+ "movaps %%xmm1, (%%esi, %%edi) \n\t" -+ "movaps %%xmm0, (%%edx, %%edi) \n\t" -+ "addl $16, %%edi \n\t" -+ "cmpl %3, %%edi \n\t" //FIXME (opt) count against 0 ++ "movaps %%xmm1, (%%"REG_S", %%"REG_D") \n\t" ++ "movaps %%xmm0, (%%"REG_d", %%"REG_D") \n\t" ++ "add $16, %%"REG_D" \n\t" ++ "cmp %3, %%"REG_D" \n\t" //FIXME (opt) count against 0 + " jb 2b \n\t" -+ "addl %2, %%esi \n\t" -+ "cmpl %1, %%esi \n\t" ++ "add %2, %%"REG_S" \n\t" ++ "cmp %1, %%"REG_S" \n\t" + " jb 1b \n\t" -+ :: "g" (buf), "m" (buf+128), "m" (two_m_plus_one<<3), "r" (two_m<<3), ++ :: "g" (buf), "m" (buf_offset), "m" (two_m_plus_one_shl3), "r" (two_m<<3), + "r" (sseW[m]) -+ : "%esi", "%edi", "%edx" ++ : "%"REG_S, "%"REG_D, "%"REG_d + ); + } + + /* Post IFFT complex multiply plus IFFT complex conjugate*/ + asm volatile( -+ "movl $-1024, %%esi \n\t" -+ ".balign 16 \n\t" ++ "mov $-1024, %%"REG_S" \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movaps (%0, %%esi), %%xmm0 \n\t" -+ "movaps (%0, %%esi), %%xmm1 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm0 \n\t" ++ "movaps (%0, %%"REG_S"), %%xmm1 \n\t" + "shufps $0xB1, %%xmm0, %%xmm0 \n\t" -+ "mulps 1024+"MANGLE(sseSinCos1c)"(%%esi), %%xmm1\n\t" -+ "mulps 1024+"MANGLE(sseSinCos1d)"(%%esi), %%xmm0\n\t" ++ "mulps 1024+"MANGLE(sseSinCos1c)"(%%"REG_S"), %%xmm1\n\t" ++ "mulps 1024+"MANGLE(sseSinCos1d)"(%%"REG_S"), %%xmm0\n\t" + "addps %%xmm1, %%xmm0 \n\t" -+ "movaps %%xmm0, (%0, %%esi) \n\t" -+ "addl $16, %%esi \n\t" ++ "movaps %%xmm0, (%0, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" + " jnz 1b \n\t" + :: "r" (buf+128) -+ : "%esi" ++ : "%"REG_S + ); + + + data_ptr = data; + delay_ptr = delay; -+ window_ptr = imdct_window; ++ window_ptr = a52_imdct_window; + + /* Window and convert to real valued signal */ + asm volatile( -+ "xorl %%edi, %%edi \n\t" // 0 -+ "xorl %%esi, %%esi \n\t" // 0 ++ "xor %%"REG_D", %%"REG_D" \n\t" // 0 ++ "xor %%"REG_S", %%"REG_S" \n\t" // 0 + "movss %3, %%xmm2 \n\t" // bias + "shufps $0x00, %%xmm2, %%xmm2 \n\t" // bias, bias, ... -+ ".balign 16 \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movlps (%0, %%esi), %%xmm0 \n\t" // ? ? A ? -+ "movlps 8(%0, %%esi), %%xmm1 \n\t" // ? ? C ? -+ "movhps -16(%0, %%edi), %%xmm1 \n\t" // ? D C ? -+ "movhps -8(%0, %%edi), %%xmm0 \n\t" // ? B A ? ++ "movlps (%0, %%"REG_S"), %%xmm0 \n\t" // ? ? A ? ++ "movlps 8(%0, %%"REG_S"), %%xmm1 \n\t" // ? ? C ? ++ "movhps -16(%0, %%"REG_D"), %%xmm1 \n\t" // ? D C ? ++ "movhps -8(%0, %%"REG_D"), %%xmm0 \n\t" // ? B A ? + "shufps $0x99, %%xmm1, %%xmm0 \n\t" // D C B A -+ "mulps "MANGLE(sseWindow)"(%%esi), %%xmm0\n\t" -+ "addps (%2, %%esi), %%xmm0 \n\t" ++ "mulps "MANGLE(sseWindow)"(%%"REG_S"), %%xmm0\n\t" ++ "addps (%2, %%"REG_S"), %%xmm0 \n\t" + "addps %%xmm2, %%xmm0 \n\t" -+ "movaps %%xmm0, (%1, %%esi) \n\t" -+ "addl $16, %%esi \n\t" -+ "subl $16, %%edi \n\t" -+ "cmpl $512, %%esi \n\t" ++ "movaps %%xmm0, (%1, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" ++ "sub $16, %%"REG_D" \n\t" ++ "cmp $512, %%"REG_S" \n\t" + " jb 1b \n\t" + :: "r" (buf+64), "r" (data_ptr), "r" (delay_ptr), "m" (bias) -+ : "%esi", "%edi" ++ : "%"REG_S, "%"REG_D + ); + data_ptr+=128; + delay_ptr+=128; +// window_ptr+=128; + + asm volatile( -+ "movl $1024, %%edi \n\t" // 512 -+ "xorl %%esi, %%esi \n\t" // 0 ++ "mov $1024, %%"REG_D" \n\t" // 512 ++ "xor %%"REG_S", %%"REG_S" \n\t" // 0 + "movss %3, %%xmm2 \n\t" // bias + "shufps $0x00, %%xmm2, %%xmm2 \n\t" // bias, bias, ... -+ ".balign 16 \n\t" ++ ASMALIGN16 + "1: \n\t" -+ "movlps (%0, %%esi), %%xmm0 \n\t" // ? ? ? A -+ "movlps 8(%0, %%esi), %%xmm1 \n\t" // ? ? ? C -+ "movhps -16(%0, %%edi), %%xmm1 \n\t" // D ? ? C -+ "movhps -8(%0, %%edi), %%xmm0 \n\t" // B ? ? A ++ "movlps (%0, %%"REG_S"), %%xmm0 \n\t" // ? ? ? A ++ "movlps 8(%0, %%"REG_S"), %%xmm1 \n\t" // ? ? ? C ++ "movhps -16(%0, %%"REG_D"), %%xmm1 \n\t" // D ? ? C ++ "movhps -8(%0, %%"REG_D"), %%xmm0 \n\t" // B ? ? A + "shufps $0xCC, %%xmm1, %%xmm0 \n\t" // D C B A -+ "mulps 512+"MANGLE(sseWindow)"(%%esi), %%xmm0\n\t" -+ "addps (%2, %%esi), %%xmm0 \n\t" ++ "mulps 512+"MANGLE(sseWindow)"(%%"REG_S"), %%xmm0\n\t" ++ "addps (%2, %%"REG_S"), %%xmm0 \n\t" + "addps %%xmm2, %%xmm0 \n\t" -+ "movaps %%xmm0, (%1, %%esi) \n\t" -+ "addl $16, %%esi \n\t" -+ "subl $16, %%edi \n\t" -+ "cmpl $512, %%esi \n\t" ++ "movaps %%xmm0, (%1, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" ++ "sub $16, %%"REG_D" \n\t" ++ "cmp $512, %%"REG_S" \n\t" + " jb 1b \n\t" + :: "r" (buf), "r" (data_ptr), "r" (delay_ptr), "m" (bias) -+ : "%esi", "%edi" ++ : "%"REG_S, "%"REG_D + ); + data_ptr+=128; +// window_ptr+=128; @@ -2426,79 +2216,83 @@ + delay_ptr = delay; + + asm volatile( -+ "xorl %%edi, %%edi \n\t" // 0 -+ "xorl %%esi, %%esi \n\t" // 0 -+ ".balign 16 \n\t" ++ "xor %%"REG_D", %%"REG_D" \n\t" // 0 ++ "xor %%"REG_S", %%"REG_S" \n\t" // 0 ++ ASMALIGN16 + "1: \n\t" -+ "movlps (%0, %%esi), %%xmm0 \n\t" // ? ? ? A -+ "movlps 8(%0, %%esi), %%xmm1 \n\t" // ? ? ? C -+ "movhps -16(%0, %%edi), %%xmm1 \n\t" // D ? ? C -+ "movhps -8(%0, %%edi), %%xmm0 \n\t" // B ? ? A ++ "movlps (%0, %%"REG_S"), %%xmm0 \n\t" // ? ? ? A ++ "movlps 8(%0, %%"REG_S"), %%xmm1 \n\t" // ? ? ? C ++ "movhps -16(%0, %%"REG_D"), %%xmm1 \n\t" // D ? ? C ++ "movhps -8(%0, %%"REG_D"), %%xmm0 \n\t" // B ? ? A + "shufps $0xCC, %%xmm1, %%xmm0 \n\t" // D C B A -+ "mulps 1024+"MANGLE(sseWindow)"(%%esi), %%xmm0\n\t" -+ "movaps %%xmm0, (%1, %%esi) \n\t" -+ "addl $16, %%esi \n\t" -+ "subl $16, %%edi \n\t" -+ "cmpl $512, %%esi \n\t" ++ "mulps 1024+"MANGLE(sseWindow)"(%%"REG_S"), %%xmm0\n\t" ++ "movaps %%xmm0, (%1, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" ++ "sub $16, %%"REG_D" \n\t" ++ "cmp $512, %%"REG_S" \n\t" + " jb 1b \n\t" + :: "r" (buf+64), "r" (delay_ptr) -+ : "%esi", "%edi" ++ : "%"REG_S, "%"REG_D + ); + delay_ptr+=128; +// window_ptr-=128; + + asm volatile( -+ "movl $1024, %%edi \n\t" // 1024 -+ "xorl %%esi, %%esi \n\t" // 0 -+ ".balign 16 \n\t" ++ "mov $1024, %%"REG_D" \n\t" // 1024 ++ "xor %%"REG_S", %%"REG_S" \n\t" // 0 ++ ASMALIGN16 + "1: \n\t" -+ "movlps (%0, %%esi), %%xmm0 \n\t" // ? ? A ? -+ "movlps 8(%0, %%esi), %%xmm1 \n\t" // ? ? C ? -+ "movhps -16(%0, %%edi), %%xmm1 \n\t" // ? D C ? -+ "movhps -8(%0, %%edi), %%xmm0 \n\t" // ? B A ? ++ "movlps (%0, %%"REG_S"), %%xmm0 \n\t" // ? ? A ? ++ "movlps 8(%0, %%"REG_S"), %%xmm1 \n\t" // ? ? C ? ++ "movhps -16(%0, %%"REG_D"), %%xmm1 \n\t" // ? D C ? ++ "movhps -8(%0, %%"REG_D"), %%xmm0 \n\t" // ? B A ? + "shufps $0x99, %%xmm1, %%xmm0 \n\t" // D C B A -+ "mulps 1536+"MANGLE(sseWindow)"(%%esi), %%xmm0\n\t" -+ "movaps %%xmm0, (%1, %%esi) \n\t" -+ "addl $16, %%esi \n\t" -+ "subl $16, %%edi \n\t" -+ "cmpl $512, %%esi \n\t" ++ "mulps 1536+"MANGLE(sseWindow)"(%%"REG_S"), %%xmm0\n\t" ++ "movaps %%xmm0, (%1, %%"REG_S") \n\t" ++ "add $16, %%"REG_S" \n\t" ++ "sub $16, %%"REG_D" \n\t" ++ "cmp $512, %%"REG_S" \n\t" + " jb 1b \n\t" + :: "r" (buf), "r" (delay_ptr) -+ : "%esi", "%edi" ++ : "%"REG_S, "%"REG_D + ); +} -+#endif //arch_x86 ++#endif // ARCH_X86 || ARCH_X86_64 + - void - imdct_do_256(sample_t data[],sample_t delay[],sample_t bias) + void a52_imdct_256(sample_t * data, sample_t * delay, sample_t bias) { -@@ -379,13 +1233,19 @@ + int i, k; +@@ -368,7 +1153,7 @@ + + void a52_imdct_init (uint32_t mm_accel) { - int i, j, k; +- int i, k; ++ int i, j, k; + double sum; -- fprintf (stderr, "No accelerated IMDCT transform found\n"); -- - /* Twiddle factors to turn IFFT into IMDCT */ - for (i = 0; i < 128; i++) { - xcos1[i] = -cos ((M_PI / 2048) * (8 * i + 1)); - xsin1[i] = -sin ((M_PI / 2048) * (8 * i + 1)); + /* compute imdct window - kaiser-bessel derived window, alpha = 5.0 */ +@@ -420,6 +1205,99 @@ + post2[i].real = cos ((M_PI / 128) * (i + 0.5)); + post2[i].imag = sin ((M_PI / 128) * (i + 0.5)); } -+#ifdef ARCH_X86 ++ for (i = 0; i < 128; i++) { ++ xcos1[i] = -cos ((M_PI / 2048) * (8 * i + 1)); ++ xsin1[i] = -sin ((M_PI / 2048) * (8 * i + 1)); ++ } ++ for (i = 0; i < 7; i++) { ++ j = 1 << i; ++ for (k = 0; k < j; k++) { ++ w[i][k].real = cos (-M_PI * k / j); ++ w[i][k].imag = sin (-M_PI * k / j); ++ } ++ } ++#if defined(ARCH_X86) || defined(ARCH_X86_64) + for (i = 0; i < 128; i++) { + sseSinCos1c[2*i+0]= xcos1[i]; + sseSinCos1c[2*i+1]= -xcos1[i]; + sseSinCos1d[2*i+0]= xsin1[i]; + sseSinCos1d[2*i+1]= xsin1[i]; + } -+#endif - - /* More twiddle factors to turn IFFT into IMDCT */ - for (i = 0; i < 64; i++) { -@@ -400,7 +1260,334 @@ - w[i][k].imag = sin (-M_PI * k / j); - } - } -+#ifdef ARCH_X86 + for (i = 1; i < 7; i++) { + j = 1 << i; + for (k = 0; k < j; k+=2) { @@ -2530,351 +2324,64 @@ + + for(i=0; i<128; i++) + { -+ sseWindow[2*i+0]= -imdct_window[2*i+0]; -+ sseWindow[2*i+1]= imdct_window[2*i+1]; ++ sseWindow[2*i+0]= -a52_imdct_window[2*i+0]; ++ sseWindow[2*i+1]= a52_imdct_window[2*i+1]; + } + + for(i=0; i<64; i++) + { -+ sseWindow[256 + 2*i+0]= -imdct_window[254 - 2*i+1]; -+ sseWindow[256 + 2*i+1]= imdct_window[254 - 2*i+0]; -+ sseWindow[384 + 2*i+0]= imdct_window[126 - 2*i+1]; -+ sseWindow[384 + 2*i+1]= -imdct_window[126 - 2*i+0]; ++ sseWindow[256 + 2*i+0]= -a52_imdct_window[254 - 2*i+1]; ++ sseWindow[256 + 2*i+1]= a52_imdct_window[254 - 2*i+0]; ++ sseWindow[384 + 2*i+0]= a52_imdct_window[126 - 2*i+1]; ++ sseWindow[384 + 2*i+1]= -a52_imdct_window[126 - 2*i+0]; + } -+#endif // arch_x86 ++#endif ++ a52_imdct_512 = imdct_do_512; ++ ifft128 = ifft128_c; ++ ifft64 = ifft64_c; + - imdct_512 = imdct_do_512; -+#ifdef ARCH_X86 ++#if defined(ARCH_X86) || defined(ARCH_X86_64) + if(mm_accel & MM_ACCEL_X86_SSE) + { + fprintf (stderr, "Using SSE optimized IMDCT transform\n"); -+ imdct_512 = imdct_do_512_sse; ++ a52_imdct_512 = imdct_do_512_sse; + } + else + if(mm_accel & MM_ACCEL_X86_3DNOWEXT) + { + fprintf (stderr, "Using 3DNowEx optimized IMDCT transform\n"); -+ imdct_512 = imdct_do_512_3dnowex; ++ a52_imdct_512 = imdct_do_512_3dnowex; + } + else + if(mm_accel & MM_ACCEL_X86_3DNOW) + { + fprintf (stderr, "Using 3DNow optimized IMDCT transform\n"); -+ imdct_512 = imdct_do_512_3dnow; ++ a52_imdct_512 = imdct_do_512_3dnow; + } + else -+#endif // arch_x86 ++#endif // ARCH_X86 || ARCH_X86_64 +#ifdef HAVE_ALTIVEC + if (mm_accel & MM_ACCEL_PPC_ALTIVEC) + { + fprintf(stderr, "Using AltiVec optimized IMDCT transform\n"); -+ imdct_512 = imdct_do_512_altivec; ++ a52_imdct_512 = imdct_do_512_altivec; + } + else +#endif -+ fprintf (stderr, "No accelerated IMDCT transform found\n"); - imdct_256 = imdct_do_256; + + #ifdef LIBA52_DJBFFT + if (mm_accel & MM_ACCEL_DJBFFT) { +@@ -430,7 +1308,5 @@ + #endif + { + fprintf (stderr, "No accelerated IMDCT transform found\n"); +- ifft128 = ifft128_c; +- ifft64 = ifft64_c; } } -+ -+static void fft_asmb(int k, complex_t *x, complex_t *wTB, -+ const complex_t *d, const complex_t *d_3) -+{ -+ register complex_t *x2k, *x3k, *x4k, *wB; -+ register float a_r, a_i, a1_r, a1_i, u_r, u_i, v_r, v_i; -+ -+ x2k = x + 2 * k; -+ x3k = x2k + 2 * k; -+ x4k = x3k + 2 * k; -+ wB = wTB + 2 * k; -+ -+ TRANSZERO(x[0],x2k[0],x3k[0],x4k[0]); -+ TRANS(x[1],x2k[1],x3k[1],x4k[1],wTB[1],wB[1],d[1],d_3[1]); -+ -+ --k; -+ for(;;) { -+ TRANS(x[2],x2k[2],x3k[2],x4k[2],wTB[2],wB[2],d[2],d_3[2]); -+ TRANS(x[3],x2k[3],x3k[3],x4k[3],wTB[3],wB[3],d[3],d_3[3]); -+ if (!--k) break; -+ x += 2; -+ x2k += 2; -+ x3k += 2; -+ x4k += 2; -+ d += 2; -+ d_3 += 2; -+ wTB += 2; -+ wB += 2; -+ } -+ -+} -+ -+static void fft_asmb16(complex_t *x, complex_t *wTB) -+{ -+ register float a_r, a_i, a1_r, a1_i, u_r, u_i, v_r, v_i; -+ int k = 2; -+ -+ /* transform x[0], x[8], x[4], x[12] */ -+ TRANSZERO(x[0],x[4],x[8],x[12]); -+ -+ /* transform x[1], x[9], x[5], x[13] */ -+ TRANS(x[1],x[5],x[9],x[13],wTB[1],wTB[5],delta16[1],delta16_3[1]); -+ -+ /* transform x[2], x[10], x[6], x[14] */ -+ TRANSHALF_16(x[2],x[6],x[10],x[14]); -+ -+ /* transform x[3], x[11], x[7], x[15] */ -+ TRANS(x[3],x[7],x[11],x[15],wTB[3],wTB[7],delta16[3],delta16_3[3]); -+ -+} -+ -+static void fft_4(complex_t *x) -+{ -+ /* delta_p = 1 here */ -+ /* x[k] = sum_{i=0..3} x[i] * w^{i*k}, w=e^{-2*pi/4} -+ */ -+ -+ register float yt_r, yt_i, yb_r, yb_i, u_r, u_i, vi_r, vi_i; -+ -+ yt_r = x[0].real; -+ yb_r = yt_r - x[2].real; -+ yt_r += x[2].real; -+ -+ u_r = x[1].real; -+ vi_i = x[3].real - u_r; -+ u_r += x[3].real; -+ -+ u_i = x[1].imag; -+ vi_r = u_i - x[3].imag; -+ u_i += x[3].imag; -+ -+ yt_i = yt_r; -+ yt_i += u_r; -+ x[0].real = yt_i; -+ yt_r -= u_r; -+ x[2].real = yt_r; -+ yt_i = yb_r; -+ yt_i += vi_r; -+ x[1].real = yt_i; -+ yb_r -= vi_r; -+ x[3].real = yb_r; -+ -+ yt_i = x[0].imag; -+ yb_i = yt_i - x[2].imag; -+ yt_i += x[2].imag; -+ -+ yt_r = yt_i; -+ yt_r += u_i; -+ x[0].imag = yt_r; -+ yt_i -= u_i; -+ x[2].imag = yt_i; -+ yt_r = yb_i; -+ yt_r += vi_i; -+ x[1].imag = yt_r; -+ yb_i -= vi_i; -+ x[3].imag = yb_i; -+} -+ -+ -+static void fft_8(complex_t *x) -+{ -+ /* delta_p = diag{1, sqrt(i)} here */ -+ /* x[k] = sum_{i=0..7} x[i] * w^{i*k}, w=e^{-2*pi/8} -+ */ -+ register float wT1_r, wT1_i, wB1_r, wB1_i, wT2_r, wT2_i, wB2_r, wB2_i; -+ -+ wT1_r = x[1].real; -+ wT1_i = x[1].imag; -+ wB1_r = x[3].real; -+ wB1_i = x[3].imag; -+ -+ x[1] = x[2]; -+ x[2] = x[4]; -+ x[3] = x[6]; -+ fft_4(&x[0]); -+ -+ -+ /* x[0] x[4] */ -+ wT2_r = x[5].real; -+ wT2_r += x[7].real; -+ wT2_r += wT1_r; -+ wT2_r += wB1_r; -+ wT2_i = wT2_r; -+ wT2_r += x[0].real; -+ wT2_i = x[0].real - wT2_i; -+ x[0].real = wT2_r; -+ x[4].real = wT2_i; -+ -+ wT2_i = x[5].imag; -+ wT2_i += x[7].imag; -+ wT2_i += wT1_i; -+ wT2_i += wB1_i; -+ wT2_r = wT2_i; -+ wT2_r += x[0].imag; -+ wT2_i = x[0].imag - wT2_i; -+ x[0].imag = wT2_r; -+ x[4].imag = wT2_i; -+ -+ /* x[2] x[6] */ -+ wT2_r = x[5].imag; -+ wT2_r -= x[7].imag; -+ wT2_r += wT1_i; -+ wT2_r -= wB1_i; -+ wT2_i = wT2_r; -+ wT2_r += x[2].real; -+ wT2_i = x[2].real - wT2_i; -+ x[2].real = wT2_r; -+ x[6].real = wT2_i; -+ -+ wT2_i = x[5].real; -+ wT2_i -= x[7].real; -+ wT2_i += wT1_r; -+ wT2_i -= wB1_r; -+ wT2_r = wT2_i; -+ wT2_r += x[2].imag; -+ wT2_i = x[2].imag - wT2_i; -+ x[2].imag = wT2_i; -+ x[6].imag = wT2_r; -+ -+ -+ /* x[1] x[5] */ -+ wT2_r = wT1_r; -+ wT2_r += wB1_i; -+ wT2_r -= x[5].real; -+ wT2_r -= x[7].imag; -+ wT2_i = wT1_i; -+ wT2_i -= wB1_r; -+ wT2_i -= x[5].imag; -+ wT2_i += x[7].real; -+ -+ wB2_r = wT2_r; -+ wB2_r += wT2_i; -+ wT2_i -= wT2_r; -+ wB2_r *= HSQRT2; -+ wT2_i *= HSQRT2; -+ wT2_r = wB2_r; -+ wB2_r += x[1].real; -+ wT2_r = x[1].real - wT2_r; -+ -+ wB2_i = x[5].real; -+ x[1].real = wB2_r; -+ x[5].real = wT2_r; -+ -+ wT2_r = wT2_i; -+ wT2_r += x[1].imag; -+ wT2_i = x[1].imag - wT2_i; -+ wB2_r = x[5].imag; -+ x[1].imag = wT2_r; -+ x[5].imag = wT2_i; -+ -+ /* x[3] x[7] */ -+ wT1_r -= wB1_i; -+ wT1_i += wB1_r; -+ wB1_r = wB2_i - x[7].imag; -+ wB1_i = wB2_r + x[7].real; -+ wT1_r -= wB1_r; -+ wT1_i -= wB1_i; -+ wB1_r = wT1_r + wT1_i; -+ wB1_r *= HSQRT2; -+ wT1_i -= wT1_r; -+ wT1_i *= HSQRT2; -+ wB2_r = x[3].real; -+ wB2_i = wB2_r + wT1_i; -+ wB2_r -= wT1_i; -+ x[3].real = wB2_i; -+ x[7].real = wB2_r; -+ wB2_i = x[3].imag; -+ wB2_r = wB2_i + wB1_r; -+ wB2_i -= wB1_r; -+ x[3].imag = wB2_i; -+ x[7].imag = wB2_r; -+} -+ -+ -+static void fft_128p(complex_t *a) -+{ -+ fft_8(&a[0]); fft_4(&a[8]); fft_4(&a[12]); -+ fft_asmb16(&a[0], &a[8]); -+ -+ fft_8(&a[16]), fft_8(&a[24]); -+ fft_asmb(4, &a[0], &a[16],&delta32[0], &delta32_3[0]); -+ -+ fft_8(&a[32]); fft_4(&a[40]); fft_4(&a[44]); -+ fft_asmb16(&a[32], &a[40]); -+ -+ fft_8(&a[48]); fft_4(&a[56]); fft_4(&a[60]); -+ fft_asmb16(&a[48], &a[56]); -+ -+ fft_asmb(8, &a[0], &a[32],&delta64[0], &delta64_3[0]); -+ -+ fft_8(&a[64]); fft_4(&a[72]); fft_4(&a[76]); -+ /* fft_16(&a[64]); */ -+ fft_asmb16(&a[64], &a[72]); -+ -+ fft_8(&a[80]); fft_8(&a[88]); -+ -+ /* fft_32(&a[64]); */ -+ fft_asmb(4, &a[64], &a[80],&delta32[0], &delta32_3[0]); -+ -+ fft_8(&a[96]); fft_4(&a[104]), fft_4(&a[108]); -+ /* fft_16(&a[96]); */ -+ fft_asmb16(&a[96], &a[104]); -+ -+ fft_8(&a[112]), fft_8(&a[120]); -+ /* fft_32(&a[96]); */ -+ fft_asmb(4, &a[96], &a[112], &delta32[0], &delta32_3[0]); -+ -+ /* fft_128(&a[0]); */ -+ fft_asmb(16, &a[0], &a[64], &delta128[0], &delta128_3[0]); -+} -+ -+ -+ ---- liba52/imdct_mlib.c 2005-03-22 19:59:35.000000000 +0100 -+++ imdct_mlib.c 2004-03-19 01:15:51.000000000 +0100 -@@ -23,11 +29,11 @@ - - #ifdef LIBA52_MLIB - --#include <inttypes.h> --#include <string.h> - #include <mlib_types.h> - #include <mlib_status.h> - #include <mlib_signal.h> -+#include <string.h> -+#include <inttypes.h> - - #include "a52.h" - #include "a52_internal.h" -@@ -42,7 +48,7 @@ - sample_t *data_ptr; - sample_t *delay_ptr; - sample_t *window_ptr; -- sample_t tmp[256] __attribute__ ((__aligned__ (16))); -+ sample_t tmp[256] __attribute__((aligned(16))); - int i; - - memcpy(tmp, data, 256 * sizeof(sample_t)); -@@ -91,7 +97,7 @@ - sample_t *data_ptr; - sample_t *delay_ptr; - sample_t *window_ptr; -- sample_t tmp[256] __attribute__ ((__aligned__ (16))); -+ sample_t tmp[256] __attribute__((aligned(16))); - int i; - - memcpy(tmp, data, 256 * sizeof(sample_t)); ---- include/mm_accel.h 2005-03-22 19:58:53.000000000 +0100 -+++ mm_accel.h 2004-03-19 01:15:52.000000000 +0100 -@@ -19,12 +25,22 @@ - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA - */ - -+#ifndef MM_ACCEL_H -+#define MM_ACCEL_H -+ - /* generic accelerations */ - #define MM_ACCEL_MLIB 0x00000001 - +--- liba52-0.7.4/mm_accel.h 2006-06-12 15:05:00.000000000 +0200 ++++ liba52/mm_accel.h 2006-06-05 02:23:04.000000000 +0200 +@@ -30,7 +34,12 @@ /* x86 accelerations */ #define MM_ACCEL_X86_MMX 0x80000000 #define MM_ACCEL_X86_3DNOW 0x40000000 @@ -2886,20 +2393,18 @@ +#define MM_ACCEL_PPC_ALTIVEC 0x00010000 uint32_t mm_accel (void); -+ -+#endif /* MM_ACCEL_H */ ---- liba52/parse.c 2005-03-22 19:59:35.000000000 +0100 -+++ parse.c 2004-04-01 15:41:29.000000000 +0200 -@@ -21,21 +27,19 @@ +--- liba52-0.7.4/parse.c 2006-06-12 15:05:07.000000000 +0200 ++++ liba52/parse.c 2006-06-12 14:51:33.000000000 +0200 +@@ -24,6 +28,7 @@ #include "config.h" --#include <inttypes.h> #include <stdlib.h> ++#include <stdio.h> #include <string.h> -+#include <inttypes.h> + #include <inttypes.h> - #include "a52.h" +@@ -31,13 +36,11 @@ #include "a52_internal.h" #include "bitstream.h" #include "tables.h" @@ -2914,110 +2419,97 @@ #endif typedef struct { -@@ -54,12 +58,28 @@ - sample_t * samples; - int i; +@@ -61,6 +64,21 @@ + return NULL; -- imdct_init (mm_accel); -- - samples = memalign (16, 256 * 12 * sizeof (sample_t)); + state->samples = memalign (16, 256 * 12 * sizeof (sample_t)); +#if defined(__MINGW32__) && defined(HAVE_SSE) + for(i=0;i<10;i++){ -+ if((int)samples%16){ ++ if((int)state->samples%16){ + sample_t* samplestmp=malloc(256 * 12 * sizeof (sample_t)); -+ free(samples); -+ samples = samplestmp; ++ free(state->samples); ++ state->samples = samplestmp; + } + else break; + } +#endif -+ if(((int)samples%16) && (mm_accel&MM_ACCEL_X86_SSE)){ ++ if(((int)state->samples%16) && (mm_accel&MM_ACCEL_X86_SSE)){ + mm_accel &=~MM_ACCEL_X86_SSE; -+ printf("liba52: unable to get 16 byte aligned memory disabling usage of SSE instructions\n"); ++ fprintf(stderr, "liba52: unable to get 16 byte aligned memory disabling usage of SSE instructions\n"); + } + - if (samples == NULL) -- return NULL; -- -+ return NULL; -+ -+ imdct_init (mm_accel); + if (state->samples == NULL) { + free (state); + return NULL; +@@ -74,6 +92,7 @@ + state->lfsr_state = 1; + + a52_imdct_init (mm_accel); + downmix_accel_init(mm_accel); -+ - for (i = 0; i < 256 * 12; i++) - samples[i] = 0; -@@ -124,7 +144,7 @@ + return state; + } +@@ -141,7 +160,7 @@ state->acmod = acmod = buf[6] >> 5; - bitstream_set_ptr (buf + 6); -- bitstream_get (3); /* skip acmod we already parsed */ -+ bitstream_skip (3); /* skip acmod we already parsed */ + a52_bitstream_set_ptr (state, buf + 6); +- bitstream_get (state, 3); /* skip acmod we already parsed */ ++ bitstream_skip (state, 3); /* skip acmod we already parsed */ - if ((acmod == 2) && (bitstream_get (2) == 2)) /* dsurmod */ + if ((acmod == 2) && (bitstream_get (state, 2) == 2)) /* dsurmod */ acmod = A52_DOLBY; -@@ -144,7 +164,7 @@ - if (state->lfeon && (*flags & A52_LFE)) - state->output |= A52_LFE; - *flags = state->output; -- // the 2* compensates for differences in imdct -+ /* the 2* compensates for differences in imdct */ - state->dynrng = state->level = 2 * *level; - state->bias = bias; - state->dynrnge = 1; -@@ -152,28 +172,28 @@ +@@ -172,28 +191,28 @@ chaninfo = !acmod; do { -- bitstream_get (5); /* dialnorm */ -+ bitstream_skip (5); /* dialnorm */ - if (bitstream_get (1)) /* compre */ -- bitstream_get (8); /* compr */ -+ bitstream_skip (8); /* compr */ - if (bitstream_get (1)) /* langcode */ -- bitstream_get (8); /* langcod */ -+ bitstream_skip (8); /* langcod */ - if (bitstream_get (1)) /* audprodie */ -- bitstream_get (7); /* mixlevel + roomtyp */ -+ bitstream_skip (7); /* mixlevel + roomtyp */ +- bitstream_get (state, 5); /* dialnorm */ ++ bitstream_skip (state, 5); /* dialnorm */ + if (bitstream_get (state, 1)) /* compre */ +- bitstream_get (state, 8); /* compr */ ++ bitstream_skip (state, 8); /* compr */ + if (bitstream_get (state, 1)) /* langcode */ +- bitstream_get (state, 8); /* langcod */ ++ bitstream_skip (state, 8); /* langcod */ + if (bitstream_get (state, 1)) /* audprodie */ +- bitstream_get (state, 7); /* mixlevel + roomtyp */ ++ bitstream_skip (state, 7); /* mixlevel + roomtyp */ } while (chaninfo--); -- bitstream_get (2); /* copyrightb + origbs */ -+ bitstream_skip (2); /* copyrightb + origbs */ +- bitstream_get (state, 2); /* copyrightb + origbs */ ++ bitstream_skip (state, 2); /* copyrightb + origbs */ - if (bitstream_get (1)) /* timecod1e */ -- bitstream_get (14); /* timecod1 */ -+ bitstream_skip (14); /* timecod1 */ - if (bitstream_get (1)) /* timecod2e */ -- bitstream_get (14); /* timecod2 */ -+ bitstream_skip (14); /* timecod2 */ + if (bitstream_get (state, 1)) /* timecod1e */ +- bitstream_get (state, 14); /* timecod1 */ ++ bitstream_skip (state, 14); /* timecod1 */ + if (bitstream_get (state, 1)) /* timecod2e */ +- bitstream_get (state, 14); /* timecod2 */ ++ bitstream_skip (state, 14); /* timecod2 */ - if (bitstream_get (1)) { /* addbsie */ + if (bitstream_get (state, 1)) { /* addbsie */ int addbsil; - addbsil = bitstream_get (6); + addbsil = bitstream_get (state, 6); do { -- bitstream_get (8); /* addbsi */ -+ bitstream_skip (8); /* addbsi */ +- bitstream_get (state, 8); /* addbsi */ ++ bitstream_skip (state, 8); /* addbsi */ } while (addbsil--); } -@@ -647,7 +667,7 @@ - if (parse_exponents (chexpstr[i], nchgrps, state->fbw_exp[i][0], - state->fbw_exp[i] + 1)) +@@ -680,7 +699,7 @@ + state->fbw_expbap[i].exp[0], + state->fbw_expbap[i].exp + 1)) return 1; -- bitstream_get (2); /* gainrng */ -+ bitstream_skip (2); /* gainrng */ +- bitstream_get (state, 2); /* gainrng */ ++ bitstream_skip (state, 2); /* gainrng */ } if (lfeexpstr != EXP_REUSE) { do_bit_alloc |= 32; -@@ -729,7 +749,7 @@ - if (bitstream_get (1)) { /* skiple */ - i = bitstream_get (9); /* skipl */ +@@ -755,7 +774,7 @@ + if (bitstream_get (state, 1)) { /* skiple */ + i = bitstream_get (state, 9); /* skipl */ while (i--) -- bitstream_get (8); -+ bitstream_skip (8); +- bitstream_get (state, 8); ++ bitstream_skip (state, 8); } - if (state->output & A52_LFE) - + samples = state->samples;