Mercurial > mplayer.hg

--- a/liba52/Makefile	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/Makefile	Thu Jun 15 22:58:06 2006 +0000
@@ -9,12 +9,11 @@
           bitstream.c \
           downmix.c \
           imdct.c \
-          imdct_mlib.c \
           parse.c \

 OBJS	= $(SRCS:.c=.o)

-CFLAGS  = $(MLIB_INC) $(OPTFLAGS) -I..
+CFLAGS  = $(OPTFLAGS) -I..

 .SUFFIXES: .c .o
--- a/liba52/a52.h	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/a52.h	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * a52.h
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -34,69 +34,7 @@
 typedef double sample_t;
 #endif

-typedef struct a52_ba_s {
-    uint16_t fsnroffst;		/* fine SNR offset */
-    uint16_t fgaincod;		/* fast gain */
-    uint16_t deltbae;		/* delta bit allocation exists */
-    int8_t deltba[50];		/* per-band delta bit allocation */
-} a52_ba_t;
-
-typedef struct a52_state_s {
-    uint8_t fscod;		/* sample rate */
-    uint8_t halfrate;		/* halfrate factor */
-    uint8_t acmod;		/* coded channels */
-    sample_t clev;		/* centre channel mix level */
-    sample_t slev;		/* surround channels mix level */
-    uint8_t lfeon;		/* coded lfe channel */
-
-    int output;			/* type of output */
-    sample_t level;		/* output level */
-    sample_t bias;		/* output bias */
-
-    int dynrnge;		/* apply dynamic range */
-    sample_t dynrng;		/* dynamic range */
-    void * dynrngdata;		/* dynamic range callback funtion and data */
-    sample_t (* dynrngcall) (sample_t range, void * dynrngdata);
-
-    uint16_t cplinu;		/* coupling in use */
-    uint16_t chincpl[5];	/* channel coupled */
-    uint16_t phsflginu;		/* phase flags in use (stereo only) */
-    uint16_t cplbndstrc[18];	/* coupling band structure */
-    uint16_t cplstrtmant;	/* coupling channel start mantissa */
-    uint16_t cplendmant;	/* coupling channel end mantissa */
-    sample_t cplco[5][18];	/* coupling coordinates */
-
-    /* derived information */
-    uint16_t cplstrtbnd;	/* coupling start band (for bit allocation) */
-    uint16_t ncplbnd;		/* number of coupling bands */
-
-    uint16_t rematflg[4];	/* stereo rematrixing */
-
-    uint16_t endmant[5];	/* channel end mantissa */
-
-    uint8_t cpl_exp[256];	/* decoded coupling channel exponents */
-    uint8_t fbw_exp[5][256];	/* decoded channel exponents */
-    uint8_t lfe_exp[7];		/* decoded lfe channel exponents */
-
-    uint16_t sdcycod;		/* slow decay */
-    uint16_t fdcycod;		/* fast decay */
-    uint16_t sgaincod;		/* slow gain */
-    uint16_t dbpbcod;		/* dB per bit - encodes the dbknee value */
-    uint16_t floorcod;		/* masking floor */
-
-    uint16_t csnroffst;		/* coarse SNR offset */
-    a52_ba_t cplba;		/* coupling bit allocation parameters */
-    a52_ba_t ba[5];		/* channel bit allocation parameters */
-    a52_ba_t lfeba;		/* lfe bit allocation parameters */
-
-    uint16_t cplfleak;		/* coupling fast leak init */
-    uint16_t cplsleak;		/* coupling slow leak init */
-
-    /* derived bit allocation information */
-    int8_t fbw_bap[5][256];
-    int8_t cpl_bap[256];
-    int8_t lfe_bap[7];
-} a52_state_t;
+typedef struct a52_state_s a52_state_t;

 #define A52_CHANNEL 0
 #define A52_MONO 1
@@ -114,14 +52,16 @@
 #define A52_LFE 16
 #define A52_ADJUST_LEVEL 32

-sample_t * a52_init (uint32_t mm_accel);
+a52_state_t * a52_init (uint32_t mm_accel);
+sample_t * a52_samples (a52_state_t * state);
 int a52_syncinfo (uint8_t * buf, int * flags,
 		  int * sample_rate, int * bit_rate);
 int a52_frame (a52_state_t * state, uint8_t * buf, int * flags,
 	       sample_t * level, sample_t bias);
 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);
+int a52_block (a52_state_t * state);
+void a52_free (a52_state_t * state);

 void* a52_resample_init(uint32_t mm_accel,int flags,int chans);
 extern int (* a52_resample) (float * _f, int16_t * s16);
--- a/liba52/a52_internal.h	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/a52_internal.h	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * a52_internal.h
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -25,6 +25,72 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */

+typedef struct {
+    uint8_t bai;		/* fine SNR offset, fast gain */
+    uint8_t deltbae;		/* delta bit allocation exists */
+    int8_t deltba[50];		/* per-band delta bit allocation */
+} ba_t;
+
+typedef struct {
+    uint8_t exp[256];		/* decoded channel exponents */
+    int8_t bap[256];		/* derived channel bit allocation */
+} expbap_t;
+
+struct a52_state_s {
+    uint8_t fscod;		/* sample rate */
+    uint8_t halfrate;		/* halfrate factor */
+    uint8_t acmod;		/* coded channels */
+    uint8_t lfeon;		/* coded lfe channel */
+    sample_t clev;		/* centre channel mix level */
+    sample_t slev;		/* surround channels mix level */
+
+    int output;			/* type of output */
+    sample_t level;		/* output level */
+    sample_t bias;		/* output bias */
+
+    int dynrnge;		/* apply dynamic range */
+    sample_t dynrng;		/* dynamic range */
+    void * dynrngdata;		/* dynamic range callback funtion and data */
+    sample_t (* dynrngcall) (sample_t range, void * dynrngdata);
+
+    uint8_t chincpl;		/* channel coupled */
+    uint8_t phsflginu;		/* phase flags in use (stereo only) */
+    uint8_t cplstrtmant;	/* coupling channel start mantissa */
+    uint8_t cplendmant;		/* coupling channel end mantissa */
+    uint32_t cplbndstrc;	/* coupling band structure */
+    sample_t cplco[5][18];	/* coupling coordinates */
+
+    /* derived information */
+    uint8_t cplstrtbnd;		/* coupling start band (for bit allocation) */
+    uint8_t ncplbnd;		/* number of coupling bands */
+
+    uint8_t rematflg;		/* stereo rematrixing */
+
+    uint8_t endmant[5];		/* channel end mantissa */
+
+    uint16_t bai;		/* bit allocation information */
+
+    uint32_t * buffer_start;
+    uint16_t lfsr_state;	/* dither state */
+    uint32_t bits_left;
+    uint32_t current_word;
+
+    uint8_t csnroffst;		/* coarse SNR offset */
+    ba_t cplba;			/* coupling bit allocation parameters */
+    ba_t ba[5];			/* channel bit allocation parameters */
+    ba_t lfeba;			/* lfe bit allocation parameters */
+
+    uint8_t cplfleak;		/* coupling fast leak init */
+    uint8_t cplsleak;		/* coupling slow leak init */
+
+    expbap_t cpl_expbap;
+    expbap_t fbw_expbap[5];
+    expbap_t lfe_expbap;
+
+    sample_t * samples;
+    int downmixed;
+};
+
 #define LEVEL_PLUS6DB 2.0
 #define LEVEL_PLUS3DB 1.4142135623730951
 #define LEVEL_3DB 0.7071067811865476
@@ -55,21 +121,20 @@
 # define REG_BP "ebp"
 #endif

-void bit_allocate (a52_state_t * state, a52_ba_t * ba, int bndstart,
+void a52_bit_allocate (a52_state_t * state, ba_t * ba, int bndstart,
 		   int start, int end, int fastleak, int slowleak,
-		   uint8_t * exp, int8_t * bap);
+		       expbap_t * expbap);

-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);
 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);
-extern void (*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);
-extern void (*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);
-extern void (* imdct_512) (sample_t * data, sample_t * delay, sample_t bias);
-void imdct_do_256_mlib (sample_t * data, sample_t * delay, sample_t bias);
-void imdct_do_512_mlib (sample_t * data, sample_t * delay, sample_t bias);
+void a52_imdct_init (uint32_t mm_accel);
+void a52_imdct_256 (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);
--- a/liba52/bit_allocate.c	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/bit_allocate.c	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * bit_allocate.c
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -121,9 +121,9 @@
     mask -= floor;				\
 } while (0)

-void bit_allocate (a52_state_t * state, a52_ba_t * ba, int bndstart,
+void a52_bit_allocate (a52_state_t * state, ba_t * ba, int bndstart,
 		   int start, int end, int fastleak, int slowleak,
-		   uint8_t * exp, int8_t * bap)
+		       expbap_t * expbap)
 {
     static int slowgain[4] = {0x540, 0x4d8, 0x478, 0x410};
     static int dbpbtab[4]  = {0xc00, 0x500, 0x300, 0x100};
@@ -131,6 +131,8 @@
 			      0xa10, 0xa90, 0xb10, 0x1400};

     int i, j;
+    uint8_t * exp;
+    int8_t * bap;
     int fdecay, fgain, sdecay, sgain, dbknee, floor, snroffset;
     int psd, mask;
     int8_t * deltba;
@@ -138,21 +140,24 @@
     int halfrate;

     halfrate = state->halfrate;
-    fdecay = (63 + 20 * state->fdcycod) >> halfrate;
-    fgain = 128 + 128 * ba->fgaincod;
-    sdecay = (15 + 2 * state->sdcycod) >> halfrate;
-    sgain = slowgain[state->sgaincod];
-    dbknee = dbpbtab[state->dbpbcod];
+    fdecay = (63 + 20 * ((state->bai >> 7) & 3)) >> halfrate;	/* fdcycod */
+    fgain = 128 + 128 * (ba->bai & 7);				/* fgaincod */
+    sdecay = (15 + 2 * (state->bai >> 9)) >> halfrate;		/* sdcycod */
+    sgain = slowgain[(state->bai >> 5) & 3];			/* sgaincod */
+    dbknee = dbpbtab[(state->bai >> 3) & 3];			/* dbpbcod */
     hth = hthtab[state->fscod];
     /*
      * if there is no delta bit allocation, make deltba point to an area
      * known to contain zeroes. baptab+156 here.
      */
     deltba = (ba->deltbae == DELTA_BIT_NONE) ? baptab + 156 : ba->deltba;
-    floor = floortab[state->floorcod];
-    snroffset = 960 - 64 * state->csnroffst - 4 * ba->fsnroffst + floor;
+    floor = floortab[state->bai & 7];				/* floorcod */
+    snroffset = 960 - 64 * state->csnroffst - 4 * (ba->bai >> 3) + floor;
     floor >>= 5;

+    exp = expbap->exp;
+    bap = expbap->bap;
+
     i = bndstart;
     j = start;
     if (start == 0) {	/* not the coupling channel */
--- a/liba52/bitstream.c	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/bitstream.c	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * bitstream.c
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -37,34 +37,27 @@

 #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;
-
-void bitstream_set_ptr (uint8_t * buf)
+void a52_bitstream_set_ptr (a52_state_t * state, uint8_t * buf)
 {
     int align;

-    align = (int)buf & 3;
-    buffer_start = (uint32_t *) (buf - align);
-    bits_left = 0;
+    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);
 }

-static inline void
-bitstream_fill_current(void)
+static inline void bitstream_fill_current (a52_state_t * state)
 {
     uint32_t tmp;

-    tmp = *(buffer_start++);
-    current_word = swab32 (tmp);
+    tmp = *(state->buffer_start++);
+    state->current_word = swab32 (tmp);
 }

 /*
@@ -76,38 +69,38 @@
  * -ah
  */

-uint32_t
-bitstream_get_bh(uint32_t num_bits)
+uint32_t a52_bitstream_get_bh (a52_state_t * state, uint32_t num_bits)
 {
     uint32_t result;

-    num_bits -= bits_left;
-    result = (current_word << (32 - bits_left)) >> (32 - bits_left);
+    num_bits -= state->bits_left;
+    result = ((state->current_word << (32 - state->bits_left)) >>
+	      (32 - state->bits_left));

-    bitstream_fill_current();
+    bitstream_fill_current (state);

     if(num_bits != 0)
-	result = (result << num_bits) | (current_word >> (32 - num_bits));
+	result = (result << num_bits) | (state->current_word >> (32 - num_bits));

-    bits_left = 32 - num_bits;
+    state->bits_left = 32 - num_bits;

     return result;
 }

-int32_t
-bitstream_get_bh_2(uint32_t num_bits)
+int32_t a52_bitstream_get_bh_2 (a52_state_t * state, uint32_t num_bits)
 {
     int32_t result;

-    num_bits -= bits_left;
-    result = (((int32_t)current_word) << (32 - bits_left)) >> (32 - bits_left);
+    num_bits -= state->bits_left;
+    result = ((((int32_t)state->current_word) << (32 - state->bits_left)) >>
+	      (32 - state->bits_left));

-    bitstream_fill_current();
+    bitstream_fill_current(state);

     if(num_bits != 0)
-	result = (result << num_bits) | (current_word >> (32 - num_bits));
+	result = (result << num_bits) | (state->current_word >> (32 - num_bits));

-    bits_left = 32 - num_bits;
+    state->bits_left = 32 - num_bits;

     return result;
 }
--- a/liba52/bitstream.h	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/bitstream.h	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * bitstream.h
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -77,7 +77,7 @@
 #	if defined (__i386__)

 #	define swab32(x) __i386_swab32(x)
-	static always_inline const uint32_t __i386_swab32(uint32_t x)
+	static inline const uint32_t __i386_swab32(uint32_t x)
 	{
 		__asm__("bswap %0" : "=r" (x) : "0" (x));
 		return x;
@@ -95,23 +95,17 @@
 #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 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);
+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);

-
-static inline uint32_t
-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;
@@ -121,30 +115,29 @@
 #else
     uint32_t result;

-    if(num_bits < bits_left) {
-	result = (current_word << (32 - bits_left)) >> (32 - num_bits);
-	bits_left -= num_bits;
+    if (num_bits < state->bits_left) {
+	result = (state->current_word << (32 - state->bits_left)) >> (32 - num_bits);
+	state->bits_left -= num_bits;
 	return result;
     }

-    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;
@@ -154,12 +147,12 @@
 #else
     int32_t result;

-    if(num_bits < bits_left) {
-	result = (((int32_t)current_word) << (32 - bits_left)) >> (32 - num_bits);
-	bits_left -= num_bits;
+    if (num_bits < state->bits_left) {
+	result = (((int32_t)state->current_word) << (32 - state->bits_left)) >> (32 - num_bits);
+	state->bits_left -= num_bits;
 	return result;
     }

-    return bitstream_get_bh_2(num_bits);
+    return a52_bitstream_get_bh_2 (state, num_bits);
 #endif
 }
--- a/liba52/downmix.c	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/downmix.c	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * downmix.c
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -40,9 +40,9 @@
 #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);
@@ -55,16 +55,16 @@

 void downmix_accel_init(uint32_t mm_accel)
 {
-    upmix= upmix_C;
-    downmix= downmix_C;
+    a52_upmix= upmix_C;
+    a52_downmix= downmix_C;
 #if defined(ARCH_X86) || defined(ARCH_X86_64)
-    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;
+    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)
 {
     static uint8_t table[11][8] = {
@@ -183,7 +183,7 @@
     return output;
 }

-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)
 {
     switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
@@ -478,7 +478,7 @@
 	samples[i] = 0;
 }

-static void downmix_C (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)) {
@@ -619,7 +619,7 @@
     }
 }

-static void upmix_C (sample_t * samples, int acmod, int output)
+void upmix_C (sample_t * samples, int acmod, int output)
 {
     switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
--- a/liba52/imdct.c	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/imdct.c	Thu Jun 15 22:58:06 2006 +0000
@@ -1,8 +1,11 @@
 /*
  * imdct.c
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
+ * The ifft algorithms in this file have been largely inspired by Dan
+ * Bernstein's work, djbfft, available at http://cr.yp.to/djbfft.html
+ *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
  * See http://liba52.sourceforge.net/ for updates.
  *
@@ -35,6 +38,9 @@

 #include <math.h>
 #include <stdio.h>
+#ifdef LIBA52_DJBFFT
+#include <fftc4.h>
+#endif
 #ifndef M_PI
 #define M_PI 3.1415926535897932384626433832795029
 #endif
@@ -45,22 +51,17 @@
 #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);
-
 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,
@@ -73,7 +74,6 @@
 	7, 23, 39, 55, 71, 87, 103, 119, 15, 31, 47,  63, 79, 95, 111, 127
 };

-/* 128 point bit-reverse LUT */
 static uint8_t attribute_used bit_reverse_512[] = {
 	0x00, 0x40, 0x20, 0x60, 0x10, 0x50, 0x30, 0x70,
 	0x08, 0x48, 0x28, 0x68, 0x18, 0x58, 0x38, 0x78,
@@ -92,36 +92,18 @@
 	0x07, 0x47, 0x27, 0x67, 0x17, 0x57, 0x37, 0x77,
 	0x0f, 0x4f, 0x2f, 0x6f, 0x1f, 0x5f, 0x3f, 0x7f};

-static uint8_t bit_reverse_256[] = {
-	0x00, 0x20, 0x10, 0x30, 0x08, 0x28, 0x18, 0x38,
-	0x04, 0x24, 0x14, 0x34, 0x0c, 0x2c, 0x1c, 0x3c,
-	0x02, 0x22, 0x12, 0x32, 0x0a, 0x2a, 0x1a, 0x3a,
-	0x06, 0x26, 0x16, 0x36, 0x0e, 0x2e, 0x1e, 0x3e,
-	0x01, 0x21, 0x11, 0x31, 0x09, 0x29, 0x19, 0x39,
-	0x05, 0x25, 0x15, 0x35, 0x0d, 0x2d, 0x1d, 0x3d,
-	0x03, 0x23, 0x13, 0x33, 0x0b, 0x2b, 0x1b, 0x3b,
-	0x07, 0x27, 0x17, 0x37, 0x0f, 0x2f, 0x1f, 0x3f};
+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,
+      4,132, 68,196, 36,164,228,100, 20,148, 84,212,244,116, 52,180,
+    252,124, 60,188, 28,156,220, 92, 12,140, 76,204,236,108, 44,172,
+      2,130, 66,194, 34,162,226, 98, 18,146, 82,210,242,114, 50,178,
+     10,138, 74,202, 42,170,234,106,250,122, 58,186, 26,154,218, 90,
+    254,126, 62,190, 30,158,222, 94, 14,142, 78,206,238,110, 46,174,
+      6,134, 70,198, 38,166,230,102,246,118, 54,182, 22,150,214, 86
+};

-#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};
-//static float __attribute__((aligned(16))) sseW0[4];
-static float __attribute__((aligned(16))) sseW1[8];
-static float __attribute__((aligned(16))) sseW2[16];
-static float __attribute__((aligned(16))) sseW3[32];
-static float __attribute__((aligned(16))) sseW4[64];
-static float __attribute__((aligned(16))) sseW5[128];
-static float __attribute__((aligned(16))) sseW6[256];
-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 __attribute__((aligned(16))) w_1[1];
@@ -136,257 +118,251 @@
 /* Twiddle factors for IMDCT */
 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];
+
+#if defined(ARCH_X86) || defined(ARCH_X86_64)
+// NOTE: SSE needs 16byte alignment or it will segfault
+//
+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};
+//static float __attribute__((aligned(16))) sseW0[4];
+static float __attribute__((aligned(16))) sseW1[8];
+static float __attribute__((aligned(16))) sseW2[16];
+static float __attribute__((aligned(16))) sseW3[32];
+static float __attribute__((aligned(16))) sseW4[64];
+static float __attribute__((aligned(16))) sseW5[128];
+static float __attribute__((aligned(16))) sseW6[256];
+static float __attribute__((aligned(16))) *sseW[7]=
+	{NULL /*sseW0*/,sseW1,sseW2,sseW3,sseW4,sseW5,sseW6};
+static float __attribute__((aligned(16))) sseWindow[512];
+#endif
+
+/* Root values for IFFT */
+static sample_t roots16[3];
+static sample_t roots32[7];
+static sample_t roots64[15];
+static sample_t roots128[31];
+
+/* Twiddle factors for IMDCT */
+static complex_t pre1[128];
+static complex_t post1[64];
+static complex_t pre2[64];
+static complex_t post2[32];
+
+static sample_t a52_imdct_window[256];

-/* Windowing function for Modified DCT - Thank you acroread */
-sample_t imdct_window[] = {
-	0.00014, 0.00024, 0.00037, 0.00051, 0.00067, 0.00086, 0.00107, 0.00130,
-	0.00157, 0.00187, 0.00220, 0.00256, 0.00297, 0.00341, 0.00390, 0.00443,
-	0.00501, 0.00564, 0.00632, 0.00706, 0.00785, 0.00871, 0.00962, 0.01061,
-	0.01166, 0.01279, 0.01399, 0.01526, 0.01662, 0.01806, 0.01959, 0.02121,
-	0.02292, 0.02472, 0.02662, 0.02863, 0.03073, 0.03294, 0.03527, 0.03770,
-	0.04025, 0.04292, 0.04571, 0.04862, 0.05165, 0.05481, 0.05810, 0.06153,
-	0.06508, 0.06878, 0.07261, 0.07658, 0.08069, 0.08495, 0.08935, 0.09389,
-	0.09859, 0.10343, 0.10842, 0.11356, 0.11885, 0.12429, 0.12988, 0.13563,
-	0.14152, 0.14757, 0.15376, 0.16011, 0.16661, 0.17325, 0.18005, 0.18699,
-	0.19407, 0.20130, 0.20867, 0.21618, 0.22382, 0.23161, 0.23952, 0.24757,
-	0.25574, 0.26404, 0.27246, 0.28100, 0.28965, 0.29841, 0.30729, 0.31626,
-	0.32533, 0.33450, 0.34376, 0.35311, 0.36253, 0.37204, 0.38161, 0.39126,
-	0.40096, 0.41072, 0.42054, 0.43040, 0.44030, 0.45023, 0.46020, 0.47019,
-	0.48020, 0.49022, 0.50025, 0.51028, 0.52031, 0.53033, 0.54033, 0.55031,
-	0.56026, 0.57019, 0.58007, 0.58991, 0.59970, 0.60944, 0.61912, 0.62873,
-	0.63827, 0.64774, 0.65713, 0.66643, 0.67564, 0.68476, 0.69377, 0.70269,
-	0.71150, 0.72019, 0.72877, 0.73723, 0.74557, 0.75378, 0.76186, 0.76981,
-	0.77762, 0.78530, 0.79283, 0.80022, 0.80747, 0.81457, 0.82151, 0.82831,
-	0.83496, 0.84145, 0.84779, 0.85398, 0.86001, 0.86588, 0.87160, 0.87716,
-	0.88257, 0.88782, 0.89291, 0.89785, 0.90264, 0.90728, 0.91176, 0.91610,
-	0.92028, 0.92432, 0.92822, 0.93197, 0.93558, 0.93906, 0.94240, 0.94560,
-	0.94867, 0.95162, 0.95444, 0.95713, 0.95971, 0.96217, 0.96451, 0.96674,
-	0.96887, 0.97089, 0.97281, 0.97463, 0.97635, 0.97799, 0.97953, 0.98099,
-	0.98236, 0.98366, 0.98488, 0.98602, 0.98710, 0.98811, 0.98905, 0.98994,
-	0.99076, 0.99153, 0.99225, 0.99291, 0.99353, 0.99411, 0.99464, 0.99513,
-	0.99558, 0.99600, 0.99639, 0.99674, 0.99706, 0.99736, 0.99763, 0.99788,
-	0.99811, 0.99831, 0.99850, 0.99867, 0.99882, 0.99895, 0.99908, 0.99919,
-	0.99929, 0.99938, 0.99946, 0.99953, 0.99959, 0.99965, 0.99969, 0.99974,
-	0.99978, 0.99981, 0.99984, 0.99986, 0.99988, 0.99990, 0.99992, 0.99993,
-	0.99994, 0.99995, 0.99996, 0.99997, 0.99998, 0.99998, 0.99998, 0.99999,
-	0.99999, 0.99999, 0.99999, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000,
-	1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000 };
+static void (* ifft128) (complex_t * buf);
+static void (* ifft64) (complex_t * buf);
+
+static inline void ifft2 (complex_t * buf)
+{
+    double r, i;
+
+    r = buf[0].real;
+    i = buf[0].imag;
+    buf[0].real += buf[1].real;
+    buf[0].imag += buf[1].imag;
+    buf[1].real = r - buf[1].real;
+    buf[1].imag = i - buf[1].imag;
+}

-
-static inline void swap_cmplx(complex_t *a, complex_t *b)
+static inline void ifft4 (complex_t * buf)
 {
-    complex_t tmp;
+    double tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;

-    tmp = *a;
-    *a = *b;
-    *b = tmp;
+    tmp1 = buf[0].real + buf[1].real;
+    tmp2 = buf[3].real + buf[2].real;
+    tmp3 = buf[0].imag + buf[1].imag;
+    tmp4 = buf[2].imag + buf[3].imag;
+    tmp5 = buf[0].real - buf[1].real;
+    tmp6 = buf[0].imag - buf[1].imag;
+    tmp7 = buf[2].imag - buf[3].imag;
+    tmp8 = buf[3].real - buf[2].real;
+
+    buf[0].real = tmp1 + tmp2;
+    buf[0].imag = tmp3 + tmp4;
+    buf[2].real = tmp1 - tmp2;
+    buf[2].imag = tmp3 - tmp4;
+    buf[1].real = tmp5 + tmp7;
+    buf[1].imag = tmp6 + tmp8;
+    buf[3].real = tmp5 - tmp7;
+    buf[3].imag = tmp6 - tmp8;
 }

+/* the basic split-radix ifft butterfly */

+#define BUTTERFLY(a0,a1,a2,a3,wr,wi) do {	\
+    tmp5 = a2.real * wr + a2.imag * wi;		\
+    tmp6 = a2.imag * wr - a2.real * wi;		\
+    tmp7 = a3.real * wr - a3.imag * wi;		\
+    tmp8 = a3.imag * wr + a3.real * wi;		\
+    tmp1 = tmp5 + tmp7;				\
+    tmp2 = tmp6 + tmp8;				\
+    tmp3 = tmp6 - tmp8;				\
+    tmp4 = tmp7 - tmp5;				\
+    a2.real = a0.real - tmp1;			\
+    a2.imag = a0.imag - tmp2;			\
+    a3.real = a1.real - tmp3;			\
+    a3.imag = a1.imag - tmp4;			\
+    a0.real += tmp1;				\
+    a0.imag += tmp2;				\
+    a1.real += tmp3;				\
+    a1.imag += tmp4;				\
+} while (0)
+
+/* split-radix ifft butterfly, specialized for wr=1 wi=0 */

-static inline complex_t cmplx_mult(complex_t a, complex_t b)
+#define BUTTERFLY_ZERO(a0,a1,a2,a3) do {	\
+    tmp1 = a2.real + a3.real;			\
+    tmp2 = a2.imag + a3.imag;			\
+    tmp3 = a2.imag - a3.imag;			\
+    tmp4 = a3.real - a2.real;			\
+    a2.real = a0.real - tmp1;			\
+    a2.imag = a0.imag - tmp2;			\
+    a3.real = a1.real - tmp3;			\
+    a3.imag = a1.imag - tmp4;			\
+    a0.real += tmp1;				\
+    a0.imag += tmp2;				\
+    a1.real += tmp3;				\
+    a1.imag += tmp4;				\
+} while (0)
+
+/* split-radix ifft butterfly, specialized for wr=wi */
+
+#define BUTTERFLY_HALF(a0,a1,a2,a3,w) do {	\
+    tmp5 = (a2.real + a2.imag) * w;		\
+    tmp6 = (a2.imag - a2.real) * w;		\
+    tmp7 = (a3.real - a3.imag) * w;		\
+    tmp8 = (a3.imag + a3.real) * w;		\
+    tmp1 = tmp5 + tmp7;				\
+    tmp2 = tmp6 + tmp8;				\
+    tmp3 = tmp6 - tmp8;				\
+    tmp4 = tmp7 - tmp5;				\
+    a2.real = a0.real - tmp1;			\
+    a2.imag = a0.imag - tmp2;			\
+    a3.real = a1.real - tmp3;			\
+    a3.imag = a1.imag - tmp4;			\
+    a0.real += tmp1;				\
+    a0.imag += tmp2;				\
+    a1.real += tmp3;				\
+    a1.imag += tmp4;				\
+} while (0)
+
+static inline void ifft8 (complex_t * buf)
 {
-    complex_t ret;
+    double tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;

-    ret.real = a.real * b.real - a.imag * b.imag;
-    ret.imag = a.real * b.imag + a.imag * b.real;
-
-    return ret;
+    ifft4 (buf);
+    ifft2 (buf + 4);
+    ifft2 (buf + 6);
+    BUTTERFLY_ZERO (buf[0], buf[2], buf[4], buf[6]);
+    BUTTERFLY_HALF (buf[1], buf[3], buf[5], buf[7], roots16[1]);
 }

-void
-imdct_do_512(sample_t data[],sample_t delay[], sample_t bias)
+static void ifft_pass (complex_t * buf, sample_t * weight, int n)
 {
+    complex_t * buf1;
+    complex_t * buf2;
+    complex_t * buf3;
+    double tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7, tmp8;
     int i;
-#ifndef USE_AC3_C
-	int k;
-    int p,q;
-    int m;
-    int two_m;
-    int two_m_plus_one;

-    sample_t tmp_b_i;
-    sample_t tmp_b_r;
-#endif
-    sample_t tmp_a_i;
-    sample_t tmp_a_r;
+    buf++;
+    buf1 = buf + n;
+    buf2 = buf + 2 * n;
+    buf3 = buf + 3 * n;
+
+    BUTTERFLY_ZERO (buf[-1], buf1[-1], buf2[-1], buf3[-1]);
+
+    i = n - 1;

-    sample_t *data_ptr;
-    sample_t *delay_ptr;
-    sample_t *window_ptr;
-
-    /* 512 IMDCT with source and dest data in 'data' */
-
-    /* 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]) ; */
-#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]));
+    do {
+	BUTTERFLY (buf[0], buf1[0], buf2[0], buf3[0], weight[n], weight[2*i]);
+	buf++;
+	buf1++;
+	buf2++;
+	buf3++;
+	weight++;
+    } while (--i);
+}
+
+static void ifft16 (complex_t * buf)
+{
+    ifft8 (buf);
+    ifft4 (buf + 8);
+    ifft4 (buf + 12);
+    ifft_pass (buf, roots16 - 4, 4);
     }

-    /* FFT Merge */
-/* unoptimized variant
-    for (m=1; m < 7; m++) {
-	if(m)
-	    two_m = (1 << m);
-	else
-	    two_m = 1;
-
-	two_m_plus_one = (1 << (m+1));
-
-	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;
-		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;
-	    }
-	}
-    }
-*/
-#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;
+static void ifft32 (complex_t * buf)
+{
+    ifft16 (buf);
+    ifft8 (buf + 16);
+    ifft8 (buf + 24);
+    ifft_pass (buf, roots32 - 8, 8);
     }

-    /* 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;
+static void ifft64_c (complex_t * buf)
+{
+    ifft32 (buf);
+    ifft16 (buf + 32);
+    ifft16 (buf + 48);
+    ifft_pass (buf, roots64 - 16, 16);
+    }
+
+static void ifft128_c (complex_t * buf)
+{
+    ifft32 (buf);
+    ifft16 (buf + 32);
+    ifft16 (buf + 48);
+    ifft_pass (buf, roots64 - 16, 16);
+
+    ifft32 (buf + 64);
+    ifft32 (buf + 96);
+    ifft_pass (buf, roots128 - 32, 32);
      }

-    /* 4-7. iterations */
-    for (m=3; m < 7; m++) {
-        two_m = (1 << m);
-
-	two_m_plus_one = two_m<<1;
+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;
+    const sample_t * window = a52_imdct_window;
+    complex_t buf[128];

-	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;
-	    }
-	}
+    for( i=0; i < 128; i++) {
+	k = fftorder[i];
+	t_r = pre1[i].real;
+	t_i = pre1[i].imag;
+
+	buf[i].real = t_i * data[255-k] + t_r * data[k];
+	buf[i].imag = t_r * data[255-k] - t_i * data[k];
     }
-#endif
-    /* Post IFFT complex multiply  plus IFFT complex conjugate*/
-    for( i=0; i < 128; i++) {
-	/* y[n] = z[n] * (xcos1[n] + j * xsin1[n]) ; */
-	tmp_a_r =        buf[i].real;
-	tmp_a_i = -1.0 * buf[i].imag;
-	buf[i].real =(tmp_a_r * xcos1[i])  -  (tmp_a_i  * xsin1[i]);
-	buf[i].imag =(tmp_a_r * xsin1[i])  +  (tmp_a_i  * xcos1[i]);
-    }
-
-    data_ptr = data;
-    delay_ptr = delay;
-    window_ptr = imdct_window;
+
+    ifft128 (buf);

+    /* Post IFFT complex multiply plus IFFT complex conjugate*/
     /* Window and convert to real valued signal */
     for(i=0; i< 64; i++) {
-	*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;
-    }
+	/* y[n] = z[n] * (xcos1[n] + j * xsin1[n]) ; */
+	t_r = post1[i].real;
+	t_i = post1[i].imag;

-    /* The trailing edge of the window goes into the delay line */
-    delay_ptr = delay;
+	a_r = t_r * buf[i].real     + t_i * buf[i].imag;
+	a_i = t_i * buf[i].real     - t_r * buf[i].imag;
+	b_r = t_i * buf[127-i].real + t_r * buf[127-i].imag;
+	b_i = t_r * buf[127-i].real - t_i * buf[127-i].imag;

-    for(i=0; i< 64; i++) {
-	*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;
+	w_1 = window[2*i];
+	w_2 = window[255-2*i];
+	data[2*i]     = delay[2*i] * w_2 - a_r * w_1 + bias;
+	data[255-2*i] = delay[2*i] * w_1 + a_r * w_2 + bias;
+	delay[2*i] = a_i;
+
+	w_1 = window[2*i+1];
+	w_2 = window[254-2*i];
+	data[2*i+1]   = delay[2*i+1] * w_2 + b_r * w_1 + bias;
+	data[254-2*i] = delay[2*i+1] * w_1 - b_r * w_2 + bias;
+	delay[2*i+1] = b_i;
     }
 }

@@ -717,7 +693,7 @@

   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++) {
@@ -995,7 +971,7 @@

     data_ptr = data;
     delay_ptr = delay;
-    window_ptr = imdct_window;
+    window_ptr = a52_imdct_window;

     /* Window and convert to real valued signal */
 	asm volatile(
@@ -1098,151 +1074,148 @@
 }
 #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)
 {
     int i,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;
-
-    sample_t *data_ptr;
-    sample_t *delay_ptr;
-    sample_t *window_ptr;
-
-    complex_t *buf_1, *buf_2;
-
-    buf_1 = &buf[0];
-    buf_2 = &buf[64];
+    sample_t t_r, t_i, a_r, a_i, b_r, b_i, c_r, c_i, d_r, d_i, w_1, w_2;
+    const sample_t * window = a52_imdct_window;
+    complex_t buf1[64], buf2[64];

     /* Pre IFFT complex multiply plus IFFT cmplx conjugate */
-    for(k=0; k<64; k++) {
-	/* X1[k] = X[2*k]  */
-	/* X2[k] = X[2*k+1]     */
-
-	p = 2 * (128-2*k-1);
-	q = 2 * (2 * k);
+    for (i = 0; i < 64; i++) {
+	k = fftorder[i];
+	t_r = pre2[i].real;
+	t_i = pre2[i].imag;

-	/* Z1[k] = (X1[128-2*k-1] + j * X1[2*k]) * (xcos2[k] + j * xsin2[k]); */
-	buf_1[k].real =         data[p] * xcos2[k] - data[q] * xsin2[k];
-	buf_1[k].imag = -1.0f * (data[q] * xcos2[k] + data[p] * xsin2[k]);
-	/* Z2[k] = (X2[128-2*k-1] + j * X2[2*k]) * (xcos2[k] + j * xsin2[k]); */
-	buf_2[k].real =          data[p + 1] * xcos2[k] - data[q + 1] * xsin2[k];
-	buf_2[k].imag = -1.0f * ( data[q + 1] * xcos2[k] + data[p + 1] * xsin2[k]);
+	buf1[i].real = t_i * data[254-k] + t_r * data[k];
+	buf1[i].imag = t_r * data[254-k] - t_i * data[k];
+
+	buf2[i].real = t_i * data[255-k] + t_r * data[k+1];
+	buf2[i].imag = t_r * data[255-k] - t_i * data[k+1];
     }

-    /* IFFT Bit reversed shuffling */
-    for(i=0; i<64; i++) {
-	k = bit_reverse_256[i];
-	if (k < i) {
-	    swap_cmplx(&buf_1[i],&buf_1[k]);
-	    swap_cmplx(&buf_2[i],&buf_2[k]);
+    ifft64 (buf1);
+    ifft64 (buf2);
+
+    /* Post IFFT complex multiply */
+    /* Window and convert to real valued signal */
+    for (i = 0; i < 32; i++) {
+	/* y1[n] = z1[n] * (xcos2[n] + j * xs in2[n]) ; */
+	t_r = post2[i].real;
+	t_i = post2[i].imag;
+
+	a_r = t_r * buf1[i].real    + t_i * buf1[i].imag;
+	a_i = t_i * buf1[i].real    - t_r * buf1[i].imag;
+	b_r = t_i * buf1[63-i].real + t_r * buf1[63-i].imag;
+	b_i = t_r * buf1[63-i].real - t_i * buf1[63-i].imag;
+
+	c_r = t_r * buf2[i].real    + t_i * buf2[i].imag;
+	c_i = t_i * buf2[i].real    - t_r * buf2[i].imag;
+	d_r = t_i * buf2[63-i].real + t_r * buf2[63-i].imag;
+	d_i = t_r * buf2[63-i].real - t_i * buf2[63-i].imag;
+
+	w_1 = window[2*i];
+	w_2 = window[255-2*i];
+	data[2*i]     = delay[2*i] * w_2 - a_r * w_1 + bias;
+	data[255-2*i] = delay[2*i] * w_1 + a_r * w_2 + bias;
+	delay[2*i] = c_i;
+
+	w_1 = window[128+2*i];
+	w_2 = window[127-2*i];
+	data[128+2*i] = delay[127-2*i] * w_2 + a_i * w_1 + bias;
+	data[127-2*i] = delay[127-2*i] * w_1 - a_i * w_2 + bias;
+	delay[127-2*i] = c_r;
+
+	w_1 = window[2*i+1];
+	w_2 = window[254-2*i];
+	data[2*i+1]   = delay[2*i+1] * w_2 - b_i * w_1 + bias;
+	data[254-2*i] = delay[2*i+1] * w_1 + b_i * w_2 + bias;
+	delay[2*i+1] = d_r;
+
+	w_1 = window[129+2*i];
+	w_2 = window[126-2*i];
+	data[129+2*i] = delay[126-2*i] * w_2 + b_r * w_1 + bias;
+	data[126-2*i] = delay[126-2*i] * w_1 - b_r * w_2 + bias;
+	delay[126-2*i] = d_i;
 	}
     }

-    /* FFT Merge */
-    for (m=0; m < 6; m++) {
-	two_m = (1 << m);
-	two_m_plus_one = (1 << (m+1));
-
-	/* FIXME */
-	if(m)
-	    two_m = (1 << m);
-	else
-	    two_m = 1;
-
-	for(k = 0; k < two_m; k++) {
-	    for(i = 0; i < 64; i += two_m_plus_one) {
-		p = k + i;
-		q = p + two_m;
-		/* Do block 1 */
-		tmp_a_r = buf_1[p].real;
-		tmp_a_i = buf_1[p].imag;
-		tmp_b_r = buf_1[q].real * w[m][k].real - buf_1[q].imag * w[m][k].imag;
-		tmp_b_i = buf_1[q].imag * w[m][k].real + buf_1[q].real * w[m][k].imag;
-		buf_1[p].real = tmp_a_r + tmp_b_r;
-		buf_1[p].imag =  tmp_a_i + tmp_b_i;
-		buf_1[q].real = tmp_a_r - tmp_b_r;
-		buf_1[q].imag =  tmp_a_i - tmp_b_i;
+static double besselI0 (double x)
+{
+    double bessel = 1;
+    int i = 100;

-		/* Do block 2 */
-		tmp_a_r = buf_2[p].real;
-		tmp_a_i = buf_2[p].imag;
-		tmp_b_r = buf_2[q].real * w[m][k].real - buf_2[q].imag * w[m][k].imag;
-		tmp_b_i = buf_2[q].imag * w[m][k].real + buf_2[q].real * w[m][k].imag;
-		buf_2[p].real = tmp_a_r + tmp_b_r;
-		buf_2[p].imag =  tmp_a_i + tmp_b_i;
-		buf_2[q].real = tmp_a_r - tmp_b_r;
-		buf_2[q].imag =  tmp_a_i - tmp_b_i;
-	    }
-	}
-    }
-
-    /* Post IFFT complex multiply */
-    for( i=0; i < 64; i++) {
-	/* y1[n] = z1[n] * (xcos2[n] + j * xs in2[n]) ; */
-	tmp_a_r =  buf_1[i].real;
-	tmp_a_i = -buf_1[i].imag;
-	buf_1[i].real =(tmp_a_r * xcos2[i])  -  (tmp_a_i  * xsin2[i]);
-	buf_1[i].imag =(tmp_a_r * xsin2[i])  +  (tmp_a_i  * xcos2[i]);
-	/* y2[n] = z2[n] * (xcos2[n] + j * xsin2[n]) ; */
-	tmp_a_r =  buf_2[i].real;
-	tmp_a_i = -buf_2[i].imag;
-	buf_2[i].real =(tmp_a_r * xcos2[i])  -  (tmp_a_i  * xsin2[i]);
-	buf_2[i].imag =(tmp_a_r * xsin2[i])  +  (tmp_a_i  * xcos2[i]);
+    do
+	bessel = bessel * x / (i * i) + 1;
+    while (--i);
+    return bessel;
     }

-    data_ptr = data;
-    delay_ptr = delay;
-    window_ptr = imdct_window;
+void a52_imdct_init (uint32_t mm_accel)
+{
+    int i, j, k;
+    double sum;
+
+    /* compute imdct window - kaiser-bessel derived window, alpha = 5.0 */
+    sum = 0;
+    for (i = 0; i < 256; i++) {
+	sum += besselI0 (i * (256 - i) * (5 * M_PI / 256) * (5 * M_PI / 256));
+	a52_imdct_window[i] = sum;
+    }
+    sum++;
+    for (i = 0; i < 256; i++)
+	a52_imdct_window[i] = sqrt (a52_imdct_window[i] / sum);
+
+    for (i = 0; i < 3; i++)
+	roots16[i] = cos ((M_PI / 8) * (i + 1));

-    /* Window and convert to real valued signal */
-    for(i=0; i< 64; i++) {
-	*data_ptr++  = -buf_1[i].imag      * *window_ptr++ + *delay_ptr++ + bias;
-	*data_ptr++  =  buf_1[64-i-1].real * *window_ptr++ + *delay_ptr++ + bias;
+    for (i = 0; i < 7; i++)
+	roots32[i] = cos ((M_PI / 16) * (i + 1));
+
+    for (i = 0; i < 15; i++)
+	roots64[i] = cos ((M_PI / 32) * (i + 1));
+
+    for (i = 0; i < 31; i++)
+	roots128[i] = cos ((M_PI / 64) * (i + 1));
+
+    for(i=0; i< 64; i++) {
+	k = fftorder[i] / 2 + 64;
+	pre1[i].real = cos ((M_PI / 256) * (k - 0.25));
+	pre1[i].imag = sin ((M_PI / 256) * (k - 0.25));
+    }
+
+    for (i = 64; i < 128; i++) {
+	k = fftorder[i] / 2 + 64;
+	pre1[i].real = -cos ((M_PI / 256) * (k - 0.25));
+	pre1[i].imag = -sin ((M_PI / 256) * (k - 0.25));
     }

     for(i=0; i< 64; i++) {
-	*data_ptr++  = -buf_1[i].real      * *window_ptr++ + *delay_ptr++ + bias;
-	*data_ptr++  =  buf_1[64-i-1].imag * *window_ptr++ + *delay_ptr++ + bias;
-    }
-
-    delay_ptr = delay;
-
-    for(i=0; i< 64; i++) {
-	*delay_ptr++ = -buf_2[i].real      * *--window_ptr;
-	*delay_ptr++ =  buf_2[64-i-1].imag * *--window_ptr;
+	post1[i].real = cos ((M_PI / 256) * (i + 0.5));
+	post1[i].imag = sin ((M_PI / 256) * (i + 0.5));
     }

     for(i=0; i< 64; i++) {
-	*delay_ptr++ =  buf_2[i].imag      * *--window_ptr;
-	*delay_ptr++ = -buf_2[64-i-1].real * *--window_ptr;
-    }
+	k = fftorder[i] / 4;
+	pre2[i].real = cos ((M_PI / 128) * (k - 0.25));
+	pre2[i].imag = sin ((M_PI / 128) * (k - 0.25));
 }

-void imdct_init (uint32_t mm_accel)
-{
-#ifdef LIBA52_MLIB
-    if (mm_accel & MM_ACCEL_MLIB) {
-        fprintf (stderr, "Using mlib for IMDCT transform\n");
-	imdct_512 = imdct_do_512_mlib;
-	imdct_256 = imdct_do_256_mlib;
-    } else
-#endif
-    {
-	int i, j, k;
-
-	/* Twiddle factors to turn IFFT into IMDCT */
+    for (i = 0; i < 32; i++) {
+	post2[i].real = cos ((M_PI / 128) * (i + 0.5));
+	post2[i].imag = sin ((M_PI / 128) * (i + 0.5));
+    }
 	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];
@@ -1250,22 +1223,6 @@
 	    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++) {
-	    xcos2[i] = -cos ((M_PI / 1024) * (8 * i + 1));
-	    xsin2[i] = -sin ((M_PI / 1024) * (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 = 1; i < 7; i++) {
 	    j = 1 << i;
 	    for (k = 0; k < j; k+=2) {
@@ -1297,37 +1254,39 @@

 	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 || ARCH_X86_64
+#endif
+	a52_imdct_512 = imdct_do_512;
+	ifft128 = ifft128_c;
+	ifft64 = ifft64_c;

-	imdct_512 = imdct_do_512;
 #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 || ARCH_X86_64
@@ -1335,264 +1294,19 @@
         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;
-    }
-}

-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)
+#ifdef LIBA52_DJBFFT
+    if (mm_accel & MM_ACCEL_DJBFFT) {
+	fprintf (stderr, "Using djbfft for IMDCT transform\n");
+	ifft128 = (void (*) (complex_t *)) fftc4_un128;
+	ifft64 = (void (*) (complex_t *)) fftc4_un64;
+    } else
+#endif
 {
-  /* 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;
+	fprintf (stderr, "No accelerated IMDCT transform found\n");
 }
-
-
-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]);
 }
-
-
-
--- a/liba52/imdct_3dnow.h	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/imdct_3dnow.h	Thu Jun 15 22:58:06 2006 +0000
@@ -469,7 +469,7 @@

     data_ptr = data;
     delay_ptr = delay;
-    window_ptr = imdct_window;
+    window_ptr = a52_imdct_window;

     /* Window and convert to real valued signal */
 #if 1
--- 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;
--- a/liba52/mm_accel.h	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/mm_accel.h	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * mm_accel.h
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -29,7 +29,7 @@
 #define MM_ACCEL_H

 /* generic accelerations */
-#define MM_ACCEL_MLIB		0x00000001
+#define MM_ACCEL_DJBFFT		0x00000001

 /* x86 accelerations */
 #define MM_ACCEL_X86_MMX	0x80000000
--- a/liba52/parse.c	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/parse.c	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * parse.c
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -28,6 +28,7 @@
 #include "config.h"

 #include <stdlib.h>
+#include <stdio.h>
 #include <string.h>
 #include <inttypes.h>

@@ -53,37 +54,52 @@

 static uint8_t halfrate[12] = {0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 2, 3};

-sample_t * a52_init (uint32_t mm_accel)
+a52_state_t * a52_init (uint32_t mm_accel)
 {
-    sample_t * samples;
+    a52_state_t * state;
     int i;

-    samples = memalign (16, 256 * 12 * sizeof (sample_t));
+    state = malloc (sizeof (a52_state_t));
+    if (state == NULL)
+	return NULL;
+
+    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)
+    if (state->samples == NULL) {
+	free (state);
 	return NULL;
+    }
+
+    for (i = 0; i < 256 * 12; i++)
+	state->samples[i] = 0;
+
+    state->downmixed = 1;
+
+    state->lfsr_state = 1;

-    imdct_init (mm_accel);
+    a52_imdct_init (mm_accel);
     downmix_accel_init(mm_accel);

-    for (i = 0; i < 256 * 12; i++)
-	samples[i] = 0;
+    return state;
+}

-    return samples;
+sample_t * a52_samples (a52_state_t * state)
+{
+    return state->samples;
 }

 int a52_syncinfo (uint8_t * buf, int * flags,
@@ -117,7 +133,7 @@
     *bit_rate = (bitrate * 1000) >> half;

     switch (buf[4] & 0xc0) {
-    case 0:	/* 48 KHz */
+    case 0:
 	*sample_rate = 48000 >> half;
 	return 4 * bitrate;
     case 0x40:
@@ -143,21 +159,21 @@
     state->halfrate = halfrate[buf[5] >> 3];
     state->acmod = acmod = buf[6] >> 5;

-    bitstream_set_ptr (buf + 6);
-    bitstream_skip (3);	/* skip acmod we already parsed */
+    a52_bitstream_set_ptr (state, buf + 6);
+    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;

     if ((acmod & 1) && (acmod != 1))
-	state->clev = clev[bitstream_get (2)];	/* cmixlev */
+	state->clev = clev[bitstream_get (state, 2)];	/* cmixlev */

     if (acmod & 4)
-	state->slev = slev[bitstream_get (2)];	/* surmixlev */
+	state->slev = slev[bitstream_get (state, 2)];	/* surmixlev */

-    state->lfeon = bitstream_get (1);
+    state->lfeon = bitstream_get (state, 1);

-    state->output = downmix_init (acmod, *flags, level,
+    state->output = a52_downmix_init (acmod, *flags, level,
 				  state->clev, state->slev);
     if (state->output < 0)
 	return 1;
@@ -169,31 +185,34 @@
     state->bias = bias;
     state->dynrnge = 1;
     state->dynrngcall = NULL;
+    state->cplba.deltbae = DELTA_BIT_NONE;
+    state->ba[0].deltbae = state->ba[1].deltbae = state->ba[2].deltbae =
+	state->ba[3].deltbae = state->ba[4].deltbae = DELTA_BIT_NONE;

     chaninfo = !acmod;
     do {
-	bitstream_skip (5);	/* dialnorm */
-	if (bitstream_get (1))	/* compre */
-	    bitstream_skip (8);	/* compr */
-	if (bitstream_get (1))	/* langcode */
-	    bitstream_skip (8);	/* langcod */
-	if (bitstream_get (1))	/* audprodie */
-	    bitstream_skip (7);	/* mixlevel + roomtyp */
+	bitstream_skip (state, 5);	/* dialnorm */
+	if (bitstream_get (state, 1))	/* compre */
+	    bitstream_skip (state, 8);	/* compr */
+	if (bitstream_get (state, 1))	/* langcode */
+	    bitstream_skip (state, 8);	/* langcod */
+	if (bitstream_get (state, 1))	/* audprodie */
+	    bitstream_skip (state, 7);	/* mixlevel + roomtyp */
     } while (chaninfo--);

-    bitstream_skip (2);		/* copyrightb + origbs */
+    bitstream_skip (state, 2);		/* copyrightb + origbs */

-    if (bitstream_get (1))	/* timecod1e */
-	bitstream_skip (14);	/* timecod1 */
-    if (bitstream_get (1))	/* timecod2e */
-	bitstream_skip (14);	/* timecod2 */
+    if (bitstream_get (state, 1))	/* timecod1e */
+	bitstream_skip (state, 14);	/* timecod1 */
+    if (bitstream_get (state, 1))	/* timecod2e */
+	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_skip (8);	/* addbsi */
+	    bitstream_skip (state, 8);	/* addbsi */
 	} while (addbsil--);
     }

@@ -211,13 +230,13 @@
     }
 }

-static int parse_exponents (int expstr, int ngrps, uint8_t exponent,
-			    uint8_t * dest)
+static int parse_exponents (a52_state_t * state, int expstr, int ngrps,
+			    uint8_t exponent, uint8_t * dest)
 {
     int exps;

     while (ngrps--) {
-	exps = bitstream_get (7);
+	exps = bitstream_get (state, 7);

 	exponent += exp_1[exps];
 	if (exponent > 24)
@@ -265,18 +284,18 @@
     return 0;
 }

-static int parse_deltba (int8_t * deltba)
+static int parse_deltba (a52_state_t * state, int8_t * deltba)
 {
     int deltnseg, deltlen, delta, j;

     memset (deltba, 0, 50);

-    deltnseg = bitstream_get (3);
+    deltnseg = bitstream_get (state, 3);
     j = 0;
     do {
-	j += bitstream_get (5);
-	deltlen = bitstream_get (4);
-	delta = bitstream_get (3);
+	j += bitstream_get (state, 5);
+	deltlen = bitstream_get (state, 4);
+	delta = bitstream_get (state, 3);
 	delta -= (delta >= 4) ? 3 : 4;
 	if (!deltlen)
 	    continue;
@@ -293,37 +312,42 @@
 {
     int i;

-    if ((state->csnroffst) || (state->cplinu && state->cplba.fsnroffst) ||
-	(state->lfeon && state->lfeba.fsnroffst))
+    if ((state->csnroffst) ||
+	(state->chincpl && state->cplba.bai >> 3) ||	/* cplinu, fsnroffst */
+	(state->lfeon && state->lfeba.bai >> 3))	/* fsnroffst */
 	return 0;
     for (i = 0; i < nfchans; i++)
-	if (state->ba[i].fsnroffst)
+	if (state->ba[i].bai >> 3)			/* fsnroffst */
 	    return 0;
     return 1;
 }

-static inline int16_t dither_gen (void)
+static inline int16_t dither_gen (a52_state_t * state)
 {
-    static uint16_t lfsr_state = 1;
-    int16_t state;
+    int16_t nstate;

-    state = dither_lut[lfsr_state >> 8] ^ (lfsr_state << 8);
+    nstate = dither_lut[state->lfsr_state >> 8] ^ (state->lfsr_state << 8);

-    lfsr_state = (uint16_t) state;
+    state->lfsr_state = (uint16_t) nstate;

-    return state;
+    return nstate;
 }

-static void coeff_get (sample_t * coeff, uint8_t * exp, int8_t * bap,
-		       quantizer_t * quantizer, sample_t level,
-		       int dither, int end)
+static void coeff_get (a52_state_t * state, sample_t * coeff,
+		       expbap_t * expbap, quantizer_t * quantizer,
+		       sample_t level, int dither, int end)
 {
     int i;
+    uint8_t * exp;
+    int8_t * bap;
     sample_t factor[25];

     for (i = 0; i <= 24; i++)
 	factor[i] = scale_factor[i] * level;

+    exp = expbap->exp;
+    bap = expbap->bap;
+
     for (i = 0; i < end; i++) {
 	int bapi;

@@ -331,7 +355,7 @@
 	switch (bapi) {
 	case 0:
 	    if (dither) {
-		coeff[i] = dither_gen() * LEVEL_3DB * factor[exp[i]];
+		coeff[i] = dither_gen (state) * LEVEL_3DB * factor[exp[i]];
 		continue;
 	    } else {
 		coeff[i] = 0;
@@ -345,7 +369,7 @@
 	    } else {
 		int code;

-		code = bitstream_get (5);
+		code = bitstream_get (state, 5);

 		quantizer->q1_ptr = 1;
 		quantizer->q1[0] = q_1_2[code];
@@ -361,7 +385,7 @@
 	    } else {
 		int code;

-		code = bitstream_get (7);
+		code = bitstream_get (state, 7);

 		quantizer->q2_ptr = 1;
 		quantizer->q2[0] = q_2_2[code];
@@ -371,7 +395,7 @@
 	    }

 	case 3:
-	    coeff[i] = q_3[bitstream_get (3)] * factor[exp[i]];
+	    coeff[i] = q_3[bitstream_get (state, 3)] * factor[exp[i]];
 	    continue;

 	case -3:
@@ -382,7 +406,7 @@
 	    } else {
 		int code;

-		code = bitstream_get (7);
+		code = bitstream_get (state, 7);

 		quantizer->q4_ptr = 0;
 		quantizer->q4 = q_4_1[code];
@@ -391,11 +415,11 @@
 	    }

 	case 4:
-	    coeff[i] = q_5[bitstream_get (4)] * factor[exp[i]];
+	    coeff[i] = q_5[bitstream_get (state, 4)] * factor[exp[i]];
 	    continue;

 	default:
-	    coeff[i] = ((bitstream_get_2 (bapi) << (16 - bapi)) *
+	    coeff[i] = ((bitstream_get_2 (state, bapi) << (16 - bapi)) *
 			  factor[exp[i]]);
 	}
     }
@@ -405,19 +429,23 @@
 				sample_t * coeff, sample_t (* samples)[256],
 				quantizer_t * quantizer, uint8_t dithflag[5])
 {
-    int sub_bnd, bnd, i, i_end, ch;
+    int cplbndstrc, bnd, i, i_end, ch;
+    uint8_t * exp;
     int8_t * bap;
-    uint8_t * exp;
     sample_t cplco[5];

-    bap = state->cpl_bap;
-    exp = state->cpl_exp;
-    sub_bnd = bnd = 0;
+    exp = state->cpl_expbap.exp;
+    bap = state->cpl_expbap.bap;
+    bnd = 0;
+    cplbndstrc = state->cplbndstrc;
     i = state->cplstrtmant;
     while (i < state->cplendmant) {
 	i_end = i + 12;
-	while (state->cplbndstrc[sub_bnd++])
+	while (cplbndstrc & 1) {
+	    cplbndstrc >>= 1;
 	    i_end += 12;
+	}
+	cplbndstrc >>= 1;
 	for (ch = 0; ch < nfchans; ch++)
 	    cplco[ch] = state->cplco[ch][bnd] * coeff[ch];
 	bnd++;
@@ -431,10 +459,10 @@
 	    case 0:
 		cplcoeff = LEVEL_3DB * scale_factor[exp[i]];
 		for (ch = 0; ch < nfchans; ch++)
-		    if (state->chincpl[ch]) {
+		    if ((state->chincpl >> ch) & 1) {
 			if (dithflag[ch])
 			    samples[ch][i] = (cplcoeff * cplco[ch] *
-					      dither_gen ());
+					      dither_gen (state));
 			else
 			    samples[ch][i] = 0;
 		    }
@@ -448,7 +476,7 @@
 		} else {
 		    int code;

-		    code = bitstream_get (5);
+		    code = bitstream_get (state, 5);

 		    quantizer->q1_ptr = 1;
 		    quantizer->q1[0] = q_1_2[code];
@@ -464,7 +492,7 @@
 		} else {
 		    int code;

-		    code = bitstream_get (7);
+		    code = bitstream_get (state, 7);

 		    quantizer->q2_ptr = 1;
 		    quantizer->q2[0] = q_2_2[code];
@@ -474,7 +502,7 @@
 		}

 	    case 3:
-		cplcoeff = q_3[bitstream_get (3)];
+		cplcoeff = q_3[bitstream_get (state, 3)];
 		break;

 	    case -3:
@@ -485,7 +513,7 @@
 		} else {
 		    int code;

-		    code = bitstream_get (7);
+		    code = bitstream_get (state, 7);

 		    quantizer->q4_ptr = 0;
 		    quantizer->q4 = q_4_1[code];
@@ -494,23 +522,23 @@
 		}

 	    case 4:
-		cplcoeff = q_5[bitstream_get (4)];
+		cplcoeff = q_5[bitstream_get (state, 4)];
 		break;

 	    default:
-		cplcoeff = bitstream_get_2 (bapi) << (16 - bapi);
+		cplcoeff = bitstream_get_2 (state, bapi) << (16 - bapi);
 	    }

 	    cplcoeff *= scale_factor[exp[i]];
 	    for (ch = 0; ch < nfchans; ch++)
-		if (state->chincpl[ch])
+		if ((state->chincpl >> ch) & 1)
 		    samples[ch][i] = cplcoeff * cplco[ch];
 	    i++;
 	}
     }
 }

-int a52_block (a52_state_t * state, sample_t * samples)
+int a52_block (a52_state_t * state)
 {
     static const uint8_t nfchans_tbl[] = {2, 1, 2, 3, 3, 4, 4, 5, 1, 1, 2};
     static int rematrix_band[4] = {25, 37, 61, 253};
@@ -520,21 +548,22 @@
     sample_t coeff[5];
     int chanbias;
     quantizer_t quantizer;
+    sample_t * samples;

     nfchans = nfchans_tbl[state->acmod];

     for (i = 0; i < nfchans; i++)
-	blksw[i] = bitstream_get (1);
+	blksw[i] = bitstream_get (state, 1);

     for (i = 0; i < nfchans; i++)
-	dithflag[i] = bitstream_get (1);
+	dithflag[i] = bitstream_get (state, 1);

-    chaninfo = !(state->acmod);
+    chaninfo = !state->acmod;
     do {
-	if (bitstream_get (1)) {	/* dynrnge */
+	if (bitstream_get (state, 1)) {	/* dynrnge */
 	    int dynrng;

-	    dynrng = bitstream_get_2 (8);
+	    dynrng = bitstream_get_2 (state, 8);
 	    if (state->dynrnge) {
 		sample_t range;

@@ -547,25 +576,25 @@
 	}
     } while (chaninfo--);

-    if (bitstream_get (1)) {	/* cplstre */
-	state->cplinu = bitstream_get (1);
-	if (state->cplinu) {
-	    static int bndtab[16] = {31, 35, 37, 39, 41, 42, 43, 44,
+    if (bitstream_get (state, 1)) {	/* cplstre */
+	state->chincpl = 0;
+	if (bitstream_get (state, 1)) {	/* cplinu */
+	    static uint8_t bndtab[16] = {31, 35, 37, 39, 41, 42, 43, 44,
 				     45, 45, 46, 46, 47, 47, 48, 48};
 	    int cplbegf;
 	    int cplendf;
 	    int ncplsubnd;

 	    for (i = 0; i < nfchans; i++)
-		state->chincpl[i] = bitstream_get (1);
+		state->chincpl |= bitstream_get (state, 1) << i;
 	    switch (state->acmod) {
 	    case 0: case 1:
 		return 1;
 	    case 2:
-		state->phsflginu = bitstream_get (1);
+		state->phsflginu = bitstream_get (state, 1);
 	    }
-	    cplbegf = bitstream_get (4);
-	    cplendf = bitstream_get (4);
+	    cplbegf = bitstream_get (state, 4);
+	    cplendf = bitstream_get (state, 4);

 	    if (cplendf + 3 - cplbegf < 0)
 		return 1;
@@ -574,28 +603,29 @@
 	    state->cplstrtmant = cplbegf * 12 + 37;
 	    state->cplendmant = cplendf * 12 + 73;

-	    for (i = 0; i < ncplsubnd - 1; i++) {
-		state->cplbndstrc[i] = bitstream_get (1);
-		state->ncplbnd -= state->cplbndstrc[i];
+	    state->cplbndstrc = 0;
+	    for (i = 0; i < ncplsubnd - 1; i++)
+		if (bitstream_get (state, 1)) {
+		    state->cplbndstrc |= 1 << i;
+		    state->ncplbnd--;
 	    }
-	    state->cplbndstrc[i] = 0;	/* last value is a sentinel */
 	}
     }

-    if (state->cplinu) {
+    if (state->chincpl) {	/* cplinu */
 	int j, cplcoe;

 	cplcoe = 0;
 	for (i = 0; i < nfchans; i++)
-	    if (state->chincpl[i])
-		if (bitstream_get (1)) {	/* cplcoe */
+	    if ((state->chincpl) >> i & 1)
+		if (bitstream_get (state, 1)) {	/* cplcoe */
 		    int mstrcplco, cplcoexp, cplcomant;

 		    cplcoe = 1;
-		    mstrcplco = 3 * bitstream_get (2);
+		    mstrcplco = 3 * bitstream_get (state, 2);
 		    for (j = 0; j < state->ncplbnd; j++) {
-			cplcoexp = bitstream_get (4);
-			cplcomant = bitstream_get (4);
+			cplcoexp = bitstream_get (state, 4);
+			cplcomant = bitstream_get (state, 4);
 			if (cplcoexp == 15)
 			    cplcomant <<= 14;
 			else
@@ -606,37 +636,38 @@
 		}
 	if ((state->acmod == 2) && state->phsflginu && cplcoe)
 	    for (j = 0; j < state->ncplbnd; j++)
-		if (bitstream_get (1))	/* phsflg */
+		if (bitstream_get (state, 1))	/* phsflg */
 		    state->cplco[1][j] = -state->cplco[1][j];
     }

-    if ((state->acmod == 2) && (bitstream_get (1))) {	/* rematstr */
+    if ((state->acmod == 2) && (bitstream_get (state, 1))) {	/* rematstr */
 	int end;

-	end = (state->cplinu) ? state->cplstrtmant : 253;
+	state->rematflg = 0;
+	end = (state->chincpl) ? state->cplstrtmant : 253;	/* cplinu */
 	i = 0;
 	do
-	    state->rematflg[i] = bitstream_get (1);
+	    state->rematflg |= bitstream_get (state, 1) << i;
 	while (rematrix_band[i++] < end);
     }

     cplexpstr = EXP_REUSE;
     lfeexpstr = EXP_REUSE;
-    if (state->cplinu)
-	cplexpstr = bitstream_get (2);
+    if (state->chincpl)	/* cplinu */
+	cplexpstr = bitstream_get (state, 2);
     for (i = 0; i < nfchans; i++)
-	chexpstr[i] = bitstream_get (2);
+	chexpstr[i] = bitstream_get (state, 2);
     if (state->lfeon)
-	lfeexpstr = bitstream_get (1);
+	lfeexpstr = bitstream_get (state, 1);

     for (i = 0; i < nfchans; i++)
 	if (chexpstr[i] != EXP_REUSE) {
-	    if (state->cplinu && state->chincpl[i])
+	    if ((state->chincpl >> i) & 1)
 		state->endmant[i] = state->cplstrtmant;
 	    else {
 		int chbwcod;

-		chbwcod = bitstream_get (6);
+		chbwcod = bitstream_get (state, 6);
 		if (chbwcod > 60)
 		    return 1;
 		state->endmant[i] = chbwcod * 3 + 73;
@@ -651,9 +682,9 @@
 	do_bit_alloc = 64;
 	ncplgrps = ((state->cplendmant - state->cplstrtmant) /
 		    (3 << (cplexpstr - 1)));
-	cplabsexp = bitstream_get (4) << 1;
-	if (parse_exponents (cplexpstr, ncplgrps, cplabsexp,
-			     state->cpl_exp + state->cplstrtmant))
+	cplabsexp = bitstream_get (state, 4) << 1;
+	if (parse_exponents (state, cplexpstr, ncplgrps, cplabsexp,
+			     state->cpl_expbap.exp + state->cplstrtmant))
 	    return 1;
     }
     for (i = 0; i < nfchans; i++)
@@ -663,99 +694,94 @@
 	    do_bit_alloc |= 1 << i;
 	    grp_size = 3 << (chexpstr[i] - 1);
 	    nchgrps = (state->endmant[i] + grp_size - 4) / grp_size;
-	    state->fbw_exp[i][0] = bitstream_get (4);
-	    if (parse_exponents (chexpstr[i], nchgrps, state->fbw_exp[i][0],
-				 state->fbw_exp[i] + 1))
+	    state->fbw_expbap[i].exp[0] = bitstream_get (state, 4);
+	    if (parse_exponents (state, chexpstr[i], nchgrps,
+				 state->fbw_expbap[i].exp[0],
+				 state->fbw_expbap[i].exp + 1))
 		return 1;
-	    bitstream_skip (2);	/* gainrng */
+	    bitstream_skip (state, 2);	/* gainrng */
 	}
     if (lfeexpstr != EXP_REUSE) {
 	do_bit_alloc |= 32;
-	state->lfe_exp[0] = bitstream_get (4);
-	if (parse_exponents (lfeexpstr, 2, state->lfe_exp[0],
-			     state->lfe_exp + 1))
+	state->lfe_expbap.exp[0] = bitstream_get (state, 4);
+	if (parse_exponents (state, lfeexpstr, 2, state->lfe_expbap.exp[0],
+			     state->lfe_expbap.exp + 1))
 	    return 1;
     }

-    if (bitstream_get (1)) {	/* baie */
+    if (bitstream_get (state, 1)) {	/* baie */
 	do_bit_alloc = -1;
-	state->sdcycod = bitstream_get (2);
-	state->fdcycod = bitstream_get (2);
-	state->sgaincod = bitstream_get (2);
-	state->dbpbcod = bitstream_get (2);
-	state->floorcod = bitstream_get (3);
+	state->bai = bitstream_get (state, 11);
     }
-    if (bitstream_get (1)) {	/* snroffste */
+    if (bitstream_get (state, 1)) {	/* snroffste */
 	do_bit_alloc = -1;
-	state->csnroffst = bitstream_get (6);
-	if (state->cplinu) {
-	    state->cplba.fsnroffst = bitstream_get (4);
-	    state->cplba.fgaincod = bitstream_get (3);
-	}
-	for (i = 0; i < nfchans; i++) {
-	    state->ba[i].fsnroffst = bitstream_get (4);
-	    state->ba[i].fgaincod = bitstream_get (3);
-	}
-	if (state->lfeon) {
-	    state->lfeba.fsnroffst = bitstream_get (4);
-	    state->lfeba.fgaincod = bitstream_get (3);
-	}
+	state->csnroffst = bitstream_get (state, 6);
+	if (state->chincpl)	/* cplinu */
+	    state->cplba.bai = bitstream_get (state, 7);
+	for (i = 0; i < nfchans; i++)
+	    state->ba[i].bai = bitstream_get (state, 7);
+	if (state->lfeon)
+	    state->lfeba.bai = bitstream_get (state, 7);
     }
-    if ((state->cplinu) && (bitstream_get (1))) {	/* cplleake */
+    if ((state->chincpl) && (bitstream_get (state, 1))) { /* cplleake */
 	do_bit_alloc |= 64;
-	state->cplfleak = 2304 - (bitstream_get (3) << 8);
-	state->cplsleak = 2304 - (bitstream_get (3) << 8);
+	state->cplfleak = 9 - bitstream_get (state, 3);
+	state->cplsleak = 9 - bitstream_get (state, 3);
     }

-    if (bitstream_get (1)) {	/* deltbaie */
+    if (bitstream_get (state, 1)) {	/* deltbaie */
 	do_bit_alloc = -1;
-	if (state->cplinu)
-	    state->cplba.deltbae = bitstream_get (2);
+	if (state->chincpl)	/* cplinu */
+	    state->cplba.deltbae = bitstream_get (state, 2);
 	for (i = 0; i < nfchans; i++)
-	    state->ba[i].deltbae = bitstream_get (2);
-	if (state->cplinu && (state->cplba.deltbae == DELTA_BIT_NEW) &&
-	    parse_deltba (state->cplba.deltba))
+	    state->ba[i].deltbae = bitstream_get (state, 2);
+	if (state->chincpl &&	/* cplinu */
+	    (state->cplba.deltbae == DELTA_BIT_NEW) &&
+	    parse_deltba (state, state->cplba.deltba))
 	    return 1;
 	for (i = 0; i < nfchans; i++)
 	    if ((state->ba[i].deltbae == DELTA_BIT_NEW) &&
-		parse_deltba (state->ba[i].deltba))
+		parse_deltba (state, state->ba[i].deltba))
 		return 1;
     }

     if (do_bit_alloc) {
 	if (zero_snr_offsets (nfchans, state)) {
-	    memset (state->cpl_bap, 0, sizeof (state->cpl_bap));
-	    memset (state->fbw_bap, 0, sizeof (state->fbw_bap));
-	    memset (state->lfe_bap, 0, sizeof (state->lfe_bap));
+	    memset (state->cpl_expbap.bap, 0, sizeof (state->cpl_expbap.bap));
+	    for (i = 0; i < nfchans; i++)
+		memset (state->fbw_expbap[i].bap, 0,
+			sizeof (state->fbw_expbap[i].bap));
+	    memset (state->lfe_expbap.bap, 0, sizeof (state->lfe_expbap.bap));
 	} else {
-	    if (state->cplinu && (do_bit_alloc & 64))
-		bit_allocate (state, &state->cplba, state->cplstrtbnd,
+	    if (state->chincpl && (do_bit_alloc & 64))	/* cplinu */
+		a52_bit_allocate (state, &state->cplba, state->cplstrtbnd,
 			      state->cplstrtmant, state->cplendmant,
-			      state->cplfleak, state->cplsleak,
-			      state->cpl_exp, state->cpl_bap);
+				  state->cplfleak << 8, state->cplsleak << 8,
+				  &state->cpl_expbap);
 	    for (i = 0; i < nfchans; i++)
 		if (do_bit_alloc & (1 << i))
-		    bit_allocate (state, state->ba + i, 0, 0,
-				  state->endmant[i], 0, 0, state->fbw_exp[i],
-				  state->fbw_bap[i]);
+		    a52_bit_allocate (state, state->ba + i, 0, 0,
+				      state->endmant[i], 0, 0,
+				      state->fbw_expbap +i);
 	    if (state->lfeon && (do_bit_alloc & 32)) {
 		state->lfeba.deltbae = DELTA_BIT_NONE;
-		bit_allocate (state, &state->lfeba, 0, 0, 7, 0, 0,
-			      state->lfe_exp, state->lfe_bap);
+		a52_bit_allocate (state, &state->lfeba, 0, 0, 7, 0, 0,
+				  &state->lfe_expbap);
 	    }
 	}
     }

-    if (bitstream_get (1)) {	/* skiple */
-	i = bitstream_get (9);	/* skipl */
+    if (bitstream_get (state, 1)) {	/* skiple */
+	i = bitstream_get (state, 9);	/* skipl */
 	while (i--)
-	    bitstream_skip (8);
+	    bitstream_skip (state, 8);
     }

+    samples = state->samples;
     if (state->output & A52_LFE)
 	samples += 256;	/* shift for LFE channel */

-    chanbias = downmix_coeff (coeff, state->acmod, state->output,
+    chanbias = a52_downmix_coeff (coeff, state->acmod, state->output,
 			      state->dynrng, state->clev, state->slev);

     quantizer.q1_ptr = quantizer.q2_ptr = quantizer.q4_ptr = -1;
@@ -764,10 +790,10 @@
     for (i = 0; i < nfchans; i++) {
 	int j;

-	coeff_get (samples + 256 * i, state->fbw_exp[i], state->fbw_bap[i],
-		   &quantizer, coeff[i], dithflag[i], state->endmant[i]);
+	coeff_get (state, samples + 256 * i, state->fbw_expbap +i, &quantizer,
+		   coeff[i], dithflag[i], state->endmant[i]);

-	if (state->cplinu && state->chincpl[i]) {
+	if ((state->chincpl >> i) & 1) {
 	    if (!done_cpl) {
 		done_cpl = 1;
 		coeff_get_coupling (state, nfchans, coeff,
@@ -783,18 +809,21 @@
     }

     if (state->acmod == 2) {
-	int j, end, band;
+	int j, end, band, rematflg;

 	end = ((state->endmant[0] < state->endmant[1]) ?
 	       state->endmant[0] : state->endmant[1]);

 	i = 0;
 	j = 13;
+	rematflg = state->rematflg;
 	do {
-	    if (!state->rematflg[i]) {
+	    if (! (rematflg & 1)) {
+		rematflg >>= 1;
 		j = rematrix_band[i++];
 		continue;
 	    }
+	    rematflg >>= 1;
 	    band = rematrix_band[i++];
 	    if (band > end)
 		band = end;
@@ -811,15 +840,15 @@

     if (state->lfeon) {
 	if (state->output & A52_LFE) {
-	    coeff_get (samples - 256, state->lfe_exp, state->lfe_bap,
-		       &quantizer, state->dynrng, 0, 7);
+	    coeff_get (state, samples - 256, &state->lfe_expbap, &quantizer,
+		       state->dynrng, 0, 7);
 	    for (i = 7; i < 256; i++)
 		(samples-256)[i] = 0;
-	    imdct_512 (samples - 256, samples + 1536 - 256, state->bias);
+	    a52_imdct_512 (samples - 256, samples + 1536 - 256, state->bias);
 	} else {
 	    /* just skip the LFE coefficients */
-	    coeff_get (samples + 1280, state->lfe_exp, state->lfe_bap,
-		       &quantizer, 0, 0, 7);
+	    coeff_get (state, samples + 1280, &state->lfe_expbap, &quantizer,
+		       0, 0, 7);
 	}
     }

@@ -830,9 +859,9 @@
 		break;

     if (i < nfchans) {
-	if (samples[2 * 1536 - 1] == (sample_t)0x776b6e21) {
-	    samples[2 * 1536 - 1] = 0;
-	    upmix (samples + 1536, state->acmod, state->output);
+	if (state->downmixed) {
+	    state->downmixed = 0;
+	    a52_upmix (samples + 1536, state->acmod, state->output);
 	}

 	for (i = 0; i < nfchans; i++) {
@@ -844,10 +873,10 @@

 	    if (coeff[i]) {
 		if (blksw[i])
-		    imdct_256 (samples + 256 * i, samples + 1536 + 256 * i,
+		    a52_imdct_256 (samples + 256 * i, samples + 1536 + 256 * i,
 			       bias);
 		else
-		    imdct_512 (samples + 256 * i, samples + 1536 + 256 * i,
+		    a52_imdct_512 (samples + 256 * i, samples + 1536 + 256 * i,
 			       bias);
 	    } else {
 		int j;
@@ -857,29 +886,35 @@
 	    }
 	}

-	downmix (samples, state->acmod, state->output, state->bias,
+	a52_downmix (samples, state->acmod, state->output, state->bias,
 		 state->clev, state->slev);
     } else {
 	nfchans = nfchans_tbl[state->output & A52_CHANNEL_MASK];

-	downmix (samples, state->acmod, state->output, 0,
+	a52_downmix (samples, state->acmod, state->output, 0,
 		 state->clev, state->slev);

-	if (samples[2 * 1536 - 1] != (sample_t)0x776b6e21) {
-	    downmix (samples + 1536, state->acmod, state->output, 0,
+	if (!state->downmixed) {
+	    state->downmixed = 1;
+	    a52_downmix (samples + 1536, state->acmod, state->output, 0,
 		     state->clev, state->slev);
-	    samples[2 * 1536 - 1] = (sample_t)0x776b6e21;
 	}

 	if (blksw[0])
 	    for (i = 0; i < nfchans; i++)
-		imdct_256 (samples + 256 * i, samples + 1536 + 256 * i,
+		a52_imdct_256 (samples + 256 * i, samples + 1536 + 256 * i,
 			   state->bias);
 	else
 	    for (i = 0; i < nfchans; i++)
-		imdct_512 (samples + 256 * i, samples + 1536 + 256 * i,
+		a52_imdct_512 (samples + 256 * i, samples + 1536 + 256 * i,
 			   state->bias);
     }

     return 0;
 }
+
+void a52_free (a52_state_t * state)
+{
+    free (state->samples);
+    free (state);
+}
--- a/liba52/tables.h	Thu Jun 15 08:43:41 2006 +0000
+++ b/liba52/tables.h	Thu Jun 15 22:58:06 2006 +0000
@@ -1,6 +1,6 @@
 /*
  * tables.h
- * Copyright (C) 2000-2001 Michel Lespinasse <walken@zoy.org>
+ * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
  *
  * This file is part of a52dec, a free ATSC A-52 stream decoder.
@@ -21,7 +21,7 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */

-static int8_t exp_1[128] = {
+static const int8_t exp_1[128] = {
     -2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,-2,
     -1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,-1,
      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
@@ -29,7 +29,7 @@
      2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
     25,25,25
 };
-static int8_t exp_2[128] = {
+static const int8_t exp_2[128] = {
     -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
     -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
     -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
@@ -37,7 +37,7 @@
     -2,-2,-2,-2,-2,-1,-1,-1,-1,-1, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2,
     25,25,25
 };
-static int8_t exp_3[128] = {
+static const int8_t exp_3[128] = {
     -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
     -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
     -2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,-2,-1, 0, 1, 2,
@@ -182,7 +182,7 @@
     0
 };

-static sample_t scale_factor[25] = {
+static const sample_t scale_factor[25] = {
     0.000030517578125,
     0.0000152587890625,
     0.00000762939453125,
--- a/libmpcodecs/ad_liba52.c	Thu Jun 15 08:43:41 2006 +0000
+++ b/libmpcodecs/ad_liba52.c	Thu Jun 15 22:58:06 2006 +0000
@@ -19,8 +19,7 @@
 #include "liba52/a52.h"
 #include "liba52/mm_accel.h"

-static sample_t * a52_samples;
-static a52_state_t a52_state;
+static a52_state_t *a52_state;
 static uint32_t a52_flags=0;
 /** Used by a52_resample_float, it defines the mapping between liba52
  * channels and output channels.  The ith nibble from the right in the
@@ -169,8 +168,8 @@
   if(gCpuCaps.has3DNow) a52_accel|=MM_ACCEL_X86_3DNOW;
   if(gCpuCaps.has3DNowExt) a52_accel|=MM_ACCEL_X86_3DNOWEXT;
   if(gCpuCaps.hasAltiVec) a52_accel|=MM_ACCEL_PPC_ALTIVEC;
-  a52_samples=a52_init (a52_accel);
-  if (a52_samples == NULL) {
+  a52_state=a52_init (a52_accel);
+  if (a52_state == NULL) {
 	mp_msg(MSGT_DECAUDIO,MSGL_ERR,"A52 init failed\n");
 	return 0;
   }
@@ -210,7 +209,7 @@
   /* test:*/
   flags=a52_flags|A52_ADJUST_LEVEL;
   mp_msg(MSGT_DECAUDIO,MSGL_V,"A52 flags before a52_frame: 0x%X\n",flags);
-  if (a52_frame (&a52_state, sh_audio->a_in_buffer, &flags, &level, bias)){
+  if (a52_frame (a52_state, sh_audio->a_in_buffer, &flags, &level, bias)){
     mp_msg(MSGT_DECAUDIO,MSGL_ERR,"a52: error decoding frame -> nosound\n");
     return 0;
   }
@@ -288,16 +287,12 @@
     sample_t level=a52_level, bias=384;
     int flags=a52_flags|A52_ADJUST_LEVEL;
     int i,len=-1;
-	if (maxlen / sh_audio->samplesize / 256 / sh_audio->channels < 6) {
-	    mp_msg(MSGT_DECAUDIO, MSGL_V, "maxlen too small in decode_audio\n");
-	    return len;
-	}
 	if (sh_audio->sample_format == AF_FORMAT_FLOAT_NE)
 	    bias = 0;
 	if(!sh_audio->a_in_buffer_len)
 	    if(a52_fillbuff(sh_audio)<0) return len; /* EOF */
 	sh_audio->a_in_buffer_len=0;
-	if (a52_frame (&a52_state, sh_audio->a_in_buffer, &flags, &level, bias)){
+	if (a52_frame (a52_state, sh_audio->a_in_buffer, &flags, &level, bias)){
 	    mp_msg(MSGT_DECAUDIO,MSGL_WARN,"a52: error decoding frame\n");
 	    return len;
 	}
@@ -305,18 +300,18 @@
 	/* handle dynrng */
 	if (a52_drc_action != DRC_NO_ACTION) {
 	    if (a52_drc_action == DRC_NO_COMPRESSION)
-		a52_dynrng(&a52_state, NULL, NULL);
+		a52_dynrng(a52_state, NULL, NULL);
 	    else
-		a52_dynrng(&a52_state, dynrng_call, NULL);
+		a52_dynrng(a52_state, dynrng_call, NULL);
 	}

 	len=0;
 	for (i = 0; i < 6; i++) {
-	    if (a52_block (&a52_state, a52_samples)){
+	    if (a52_block (a52_state)){
 		mp_msg(MSGT_DECAUDIO,MSGL_WARN,"a52: error at resampling\n");
 		break;
 	    }
-	    len+=2*a52_resample(a52_samples,(int16_t *)&buf[len]);
+	    len+=2*a52_resample(a52_samples(a52_state),(int16_t *)&buf[len]);
 	}
 	assert(len <= maxlen);
   return len;