transform.c

mpeg层3音频编码程序

        u_p[ 6][0] = -d10 +d13;
/*24*/  u_p[ 8][0] =d3;
        u_p[10][0] = -d8 -d9 +d11 -d13;
/*28*/  u_p[12][0] = +d7;
        u_p[14][0] = +d15;

        /* the other 32 are stored for use with the next granule
         */

        u_p = (float (*)[16]) &u[ch][!div][0][start];

/*0*/   u_p[16][0] = d0;
        u_p[14][0] = -(+d8 );
/*4*/   u_p[12][0] = -(+d4 );
        u_p[10][0] = -(-d8 +d12 );
/*8*/   u_p[ 8][0] = -(+d2 );
        u_p[ 6][0] = -(+d8 +d10 -d12 );
/*12*/  u_p[ 4][0] = -(-d4 +d6 );
        u_p[ 2][0] = -d14;
        u_p[ 0][0] = -d1;
	}

	{
	float c0,c1,c2,c3,c4,c5,c6,c7,c8,c9,c10,c11,c12,c13,c14,c15;

/* memory separation, second part
*/
/* 2
*/
        c0=d16 + d24; c8= (d16 - d24) *  b2;
        c1=d17 + d25; c9= (d17 - d25) *  b6;
        c2=d18 + d26; c10= (d18 - d26) * b14;
        c3=d19 + d27; c11= (d19 - d27) * b10;
        c4=d20 + d28; c12= (d20 - d28) * b30;
        c5=d21 + d29; c13= (d21 - d29) * b26;
        c6=d22 + d30; c14= (d22 - d30) * b18;
        c7=d23 + d31; c15= (d23 - d31) * b22;

/* 3
*/
        d16= c0+ c4; d20= (c0 - c4) *  b4;
        d17= c1+ c5; d21= (c1 - c5) * b12;
        d18= c2+ c6; d22= (c2 - c6) * b28;
        d19= c3+ c7; d23= (c3 - c7) * b20;

        d24= c8+ c12; d28= (c8 - c12) *  b4;
        d25= c9+ c13; d29= (c9 - c13) * b12;
        d26= c10+ c14; d30= (c10 - c14) * b28;
        d27= c11+ c15; d31= (c11 - c15) * b20;

/* 4
*/
        {
        float rb8 = b8;
        float rb24 = b24;

/**/    c0= d16+ d18; c2= (d16 - d18) *  rb8;
        c1= d17+ d19; c3= (d17 - d19) * rb24;
/**/    c4= d20+ d22; c6= (d20 - d22) *  rb8;
        c5= d21+ d23; c7= (d21 - d23) * rb24;
/**/    c8= d24+ d26; c10= (d24 - d26) *  rb8;
        c9= d25+ d27; c11= (d25 - d27) * rb24;
/**/    c12= d28+ d30; c14= (d28 - d30) *  rb8;
        c13= d29+ d31; c15= (d29 - d31) * rb24;
        }

/* 5
*/
        {
        float rb16 = b16;
        d16= c0+ c1; d17= (c0 - c1) * rb16;
        d18= c2+ c3; d19= (c2 - c3) * rb16;

        d20= c4+ c5; d21= (c4 - c5) * rb16;
        d20+=d16; d21+=d17;
        d22= c6+ c7; d23= (c6 - c7) * rb16;
        d22+=d16; d22+=d18;
        d23+=d16; d23+=d17; d23+=d19;


        d24= c8+ c9; d25= (c8 - c9) * rb16;
        d26= c10+ c11; d27= (c10 - c11) * rb16;
        d26+=d24;
        d27+=d24; d27+=d25;

        d28= c12+ c13; d29= (c12 - c13) * rb16;
        d28-=d20; d29+=d28; d29-=d21;
        d30= c14+ c15; d31= (c14 - c15) * rb16;
        d30-=d22;
        d31-=d23;
	}

	/* step 6: final resolving & reordering 
	 * the other 32 are stored for use with the next granule
	 */
	
	u_p = (float (*)[16]) &u[ch][!div][0][start];

	u_p[ 1][0] = -(+d30 );	
	u_p[ 3][0] = -(+d22 -d26 );
	u_p[ 5][0] = -(-d18 -d20 +d26 );
	u_p[ 7][0] = -(+d18 -d28 );
	u_p[ 9][0] = -(+d28 );
	u_p[11][0] = -(+d20 -d24 );
	u_p[13][0] = -(-d16 +d24 );
	u_p[15][0] = -(+d16 );

	/* the other 32 are stored for use with the next granule
	 */

	u_p = (float (*)[16]) &u[ch][div][0][start];

	u_p[15][0] = +d31;
	u_p[13][0] = +d23 -d27;
	u_p[11][0] = -d19 -d20 -d21 +d27;
	u_p[ 9][0] = +d19 -d29;
	u_p[ 7][0] = -d18 +d29;
	u_p[ 5][0] = +d18 +d20 +d21 -d25 -d26;
	u_p[ 3][0] = -d17 -d22 +d25 +d26;
	u_p[ 1][0] = +d17 -d30;
	}
	}

#if defined(PENTIUM_RDTSC)
	__asm__(".byte 0x0f,0x31" : "=a" (cnt3), "=d" (cnt4));
#endif

	/* we're doing dewindowing and calculating final samples now
	 */

#if defined(ARCH_i586)
	/* x86 assembler optimisations.  These optimisations are tuned
           specifically for Intel Pentiums. */

	    asm("movl $15,%%eax\n\t"\
        "1:\n\t"\
	"flds (%0)\n\t"\
	"fmuls (%1)\n\t"\
	"flds 4(%0)\n\t"\
	"fmuls 4(%1)\n\t"\
	"flds 8(%0)\n\t"\
	"fmuls 8(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 12(%0)\n\t"\
	"fmuls 12(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 16(%0)\n\t"\
	"fmuls 16(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 20(%0)\n\t"\
	"fmuls 20(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 24(%0)\n\t"\
	"fmuls 24(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 28(%0)\n\t"\
	"fmuls 28(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 32(%0)\n\t"\
	"fmuls 32(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 36(%0)\n\t"\
	"fmuls 36(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 40(%0)\n\t"\
	"fmuls 40(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 44(%0)\n\t"\
	"fmuls 44(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 48(%0)\n\t"\
	"fmuls 48(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 52(%0)\n\t"\
	"fmuls 52(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 56(%0)\n\t"\
	"fmuls 56(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 60(%0)\n\t"\
	"fmuls 60(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"addl $64,%0\n\t"\
	"addl $128,%1\n\t"\
	"subl $4,%%esp\n\t"\
	"faddp\n\t"\
	"fistpl (%%esp)\n\t"\
	"popl %%ecx\n\t"\
	"cmpl $32767,%%ecx\n\t"\
	"jle 2f\n\t"\
	"movw $32767,%%cx\n\t"\
	"jmp 3f\n\t"\
	"2: cmpl $-32768,%%ecx\n\t"\
	"jge 3f\n\t"\
	"movw $-32768,%%cx\n\t"\
	"3: movw %%cx,(%2)\n\t"\
	"addl %3,%2\n\t"\
	"decl %%eax\n\t"\
	"jns 1b\n\t"\

	"testb $1,%4\n\t"\
        "je 4f\n\t"

	"flds (%0)\n\t"\
	"fmuls (%1)\n\t"\
	"flds 8(%0)\n\t"\
	"fmuls 8(%1)\n\t"\
	"flds 16(%0)\n\t"\
	"fmuls 16(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 24(%0)\n\t"\
	"fmuls 24(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 32(%0)\n\t"\
	"fmuls 32(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 40(%0)\n\t"\
	"fmuls 40(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 48(%0)\n\t"\
	"fmuls 48(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 56(%0)\n\t"\
	"fmuls 56(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"subl $4,%%esp\n\t"\
	"subl $64,%0\n\t"\
	"subl $192,%1\n\t"\
	"faddp\n\t"\
	"fistpl (%%esp)\n\t"\
	"popl %%ecx\n\t"\
	"cmpl $32767,%%ecx\n\t"\
	"jle 2f\n\t"\
	"movw $32767,%%cx\n\t"\
	"jmp 3f\n\t"\
	"2: cmpl $-32768,%%ecx\n\t"\
	"jge 3f\n\t"\
	"movw $-32768,%%cx\n\t"\
	"3: movw %%cx,(%2)\n\t"\

	"movl %5,%%ecx\n\t"\
	"sall $3,%%ecx\n\t"\
	"addl %%ecx,%1\n\t"\
	"addl %3,%2\n\t"\
	"movl $14,%%eax\n\t"\

	"1:flds 4(%0)\n\t"\
	"fmuls 56(%1)\n\t"\
	"flds (%0)\n\t"\
	"fmuls 60(%1)\n\t"\
	"flds 12(%0)\n\t"\
	"fmuls 48(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubp\n\t"\
	"flds 8(%0)\n\t"\
	"fmuls 52(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 20(%0)\n\t"\
	"fmuls 40(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 16(%0)\n\t"\
	"fmuls 44(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 28(%0)\n\t"\
	"fmuls 32(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 24(%0)\n\t"\
	"fmuls 36(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 36(%0)\n\t"\
	"fmuls 24(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 32(%0)\n\t"\
	"fmuls 28(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 44(%0)\n\t"\
	"fmuls 16(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 40(%0)\n\t"\
	"fmuls 20(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 52(%0)\n\t"\
	"fmuls 8(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 48(%0)\n\t"\
	"fmuls 12(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 60(%0)\n\t"\
	"fmuls (%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 56(%0)\n\t"\
	"fmuls 4(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"subl $64,%0\n\t"\
	"subl $128,%1\n\t"\
	"subl $4,%%esp\n\t"\
	"fsubp\n\t"\
	"fistpl (%%esp)\n\t"\
	"popl %%ecx\n\t"\
	"cmpl $32767,%%ecx\n\t"\
	"jle 2f\n\t"\
	"movw $32767,%%cx\n\t"\
	"jmp 3f\n\t"\
	"2: cmpl $-32768,%%ecx\n\t"\
	"jge 3f\n\t"\
	"movw $-32768,%%cx\n\t"\
	"3: movw %%cx,(%2)\n\t"\
	"addl %3,%2\n\t"\
	"decl %%eax\n\t"\
	"jns 1b\n\t"\
	"jmp 5f\n\t"\

	"4:flds 4(%0)\n\t"\
	"fmuls 4(%1)\n\t"\
	"flds 12(%0)\n\t"\
	"fmuls 12(%1)\n\t"\
	"flds 20(%0)\n\t"\
	"fmuls 20(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 28(%0)\n\t"\
	"fmuls 28(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 36(%0)\n\t"\
	"fmuls 36(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 44(%0)\n\t"\
	"fmuls 44(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 52(%0)\n\t"\
	"fmuls 52(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 60(%0)\n\t"\
	"fmuls 60(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"subl $4,%%esp\n\t"\
	"subl $64,%0\n\t"\
	"subl $192,%1\n\t"\
	"faddp\n\t"\
	"fistpl (%%esp)\n\t"\
	"popl %%ecx\n\t"\
	"cmpl $32767,%%ecx\n\t"\
	"jle 2f\n\t"\
	"movw $32767,%%cx\n\t"\
	"jmp 3f\n\t"\
	"2: cmpl $-32768,%%ecx\n\t"\
	"jge 3f\n\t"\
	"movw $-32768,%%cx\n\t"\
	"3: movw %%cx,(%2)\n\t"\

	"movl %5,%%ecx\n\t"\
	"sall $3,%%ecx\n\t"\
	"addl %%ecx,%1\n\t"\
	"addl %3,%2\n\t"\

	"movl $14,%%eax\n\t"\
	"1:flds (%0)\n\t"\
	"fmuls 60(%1)\n\t"\
	"flds 4(%0)\n\t"\
	"fmuls 56(%1)\n\t"\
	"flds 8(%0)\n\t"\
	"fmuls 52(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubp\n\t"\
	"flds 12(%0)\n\t"\
	"fmuls 48(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 16(%0)\n\t"\
	"fmuls 44(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 20(%0)\n\t"\
	"fmuls 40(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 24(%0)\n\t"\
	"fmuls 36(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 28(%0)\n\t"\
	"fmuls 32(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 32(%0)\n\t"\
	"fmuls 28(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 36(%0)\n\t"\
	"fmuls 24(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 40(%0)\n\t"\
	"fmuls 20(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 44(%0)\n\t"\
	"fmuls 16(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 48(%0)\n\t"\
	"fmuls 12(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 52(%0)\n\t"\
	"fmuls 8(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"flds 56(%0)\n\t"\
	"fmuls 4(%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"fsubrp\n\t"\
	"flds 60(%0)\n\t"\
	"fmuls (%1)\n\t"\
	"fxch %%st(2)\n\t"\
	"faddp\n\t"\
	"subl $64,%0\n\t"\
	"subl $128,%1\n\t"\
	"subl $4,%%esp\n\t"\
	"fsubp\n\t"\
	"fistpl (%%esp)\n\t"\
	"popl %%ecx\n\t"\
	"cmpl $32767,%%ecx\n\t"\
	"jle 2f\n\t"\
	"movw $32767,%%cx\n\t"\
	"jmp 3f\n\t"\
	"2: cmpl $-32768,%%ecx\n\t"\
	"jge 3f\n\t"\
	"movw $-32768,%%cx\n\t"\
	"3: movw %%cx,(%2)\n\t"\
	"addl %3,%2\n\t"\
	"decl %%eax\n\t"\
	"jns 1b\n\t"\

	"5:"\
	    : : "b" (u[ch][div]), "d" (t_dewindow[0] + 16 - start), "S" (&sample_buffer[f>>(2-nch)][nch==2?0:(f&1?16:0)][ch]), "m" (sizeof(short) * nch), "m" (div), "m" (start)\
	    : "eax", "ecx", "memory");
#else
	{
	  short *samples = (&sample_buffer[f>>(2-nch)][nch==2?0:(f&1?16:0)][ch]);
	  int out, j;

#define PUT_SAMPLE(out)			\
		if (out > 32767)	\
		  *samples = 32767;	\
		else			\
		  if (out < -32768)	\
		    *samples = -32768;	\
		  else			\
		    *samples = out;	\
					\
		samples += nch;

#if defined(SUPERHACK)
	  /* These is a simple implementation which should be nicer to the
             cache; computation of samples are done in one pass rather than
             two.  However, for various reasons which I do not have time to
             investigate, it runs quite a lot slower than two pass
             computations.  If you have time, you are welcome to look into
             it. */

          {
            float (*u_ptr)[16] = u[ch][div];
	    const float *dewindow2 = t_dewindow[0] + start;

	    {
	      float outf1, outf2, outf3, outf4;

	      outf1  = u_ptr[0][ 0] * dewindow[0x0];
	      outf2  = u_ptr[0][ 1] * dewindow[0x1];