seal 0.1.0

data/mpg123/src/libmpg123/dct64_altivec.c

@@ -0,0 +1,315 @@
+/*
+	dct64_altivec.c: Discrete Cosine Tansform (DCT) for Altivec
+
+	copyright ?-2006 by the mpg123 project - free software under the terms of the LGPL 2.1
+	see COPYING and AUTHORS files in distribution or http://mpg123.org
+	initially written by Michael Hipp
+	altivec optimization by tmkk
+*/
+
+/*
+ * Discrete Cosine Tansform (DCT) for subband synthesis
+ *
+ * -funroll-loops (for gcc) will remove the loops for better performance
+ * using loops in the source-code enhances readabillity
+ *
+ *
+ * TODO: write an optimized version for the down-sampling modes
+ *       (in these modes the bands 16-31 (2:1) or 8-31 (4:1) are zero 
+ */
+
+#include "mpg123lib_intern.h"
+
+#ifndef __APPLE__
+#include <altivec.h>
+#endif
+
+void dct64_altivec(real *out0,real *out1,real *samples)
+{
+  ALIGNED(16) real bufs[32];
+
+	{
+		register real *b1,*costab;
+		
+		vector unsigned char vinvert,vperm1,vperm2,vperm3,vperm4;
+		vector float v1,v2,v3,v4,v5,v6,v7,v8;
+		vector float vbs1,vbs2,vbs3,vbs4,vbs5,vbs6,vbs7,vbs8;
+		vector float vbs9,vbs10,vbs11,vbs12,vbs13,vbs14,vbs15,vbs16;
+		vector float vzero;
+		b1 = samples;
+		costab = pnts[0];
+		
+		vzero = vec_xor(vzero,vzero);
+#ifdef __APPLE__
+		vinvert = (vector unsigned char)(12,13,14,15,8,9,10,11,4,5,6,7,0,1,2,3);
+#else
+		vinvert = (vector unsigned char){12,13,14,15,8,9,10,11,4,5,6,7,0,1,2,3};
+#endif
+		vperm1 = vec_lvsl(0,b1);
+		vperm2 = vec_perm(vperm1,vperm1,vinvert);
+		
+		v1 = vec_ld(0,b1);
+		v2 = vec_ld(16,b1);
+		v3 = vec_ld(112,b1);
+		v4 = vec_ld(127,b1);
+		v5 = vec_perm(v1,v2,vperm1); /* b1[0,1,2,3] */
+		v6 = vec_perm(v3,v4,vperm2); /* b1[31,30,29,28] */
+		
+		vbs1 = vec_add(v5,v6);
+		vbs8 = vec_sub(v5,v6);
+		
+		v1 = vec_ld(32,b1);
+		v4 = vec_ld(96,b1);
+		v5 = vec_perm(v2,v1,vperm1); /* b1[4,5,6,7] */
+		v6 = vec_perm(v4,v3,vperm2); /* b1[27,26,25,24] */
+		
+		vbs2 = vec_add(v5,v6);
+		vbs7 = vec_sub(v5,v6);
+		
+		v2 = vec_ld(48,b1);
+		v3 = vec_ld(80,b1);
+		v5 = vec_perm(v1,v2,vperm1); /* b1[8,9,10,11] */
+		v6 = vec_perm(v3,v4,vperm2); /* b1[23,22,21,20] */
+		
+		vbs3 = vec_add(v5,v6);
+		vbs6 = vec_sub(v5,v6);
+		
+		v1 = vec_ld(64,b1);
+		v5 = vec_perm(v2,v1,vperm1); /* b1[12,13,14,15] */
+		v6 = vec_perm(v1,v3,vperm2); /* b1[19,18,17,16] */
+		
+		vbs4 = vec_add(v5,v6);
+		vbs5 = vec_sub(v5,v6);
+		
+		v1 = vec_ld(0,costab);
+		vbs8 = vec_madd(vbs8,v1,vzero);
+		v2 = vec_ld(16,costab);
+		vbs7 = vec_madd(vbs7,v2,vzero);
+		v3 = vec_ld(32,costab);
+		vbs6 = vec_madd(vbs6,v3,vzero);
+		v4 = vec_ld(48,costab);
+		vbs5 = vec_madd(vbs5,v4,vzero);
+		vbs6 = vec_perm(vbs6,vbs6,vinvert);
+		vbs5 = vec_perm(vbs5,vbs5,vinvert);
+		
+		
+		costab = pnts[1];
+		
+		v1 = vec_perm(vbs4,vbs4,vinvert);
+		vbs9 = vec_add(vbs1,v1);
+		v3 = vec_sub(vbs1,v1);
+		v5 = vec_ld(0,costab);
+		v2 = vec_perm(vbs3,vbs3,vinvert);
+		vbs10 = vec_add(vbs2,v2);
+		v4 = vec_sub(vbs2,v2);
+		v6 = vec_ld(16,costab);
+		vbs12 = vec_madd(v3,v5,vzero);
+		vbs11 = vec_madd(v4,v6,vzero);
+		
+		v7 = vec_sub(vbs7,vbs6);
+		v8 = vec_sub(vbs8,vbs5);
+		vbs13 = vec_add(vbs5,vbs8);
+		vbs14 = vec_add(vbs6,vbs7);
+		vbs15 = vec_madd(v7,v6,vzero);
+		vbs16 = vec_madd(v8,v5,vzero);
+		
+		
+		costab = pnts[2];
+		
+		v1 = vec_perm(vbs10,vbs10,vinvert);
+		v5 = vec_perm(vbs14,vbs14,vinvert);
+		vbs1 = vec_add(v1,vbs9);
+		vbs5 = vec_add(v5,vbs13);
+		v2 = vec_sub(vbs9,v1);
+		v6 = vec_sub(vbs13,v5);
+		v3 = vec_ld(0,costab);
+		vbs11 = vec_perm(vbs11,vbs11,vinvert);
+		vbs15 = vec_perm(vbs15,vbs15,vinvert);
+		vbs3 = vec_add(vbs11,vbs12);
+		vbs7 = vec_add(vbs15,vbs16);
+		v4 = vec_sub(vbs12,vbs11);
+		v7 = vec_sub(vbs16,vbs15);
+		vbs2 = vec_madd(v2,v3,vzero);
+		vbs4 = vec_madd(v4,v3,vzero);
+		vbs6 = vec_madd(v6,v3,vzero);
+		vbs8 = vec_madd(v7,v3,vzero);
+		
+		vbs2 = vec_perm(vbs2,vbs2,vinvert);
+		vbs4 = vec_perm(vbs4,vbs4,vinvert);
+		vbs6 = vec_perm(vbs6,vbs6,vinvert);
+		vbs8 = vec_perm(vbs8,vbs8,vinvert);
+		
+		
+		costab = pnts[3];
+		
+#ifdef __APPLE__
+		vperm1 = (vector unsigned char)(0,1,2,3,4,5,6,7,16,17,18,19,20,21,22,23);
+		vperm2 = (vector unsigned char)(12,13,14,15,8,9,10,11,28,29,30,31,24,25,26,27);
+		vperm3 = (vector unsigned char)(0,1,2,3,4,5,6,7,20,21,22,23,16,17,18,19);
+#else
+		vperm1 = (vector unsigned char){0,1,2,3,4,5,6,7,16,17,18,19,20,21,22,23};
+		vperm2 = (vector unsigned char){12,13,14,15,8,9,10,11,28,29,30,31,24,25,26,27};
+		vperm3 = (vector unsigned char){0,1,2,3,4,5,6,7,20,21,22,23,16,17,18,19};
+#endif
+		vperm4 = vec_add(vperm3,vec_splat_u8(8));
+		
+		v1 = vec_ld(0,costab);
+		v2 = vec_splat(v1,0);
+		v3 = vec_splat(v1,1);
+		v1 = vec_mergeh(v2,v3);
+		
+		v2 = vec_perm(vbs1,vbs3,vperm1);
+		v3 = vec_perm(vbs2,vbs4,vperm1);
+		v4 = vec_perm(vbs1,vbs3,vperm2);
+		v5 = vec_perm(vbs2,vbs4,vperm2);
+		v6 = vec_sub(v2,v4);
+		v7 = vec_sub(v3,v5);
+		v2 = vec_add(v2,v4);
+		v3 = vec_add(v3,v5);
+		v4 = vec_madd(v6,v1,vzero);
+		v5 = vec_nmsub(v7,v1,vzero);
+		vbs9 = vec_perm(v2,v4,vperm3);
+		vbs11 = vec_perm(v2,v4,vperm4);
+		vbs10 = vec_perm(v3,v5,vperm3);
+		vbs12 = vec_perm(v3,v5,vperm4);
+		
+		v2 = vec_perm(vbs5,vbs7,vperm1);
+		v3 = vec_perm(vbs6,vbs8,vperm1);
+		v4 = vec_perm(vbs5,vbs7,vperm2);
+		v5 = vec_perm(vbs6,vbs8,vperm2);
+		v6 = vec_sub(v2,v4);
+		v7 = vec_sub(v3,v5);
+		v2 = vec_add(v2,v4);
+		v3 = vec_add(v3,v5);
+		v4 = vec_madd(v6,v1,vzero);
+		v5 = vec_nmsub(v7,v1,vzero);
+		vbs13 = vec_perm(v2,v4,vperm3);
+		vbs15 = vec_perm(v2,v4,vperm4);
+		vbs14 = vec_perm(v3,v5,vperm3);
+		vbs16 = vec_perm(v3,v5,vperm4);
+		
+		
+		costab = pnts[4];
+		
+		v1 = vec_lde(0,costab);
+#ifdef __APPLE__
+		v2 = (vector float)(1.0f,-1.0f,1.0f,-1.0f);
+#else
+		v2 = (vector float){1.0f,-1.0f,1.0f,-1.0f};
+#endif
+		v3 = vec_splat(v1,0);
+		v1 = vec_madd(v2,v3,vzero);
+		
+		v2 = vec_mergeh(vbs9,vbs10);
+		v3 = vec_mergel(vbs9,vbs10);
+		v4 = vec_mergeh(vbs11,vbs12);
+		v5 = vec_mergel(vbs11,vbs12);
+		v6 = vec_mergeh(v2,v3);
+		v7 = vec_mergel(v2,v3);
+		v2 = vec_mergeh(v4,v5);
+		v3 = vec_mergel(v4,v5); 
+		v4 = vec_sub(v6,v7);
+		v5 = vec_sub(v2,v3);
+		v6 = vec_add(v6,v7);
+		v7 = vec_add(v2,v3);
+		v2 = vec_madd(v4,v1,vzero);
+		v3 = vec_madd(v5,v1,vzero);
+		vbs1 = vec_mergeh(v6,v2);
+		vbs2 = vec_mergel(v6,v2);
+		vbs3 = vec_mergeh(v7,v3);
+		vbs4 = vec_mergel(v7,v3);
+		
+		v2 = vec_mergeh(vbs13,vbs14);
+		v3 = vec_mergel(vbs13,vbs14);
+		v4 = vec_mergeh(vbs15,vbs16);
+		v5 = vec_mergel(vbs15,vbs16);
+		v6 = vec_mergeh(v2,v3);
+		v7 = vec_mergel(v2,v3);
+		v2 = vec_mergeh(v4,v5);
+		v3 = vec_mergel(v4,v5); 
+		v4 = vec_sub(v6,v7);
+		v5 = vec_sub(v2,v3);
+		v6 = vec_add(v6,v7);
+		v7 = vec_add(v2,v3);
+		v2 = vec_madd(v4,v1,vzero);
+		v3 = vec_madd(v5,v1,vzero);
+		vbs5 = vec_mergeh(v6,v2);
+		vbs6 = vec_mergel(v6,v2);
+		vbs7 = vec_mergeh(v7,v3);
+		vbs8 = vec_mergel(v7,v3);
+		
+		vec_st(vbs1,0,bufs);
+		vec_st(vbs2,16,bufs);
+		vec_st(vbs3,32,bufs);
+		vec_st(vbs4,48,bufs);
+		vec_st(vbs5,64,bufs);
+		vec_st(vbs6,80,bufs);
+		vec_st(vbs7,96,bufs);
+		vec_st(vbs8,112,bufs);
+	}
+
+ {
+  register real *b1;
+  register int i;
+
+  for(b1=bufs,i=8;i;i--,b1+=4)
+    b1[2] += b1[3];
+
+  for(b1=bufs,i=4;i;i--,b1+=8)
+  {
+    b1[4] += b1[6];
+    b1[6] += b1[5];
+    b1[5] += b1[7];
+  }
+
+  for(b1=bufs,i=2;i;i--,b1+=16)
+  {
+    b1[8]  += b1[12];
+    b1[12] += b1[10];
+    b1[10] += b1[14];
+    b1[14] += b1[9];
+    b1[9]  += b1[13];
+    b1[13] += b1[11];
+    b1[11] += b1[15];
+  }
+ }
+
+
+  out0[0x10*16] = bufs[0];
+  out0[0x10*15] = bufs[16+0]  + bufs[16+8];
+  out0[0x10*14] = bufs[8];
+  out0[0x10*13] = bufs[16+8]  + bufs[16+4];
+  out0[0x10*12] = bufs[4];
+  out0[0x10*11] = bufs[16+4]  + bufs[16+12];
+  out0[0x10*10] = bufs[12];
+  out0[0x10* 9] = bufs[16+12] + bufs[16+2];
+  out0[0x10* 8] = bufs[2];
+  out0[0x10* 7] = bufs[16+2]  + bufs[16+10];
+  out0[0x10* 6] = bufs[10];
+  out0[0x10* 5] = bufs[16+10] + bufs[16+6];
+  out0[0x10* 4] = bufs[6];
+  out0[0x10* 3] = bufs[16+6]  + bufs[16+14];
+  out0[0x10* 2] = bufs[14];
+  out0[0x10* 1] = bufs[16+14] + bufs[16+1];
+  out0[0x10* 0] = bufs[1];
+
+  out1[0x10* 0] = bufs[1];
+  out1[0x10* 1] = bufs[16+1]  + bufs[16+9];
+  out1[0x10* 2] = bufs[9];
+  out1[0x10* 3] = bufs[16+9]  + bufs[16+5];
+  out1[0x10* 4] = bufs[5];
+  out1[0x10* 5] = bufs[16+5]  + bufs[16+13];
+  out1[0x10* 6] = bufs[13];
+  out1[0x10* 7] = bufs[16+13] + bufs[16+3];
+  out1[0x10* 8] = bufs[3];
+  out1[0x10* 9] = bufs[16+3]  + bufs[16+11];
+  out1[0x10*10] = bufs[11];
+  out1[0x10*11] = bufs[16+11] + bufs[16+7];
+  out1[0x10*12] = bufs[7];
+  out1[0x10*13] = bufs[16+7]  + bufs[16+15];
+  out1[0x10*14] = bufs[15];
+  out1[0x10*15] = bufs[16+15];
+
+}
+
+

data/mpg123/src/libmpg123/dct64_i386.c

@@ -0,0 +1,336 @@
+/*
+	dct64_i386.c: DCT64, a C variant for i386
+
+	copyright ?-2006 by the mpg123 project - free software under the terms of the LGPL 2.1
+	see COPYING and AUTHORS files in distribution or http://mpg123.org
+	initially written by Michael Hipp
+*/
+
+/*
+ * Discrete Cosine Tansform (DCT) for subband synthesis
+ * optimized for machines with no auto-increment. 
+ * The performance is highly compiler dependend. Maybe
+ * the dct64.c version for 'normal' processor may be faster
+ * even for Intel processors.
+ */
+
+#include "mpg123lib_intern.h"
+
+static void dct64_1(real *out0,real *out1,real *b1,real *b2,real *samples)
+{
+ {
+  register real *costab = pnts[0];
+
+  b1[0x00] = samples[0x00] + samples[0x1F];
+  b1[0x01] = samples[0x01] + samples[0x1E];
+  b1[0x1F] = REAL_MUL(samples[0x00] - samples[0x1F], costab[0x0]);
+  b1[0x1E] = REAL_MUL(samples[0x01] - samples[0x1E], costab[0x1]);
+
+  b1[0x02] = samples[0x02] + samples[0x1D];
+  b1[0x03] = samples[0x03] + samples[0x1C];
+  b1[0x1D] = REAL_MUL(samples[0x02] - samples[0x1D], costab[0x2]);
+  b1[0x1C] = REAL_MUL(samples[0x03] - samples[0x1C], costab[0x3]);
+
+  b1[0x04] = samples[0x04] + samples[0x1B];
+  b1[0x05] = samples[0x05] + samples[0x1A];
+  b1[0x1B] = REAL_MUL(samples[0x04] - samples[0x1B], costab[0x4]);
+  b1[0x1A] = REAL_MUL(samples[0x05] - samples[0x1A], costab[0x5]);
+
+  b1[0x06] = samples[0x06] + samples[0x19];
+  b1[0x07] = samples[0x07] + samples[0x18];
+  b1[0x19] = REAL_MUL(samples[0x06] - samples[0x19], costab[0x6]);
+  b1[0x18] = REAL_MUL(samples[0x07] - samples[0x18], costab[0x7]);
+
+  b1[0x08] = samples[0x08] + samples[0x17];
+  b1[0x09] = samples[0x09] + samples[0x16];
+  b1[0x17] = REAL_MUL(samples[0x08] - samples[0x17], costab[0x8]);
+  b1[0x16] = REAL_MUL(samples[0x09] - samples[0x16], costab[0x9]);
+
+  b1[0x0A] = samples[0x0A] + samples[0x15];
+  b1[0x0B] = samples[0x0B] + samples[0x14];
+  b1[0x15] = REAL_MUL(samples[0x0A] - samples[0x15], costab[0xA]);
+  b1[0x14] = REAL_MUL(samples[0x0B] - samples[0x14], costab[0xB]);
+
+  b1[0x0C] = samples[0x0C] + samples[0x13];
+  b1[0x0D] = samples[0x0D] + samples[0x12];
+  b1[0x13] = REAL_MUL(samples[0x0C] - samples[0x13], costab[0xC]);
+  b1[0x12] = REAL_MUL(samples[0x0D] - samples[0x12], costab[0xD]);
+
+  b1[0x0E] = samples[0x0E] + samples[0x11];
+  b1[0x0F] = samples[0x0F] + samples[0x10];
+  b1[0x11] = REAL_MUL(samples[0x0E] - samples[0x11], costab[0xE]);
+  b1[0x10] = REAL_MUL(samples[0x0F] - samples[0x10], costab[0xF]);
+
+ }
+
+
+ {
+  register real *costab = pnts[1];
+
+  b2[0x00] = b1[0x00] + b1[0x0F]; 
+  b2[0x01] = b1[0x01] + b1[0x0E]; 
+  b2[0x0F] = REAL_MUL(b1[0x00] - b1[0x0F], costab[0]);
+  b2[0x0E] = REAL_MUL(b1[0x01] - b1[0x0E], costab[1]);
+
+  b2[0x02] = b1[0x02] + b1[0x0D]; 
+  b2[0x03] = b1[0x03] + b1[0x0C]; 
+  b2[0x0D] = REAL_MUL(b1[0x02] - b1[0x0D], costab[2]);
+  b2[0x0C] = REAL_MUL(b1[0x03] - b1[0x0C], costab[3]);
+
+  b2[0x04] = b1[0x04] + b1[0x0B]; 
+  b2[0x05] = b1[0x05] + b1[0x0A]; 
+  b2[0x0B] = REAL_MUL(b1[0x04] - b1[0x0B], costab[4]);
+  b2[0x0A] = REAL_MUL(b1[0x05] - b1[0x0A], costab[5]);
+
+  b2[0x06] = b1[0x06] + b1[0x09]; 
+  b2[0x07] = b1[0x07] + b1[0x08]; 
+  b2[0x09] = REAL_MUL(b1[0x06] - b1[0x09], costab[6]);
+  b2[0x08] = REAL_MUL(b1[0x07] - b1[0x08], costab[7]);
+
+  /* */
+
+  b2[0x10] = b1[0x10] + b1[0x1F];
+  b2[0x11] = b1[0x11] + b1[0x1E];
+  b2[0x1F] = REAL_MUL(b1[0x1F] - b1[0x10], costab[0]);
+  b2[0x1E] = REAL_MUL(b1[0x1E] - b1[0x11], costab[1]);
+
+  b2[0x12] = b1[0x12] + b1[0x1D];
+  b2[0x13] = b1[0x13] + b1[0x1C];
+  b2[0x1D] = REAL_MUL(b1[0x1D] - b1[0x12], costab[2]);
+  b2[0x1C] = REAL_MUL(b1[0x1C] - b1[0x13], costab[3]);
+
+  b2[0x14] = b1[0x14] + b1[0x1B];
+  b2[0x15] = b1[0x15] + b1[0x1A];
+  b2[0x1B] = REAL_MUL(b1[0x1B] - b1[0x14], costab[4]);
+  b2[0x1A] = REAL_MUL(b1[0x1A] - b1[0x15], costab[5]);
+
+  b2[0x16] = b1[0x16] + b1[0x19];
+  b2[0x17] = b1[0x17] + b1[0x18];
+  b2[0x19] = REAL_MUL(b1[0x19] - b1[0x16], costab[6]);
+  b2[0x18] = REAL_MUL(b1[0x18] - b1[0x17], costab[7]);
+ }
+
+ {
+  register real *costab = pnts[2];
+
+  b1[0x00] = b2[0x00] + b2[0x07];
+  b1[0x07] = REAL_MUL(b2[0x00] - b2[0x07], costab[0]);
+  b1[0x01] = b2[0x01] + b2[0x06];
+  b1[0x06] = REAL_MUL(b2[0x01] - b2[0x06], costab[1]);
+  b1[0x02] = b2[0x02] + b2[0x05];
+  b1[0x05] = REAL_MUL(b2[0x02] - b2[0x05], costab[2]);
+  b1[0x03] = b2[0x03] + b2[0x04];
+  b1[0x04] = REAL_MUL(b2[0x03] - b2[0x04], costab[3]);
+
+  b1[0x08] = b2[0x08] + b2[0x0F];
+  b1[0x0F] = REAL_MUL(b2[0x0F] - b2[0x08], costab[0]);
+  b1[0x09] = b2[0x09] + b2[0x0E];
+  b1[0x0E] = REAL_MUL(b2[0x0E] - b2[0x09], costab[1]);
+  b1[0x0A] = b2[0x0A] + b2[0x0D];
+  b1[0x0D] = REAL_MUL(b2[0x0D] - b2[0x0A], costab[2]);
+  b1[0x0B] = b2[0x0B] + b2[0x0C];
+  b1[0x0C] = REAL_MUL(b2[0x0C] - b2[0x0B], costab[3]);
+
+  b1[0x10] = b2[0x10] + b2[0x17];
+  b1[0x17] = REAL_MUL(b2[0x10] - b2[0x17], costab[0]);
+  b1[0x11] = b2[0x11] + b2[0x16];
+  b1[0x16] = REAL_MUL(b2[0x11] - b2[0x16], costab[1]);
+  b1[0x12] = b2[0x12] + b2[0x15];
+  b1[0x15] = REAL_MUL(b2[0x12] - b2[0x15], costab[2]);
+  b1[0x13] = b2[0x13] + b2[0x14];
+  b1[0x14] = REAL_MUL(b2[0x13] - b2[0x14], costab[3]);
+
+  b1[0x18] = b2[0x18] + b2[0x1F];
+  b1[0x1F] = REAL_MUL(b2[0x1F] - b2[0x18], costab[0]);
+  b1[0x19] = b2[0x19] + b2[0x1E];
+  b1[0x1E] = REAL_MUL(b2[0x1E] - b2[0x19], costab[1]);
+  b1[0x1A] = b2[0x1A] + b2[0x1D];
+  b1[0x1D] = REAL_MUL(b2[0x1D] - b2[0x1A], costab[2]);
+  b1[0x1B] = b2[0x1B] + b2[0x1C];
+  b1[0x1C] = REAL_MUL(b2[0x1C] - b2[0x1B], costab[3]);
+ }
+
+ {
+  register real const cos0 = pnts[3][0];
+  register real const cos1 = pnts[3][1];
+
+  b2[0x00] = b1[0x00] + b1[0x03];
+  b2[0x03] = REAL_MUL(b1[0x00] - b1[0x03], cos0);
+  b2[0x01] = b1[0x01] + b1[0x02];
+  b2[0x02] = REAL_MUL(b1[0x01] - b1[0x02], cos1);
+
+  b2[0x04] = b1[0x04] + b1[0x07];
+  b2[0x07] = REAL_MUL(b1[0x07] - b1[0x04], cos0);
+  b2[0x05] = b1[0x05] + b1[0x06];
+  b2[0x06] = REAL_MUL(b1[0x06] - b1[0x05], cos1);
+
+  b2[0x08] = b1[0x08] + b1[0x0B];
+  b2[0x0B] = REAL_MUL(b1[0x08] - b1[0x0B], cos0);
+  b2[0x09] = b1[0x09] + b1[0x0A];
+  b2[0x0A] = REAL_MUL(b1[0x09] - b1[0x0A], cos1);
+  
+  b2[0x0C] = b1[0x0C] + b1[0x0F];
+  b2[0x0F] = REAL_MUL(b1[0x0F] - b1[0x0C], cos0);
+  b2[0x0D] = b1[0x0D] + b1[0x0E];
+  b2[0x0E] = REAL_MUL(b1[0x0E] - b1[0x0D], cos1);
+
+  b2[0x10] = b1[0x10] + b1[0x13];
+  b2[0x13] = REAL_MUL(b1[0x10] - b1[0x13], cos0);
+  b2[0x11] = b1[0x11] + b1[0x12];
+  b2[0x12] = REAL_MUL(b1[0x11] - b1[0x12], cos1);
+
+  b2[0x14] = b1[0x14] + b1[0x17];
+  b2[0x17] = REAL_MUL(b1[0x17] - b1[0x14], cos0);
+  b2[0x15] = b1[0x15] + b1[0x16];
+  b2[0x16] = REAL_MUL(b1[0x16] - b1[0x15], cos1);
+
+  b2[0x18] = b1[0x18] + b1[0x1B];
+  b2[0x1B] = REAL_MUL(b1[0x18] - b1[0x1B], cos0);
+  b2[0x19] = b1[0x19] + b1[0x1A];
+  b2[0x1A] = REAL_MUL(b1[0x19] - b1[0x1A], cos1);
+
+  b2[0x1C] = b1[0x1C] + b1[0x1F];
+  b2[0x1F] = REAL_MUL(b1[0x1F] - b1[0x1C], cos0);
+  b2[0x1D] = b1[0x1D] + b1[0x1E];
+  b2[0x1E] = REAL_MUL(b1[0x1E] - b1[0x1D], cos1);
+ }
+
+ {
+  register real const cos0 = pnts[4][0];
+
+  b1[0x00] = b2[0x00] + b2[0x01];
+  b1[0x01] = REAL_MUL(b2[0x00] - b2[0x01], cos0);
+  b1[0x02] = b2[0x02] + b2[0x03];
+  b1[0x03] = REAL_MUL(b2[0x03] - b2[0x02], cos0);
+  b1[0x02] += b1[0x03];
+
+  b1[0x04] = b2[0x04] + b2[0x05];
+  b1[0x05] = REAL_MUL(b2[0x04] - b2[0x05], cos0);
+  b1[0x06] = b2[0x06] + b2[0x07];
+  b1[0x07] = REAL_MUL(b2[0x07] - b2[0x06], cos0);
+  b1[0x06] += b1[0x07];
+  b1[0x04] += b1[0x06];
+  b1[0x06] += b1[0x05];
+  b1[0x05] += b1[0x07];
+
+  b1[0x08] = b2[0x08] + b2[0x09];
+  b1[0x09] = REAL_MUL(b2[0x08] - b2[0x09], cos0);
+  b1[0x0A] = b2[0x0A] + b2[0x0B];
+  b1[0x0B] = REAL_MUL(b2[0x0B] - b2[0x0A], cos0);
+  b1[0x0A] += b1[0x0B];
+
+  b1[0x0C] = b2[0x0C] + b2[0x0D];
+  b1[0x0D] = REAL_MUL(b2[0x0C] - b2[0x0D], cos0);
+  b1[0x0E] = b2[0x0E] + b2[0x0F];
+  b1[0x0F] = REAL_MUL(b2[0x0F] - b2[0x0E], cos0);
+  b1[0x0E] += b1[0x0F];
+  b1[0x0C] += b1[0x0E];
+  b1[0x0E] += b1[0x0D];
+  b1[0x0D] += b1[0x0F];
+
+  b1[0x10] = b2[0x10] + b2[0x11];
+  b1[0x11] = REAL_MUL(b2[0x10] - b2[0x11], cos0);
+  b1[0x12] = b2[0x12] + b2[0x13];
+  b1[0x13] = REAL_MUL(b2[0x13] - b2[0x12], cos0);
+  b1[0x12] += b1[0x13];
+
+  b1[0x14] = b2[0x14] + b2[0x15];
+  b1[0x15] = REAL_MUL(b2[0x14] - b2[0x15], cos0);
+  b1[0x16] = b2[0x16] + b2[0x17];
+  b1[0x17] = REAL_MUL(b2[0x17] - b2[0x16], cos0);
+  b1[0x16] += b1[0x17];
+  b1[0x14] += b1[0x16];
+  b1[0x16] += b1[0x15];
+  b1[0x15] += b1[0x17];
+
+  b1[0x18] = b2[0x18] + b2[0x19];
+  b1[0x19] = REAL_MUL(b2[0x18] - b2[0x19], cos0);
+  b1[0x1A] = b2[0x1A] + b2[0x1B];
+  b1[0x1B] = REAL_MUL(b2[0x1B] - b2[0x1A], cos0);
+  b1[0x1A] += b1[0x1B];
+
+  b1[0x1C] = b2[0x1C] + b2[0x1D];
+  b1[0x1D] = REAL_MUL(b2[0x1C] - b2[0x1D], cos0);
+  b1[0x1E] = b2[0x1E] + b2[0x1F];
+  b1[0x1F] = REAL_MUL(b2[0x1F] - b2[0x1E], cos0);
+  b1[0x1E] += b1[0x1F];
+  b1[0x1C] += b1[0x1E];
+  b1[0x1E] += b1[0x1D];
+  b1[0x1D] += b1[0x1F];
+ }
+
+ out0[0x10*16] = REAL_SCALE_DCT64(b1[0x00]);
+ out0[0x10*12] = REAL_SCALE_DCT64(b1[0x04]);
+ out0[0x10* 8] = REAL_SCALE_DCT64(b1[0x02]);
+ out0[0x10* 4] = REAL_SCALE_DCT64(b1[0x06]);
+ out0[0x10* 0] = REAL_SCALE_DCT64(b1[0x01]);
+ out1[0x10* 0] = REAL_SCALE_DCT64(b1[0x01]);
+ out1[0x10* 4] = REAL_SCALE_DCT64(b1[0x05]);
+ out1[0x10* 8] = REAL_SCALE_DCT64(b1[0x03]);
+ out1[0x10*12] = REAL_SCALE_DCT64(b1[0x07]);
+
+#if 1
+ out0[0x10*14] = REAL_SCALE_DCT64(b1[0x08] + b1[0x0C]);
+ out0[0x10*10] = REAL_SCALE_DCT64(b1[0x0C] + b1[0x0a]);
+ out0[0x10* 6] = REAL_SCALE_DCT64(b1[0x0A] + b1[0x0E]);
+ out0[0x10* 2] = REAL_SCALE_DCT64(b1[0x0E] + b1[0x09]);
+ out1[0x10* 2] = REAL_SCALE_DCT64(b1[0x09] + b1[0x0D]);
+ out1[0x10* 6] = REAL_SCALE_DCT64(b1[0x0D] + b1[0x0B]);
+ out1[0x10*10] = REAL_SCALE_DCT64(b1[0x0B] + b1[0x0F]);
+ out1[0x10*14] = REAL_SCALE_DCT64(b1[0x0F]);
+#else
+ b1[0x08] += b1[0x0C];
+ out0[0x10*14] = REAL_SCALE_DCT64(b1[0x08]);
+ b1[0x0C] += b1[0x0a];
+ out0[0x10*10] = REAL_SCALE_DCT64(b1[0x0C]);
+ b1[0x0A] += b1[0x0E];
+ out0[0x10* 6] = REAL_SCALE_DCT64(b1[0x0A]);
+ b1[0x0E] += b1[0x09];
+ out0[0x10* 2] = REAL_SCALE_DCT64(b1[0x0E]);
+ b1[0x09] += b1[0x0D];
+ out1[0x10* 2] = REAL_SCALE_DCT64(b1[0x09]);
+ b1[0x0D] += b1[0x0B];
+ out1[0x10* 6] = REAL_SCALE_DCT64(b1[0x0D]);
+ b1[0x0B] += b1[0x0F];
+ out1[0x10*10] = REAL_SCALE_DCT64(b1[0x0B]);
+ out1[0x10*14] = REAL_SCALE_DCT64(b1[0x0F]);
+#endif
+
+ { 
+ real tmp;
+ tmp = b1[0x18] + b1[0x1C];
+ out0[0x10*15] = REAL_SCALE_DCT64(tmp + b1[0x10]);
+ out0[0x10*13] = REAL_SCALE_DCT64(tmp + b1[0x14]);
+ tmp = b1[0x1C] + b1[0x1A];
+ out0[0x10*11] = REAL_SCALE_DCT64(tmp + b1[0x14]);
+ out0[0x10* 9] = REAL_SCALE_DCT64(tmp + b1[0x12]);
+ tmp = b1[0x1A] + b1[0x1E];
+ out0[0x10* 7] = REAL_SCALE_DCT64(tmp + b1[0x12]);
+ out0[0x10* 5] = REAL_SCALE_DCT64(tmp + b1[0x16]);
+ tmp = b1[0x1E] + b1[0x19];
+ out0[0x10* 3] = REAL_SCALE_DCT64(tmp + b1[0x16]);
+ out0[0x10* 1] = REAL_SCALE_DCT64(tmp + b1[0x11]);
+ tmp = b1[0x19] + b1[0x1D];
+ out1[0x10* 1] = REAL_SCALE_DCT64(tmp + b1[0x11]);
+ out1[0x10* 3] = REAL_SCALE_DCT64(tmp + b1[0x15]); 
+ tmp = b1[0x1D] + b1[0x1B];
+ out1[0x10* 5] = REAL_SCALE_DCT64(tmp + b1[0x15]);
+ out1[0x10* 7] = REAL_SCALE_DCT64(tmp + b1[0x13]);
+ tmp = b1[0x1B] + b1[0x1F];
+ out1[0x10* 9] = REAL_SCALE_DCT64(tmp + b1[0x13]);
+ out1[0x10*11] = REAL_SCALE_DCT64(tmp + b1[0x17]);
+ out1[0x10*13] = REAL_SCALE_DCT64(b1[0x17] + b1[0x1F]);
+ out1[0x10*15] = REAL_SCALE_DCT64(b1[0x1F]);
+ }
+}
+
+/*
+ * the call via dct64 is a trick to force GCC to use
+ * (new) registers for the b1,b2 pointer to the bufs[xx] field
+ */
+void dct64_i386(real *a,real *b,real *c)
+{
+  real bufs[0x40];
+  dct64_1(a,b,bufs,bufs+0x20,c);
+}
+