*/
/**
- * @file faandct.c
+ * @file
* @brief
* Floating point AAN DCT
* @author Michael Niedermayer <michaelni@gmx.at>
*/
-#include "dsputil.h"
#include "faandct.h"
+#include "libavutil/internal.h"
+#include "libavutil/libm.h"
-#define FLOAT float
-#ifdef FAAN_POSTSCALE
-# define SCALE(x) postscale[x]
-#else
-# define SCALE(x) 1
-#endif
+typedef float FLOAT;
-//numbers generated by simple c code (not as accurate as they could be)
-/*
-for(i=0; i<8; i++){
- printf("#define B%d %1.20llf\n", i, (long double)1.0/(cosl(i*acosl(-1.0)/(long double)16.0)*sqrtl(2)));
-}
+/* numbers generated by arbitrary precision arithmetic followed by truncation
+to 36 fractional digits (enough for a 128-bit IEEE quad, see /usr/include/math.h
+for this approach). Unfortunately, long double is not always available correctly,
+e.g ppc has issues.
+TODO: add L suffixes when ppc and toolchains sort out their stuff.
*/
-#define B0 1.00000000000000000000
-#define B1 0.72095982200694791383 // (cos(pi*1/16)sqrt(2))^-1
-#define B2 0.76536686473017954350 // (cos(pi*2/16)sqrt(2))^-1
-#define B3 0.85043009476725644878 // (cos(pi*3/16)sqrt(2))^-1
-#define B4 1.00000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
-#define B5 1.27275858057283393842 // (cos(pi*5/16)sqrt(2))^-1
-#define B6 1.84775906502257351242 // (cos(pi*6/16)sqrt(2))^-1
-#define B7 3.62450978541155137218 // (cos(pi*7/16)sqrt(2))^-1
-
-
-#define A1 0.70710678118654752438 // cos(pi*4/16)
+#define B0 1.000000000000000000000000000000000000
+#define B1 0.720959822006947913789091890943021267 // (cos(pi*1/16)sqrt(2))^-1
+#define B2 0.765366864730179543456919968060797734 // (cos(pi*2/16)sqrt(2))^-1
+#define B3 0.850430094767256448766702844371412325 // (cos(pi*3/16)sqrt(2))^-1
+#define B4 1.000000000000000000000000000000000000 // (cos(pi*4/16)sqrt(2))^-1
+#define B5 1.272758580572833938461007018281767032 // (cos(pi*5/16)sqrt(2))^-1
+#define B6 1.847759065022573512256366378793576574 // (cos(pi*6/16)sqrt(2))^-1
+#define B7 3.624509785411551372409941227504289587 // (cos(pi*7/16)sqrt(2))^-1
+
+#define A1 M_SQRT1_2 // cos(pi*4/16)
#define A2 0.54119610014619698435 // cos(pi*6/16)sqrt(2)
#define A5 0.38268343236508977170 // cos(pi*6/16)
#define A4 1.30656296487637652774 // cos(pi*2/16)sqrt(2)
B7*B0, B7*B1, B7*B2, B7*B3, B7*B4, B7*B5, B7*B6, B7*B7,
};
-static av_always_inline void row_fdct(FLOAT temp[64], DCTELEM * data)
+static av_always_inline void row_fdct(FLOAT temp[64], int16_t *data)
{
FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
FLOAT tmp10, tmp11, tmp12, tmp13;
FLOAT z2, z4, z11, z13;
- FLOAT av_unused z5;
int i;
for (i=0; i<8*8; i+=8) {
tmp6 += tmp7;
#if 0
- z5= (tmp4 - tmp6) * A5;
- z2= tmp4*A2 + z5;
- z4= tmp6*A4 + z5;
+ {
+ FLOAT z5;
+ z5 = (tmp4 - tmp6) * A5;
+ z2 = tmp4 * A2 + z5;
+ z4 = tmp6 * A4 + z5;
+ }
#else
z2= tmp4*(A2+A5) - tmp6*A5;
z4= tmp6*(A4-A5) + tmp4*A5;
}
}
-void ff_faandct(DCTELEM * data)
+void ff_faandct(int16_t *data)
{
FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
FLOAT tmp10, tmp11, tmp12, tmp13;
FLOAT z2, z4, z11, z13;
- FLOAT av_unused z5;
FLOAT temp[64];
int i;
tmp11= tmp1 + tmp2;
tmp12= tmp1 - tmp2;
- data[8*0 + i]= lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
- data[8*4 + i]= lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
+ data[8*0 + i]= lrintf(postscale[8*0 + i] * (tmp10 + tmp11));
+ data[8*4 + i]= lrintf(postscale[8*4 + i] * (tmp10 - tmp11));
tmp12 += tmp13;
tmp12 *= A1;
- data[8*2 + i]= lrintf(SCALE(8*2 + i) * (tmp13 + tmp12));
- data[8*6 + i]= lrintf(SCALE(8*6 + i) * (tmp13 - tmp12));
+ data[8*2 + i]= lrintf(postscale[8*2 + i] * (tmp13 + tmp12));
+ data[8*6 + i]= lrintf(postscale[8*6 + i] * (tmp13 - tmp12));
tmp4 += tmp5;
tmp5 += tmp6;
tmp6 += tmp7;
#if 0
- z5= (tmp4 - tmp6) * A5;
- z2= tmp4*A2 + z5;
- z4= tmp6*A4 + z5;
+ {
+ FLOAT z5;
+ z5 = (tmp4 - tmp6) * A5;
+ z2 = tmp4 * A2 + z5;
+ z4 = tmp6 * A4 + z5;
+ }
#else
z2= tmp4*(A2+A5) - tmp6*A5;
z4= tmp6*(A4-A5) + tmp4*A5;
z11= tmp7 + tmp5;
z13= tmp7 - tmp5;
- data[8*5 + i]= lrintf(SCALE(8*5 + i) * (z13 + z2));
- data[8*3 + i]= lrintf(SCALE(8*3 + i) * (z13 - z2));
- data[8*1 + i]= lrintf(SCALE(8*1 + i) * (z11 + z4));
- data[8*7 + i]= lrintf(SCALE(8*7 + i) * (z11 - z4));
+ data[8*5 + i]= lrintf(postscale[8*5 + i] * (z13 + z2));
+ data[8*3 + i]= lrintf(postscale[8*3 + i] * (z13 - z2));
+ data[8*1 + i]= lrintf(postscale[8*1 + i] * (z11 + z4));
+ data[8*7 + i]= lrintf(postscale[8*7 + i] * (z11 - z4));
}
}
-void ff_faandct248(DCTELEM * data)
+void ff_faandct248(int16_t *data)
{
FLOAT tmp0, tmp1, tmp2, tmp3, tmp4, tmp5, tmp6, tmp7;
FLOAT tmp10, tmp11, tmp12, tmp13;
tmp12 = tmp1 - tmp2;
tmp13 = tmp0 - tmp3;
- data[8*0 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
- data[8*4 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
+ data[8*0 + i] = lrintf(postscale[8*0 + i] * (tmp10 + tmp11));
+ data[8*4 + i] = lrintf(postscale[8*4 + i] * (tmp10 - tmp11));
tmp12 += tmp13;
tmp12 *= A1;
- data[8*2 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + tmp12));
- data[8*6 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - tmp12));
+ data[8*2 + i] = lrintf(postscale[8*2 + i] * (tmp13 + tmp12));
+ data[8*6 + i] = lrintf(postscale[8*6 + i] * (tmp13 - tmp12));
tmp10 = tmp4 + tmp7;
tmp11 = tmp5 + tmp6;
tmp12 = tmp5 - tmp6;
tmp13 = tmp4 - tmp7;
- data[8*1 + i] = lrintf(SCALE(8*0 + i) * (tmp10 + tmp11));
- data[8*5 + i] = lrintf(SCALE(8*4 + i) * (tmp10 - tmp11));
+ data[8*1 + i] = lrintf(postscale[8*0 + i] * (tmp10 + tmp11));
+ data[8*5 + i] = lrintf(postscale[8*4 + i] * (tmp10 - tmp11));
tmp12 += tmp13;
tmp12 *= A1;
- data[8*3 + i] = lrintf(SCALE(8*2 + i) * (tmp13 + tmp12));
- data[8*7 + i] = lrintf(SCALE(8*6 + i) * (tmp13 - tmp12));
+ data[8*3 + i] = lrintf(postscale[8*2 + i] * (tmp13 + tmp12));
+ data[8*7 + i] = lrintf(postscale[8*6 + i] * (tmp13 - tmp12));
}
}