Fixed-point FFT and MDCT

libavcodec/Makefile

@@ -31,15 +31,16 @@ OBJS-$(CONFIG_ENCODERS)                += faandct.o jfdctfst.o jfdctint.o
 OBJS-$(CONFIG_DCT)                     += dct.o
 OBJS-$(CONFIG_DWT)                     += dwt.o
 OBJS-$(CONFIG_DXVA2)                   += dxva2.o
-FFT-OBJS-$(CONFIG_HARDCODED_TABLES)    += cos_tables.o
-OBJS-$(CONFIG_FFT)                     += avfft.o fft.o $(FFT-OBJS-yes)
+FFT-OBJS-$(CONFIG_HARDCODED_TABLES)    += cos_tables.o cos_fixed_tables.o
+OBJS-$(CONFIG_FFT)                     += avfft.o fft_fixed.o fft_float.o \
+                                          $(FFT-OBJS-yes)
 OBJS-$(CONFIG_GOLOMB)                  += golomb.o
 OBJS-$(CONFIG_H264DSP)                 += h264dsp.o h264idct.o
 OBJS-$(CONFIG_H264PRED)                += h264pred.o
 OBJS-$(CONFIG_HUFFMAN)                 += huffman.o
 OBJS-$(CONFIG_LPC)                     += lpc.o
 OBJS-$(CONFIG_LSP)                     += lsp.o
-OBJS-$(CONFIG_MDCT)                    += mdct.o
+OBJS-$(CONFIG_MDCT)                    += mdct_fixed.o mdct_float.o
 RDFT-OBJS-$(CONFIG_HARDCODED_TABLES)   += sin_tables.o
 OBJS-$(CONFIG_RDFT)                    += rdft.o $(RDFT-OBJS-yes)
 OBJS-$(CONFIG_SINEWIN)                 += sinewin.o
@@ -672,6 +673,9 @@ $(SUBDIR)dct-test$(EXESUF): $(SUBDIR)dctref.o
 $(SUBDIR)cos_tables.c: $(SUBDIR)costablegen$(HOSTEXESUF)
 	$(M)./$< > $@
 
+$(SUBDIR)cos_fixed_tables.c: $(SUBDIR)costablegen$(HOSTEXESUF)
+	$(M)./$< cos fixed > $@
+
 $(SUBDIR)sin_tables.c: $(SUBDIR)costablegen$(HOSTEXESUF)
 	$(M)./$< sin > $@
 

libavcodec/costablegen.c

@@ -29,14 +29,33 @@
 #endif
 #define BITS 16
 #define FLOATFMT "%.18e"
+#define FIXEDFMT "%6d"
+
+static int clip_f15(int v)
+{
+    return v < -32767 ? -32767 :
+           v >  32767 ?  32767 :
+           v;
+}
+
+static void printval(double val, int fixed)
+{
+    if (fixed)
+        printf(" "FIXEDFMT",", clip_f15(lrint(val * (double)(1<<15))));
+    else
+        printf(" "FLOATFMT",", val);
+
+}
 
 int main(int argc, char *argv[])
 {
     int i, j;
-    int do_sin = argc == 2 && !strcmp(argv[1], "sin");
+    int do_sin = argc > 1 && !strcmp(argv[1], "sin");
+    int fixed  = argc > 2 && !strcmp(argv[2], "fixed");
     double (*func)(double) = do_sin ? sin : cos;
 
     printf("/* This file was generated by libavcodec/costablegen */\n");
+    printf("#define CONFIG_FFT_FLOAT %d\n", !fixed);
     printf("#include \"libavcodec/%s\"\n", do_sin ? "rdft.h" : "fft.h");
     for (i = 4; i <= BITS; i++) {
         int m = 1 << i;
@@ -46,11 +65,12 @@ int main(int argc, char *argv[])
             int idx = j > m/4 ? m/2 - j : j;
             if (do_sin && j >= m/4)
                 idx = m/4 - j;
-            printf(" "FLOATFMT",", func(idx*freq));
+            printval(func(idx*freq), fixed);
             if ((j & 3) == 3)
                 printf("\n   ");
         }
-        printf(" "FLOATFMT"\n};\n", func(do_sin ? -(m/4 - 1)*freq : freq));
+        printval(func(do_sin ? -(m/4 - 1)*freq : freq), fixed);
+        printf("\n};\n");
     }
     return 0;
 }

libavcodec/fft-internal.h

@@ -0,0 +1,67 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVCODEC_FFT_INTERNAL_H
+#define AVCODEC_FFT_INTERNAL_H
+
+#if CONFIG_FFT_FLOAT
+
+#define FIX15(v) (v)
+#define sqrthalf (float)M_SQRT1_2
+
+#define BF(x, y, a, b) do {                     \
+        x = a - b;                              \
+        y = a + b;                              \
+    } while (0)
+
+#define CMUL(dre, dim, are, aim, bre, bim) do { \
+        (dre) = (are) * (bre) - (aim) * (bim);  \
+        (dim) = (are) * (bim) + (aim) * (bre);  \
+    } while (0)
+
+#else
+
+#include "libavutil/intmath.h"
+#include "mathops.h"
+
+#define SCALE_FLOAT(a, bits) lrint((a) * (double)(1 << (bits)))
+#define FIX15(a) av_clip(SCALE_FLOAT(a, 15), -32767, 32767)
+
+#define sqrthalf ((int16_t)((1<<15)*M_SQRT1_2))
+
+#define BF(x, y, a, b) do {                     \
+        x = (a - b) >> 1;                       \
+        y = (a + b) >> 1;                       \
+    } while (0)
+
+#define CMUL(dre, dim, are, aim, bre, bim) do {                 \
+        (dre) = (MUL16(are, bre) - MUL16(aim, bim)) >> 15;      \
+        (dim) = (MUL16(are, bim) + MUL16(aim, bre)) >> 15;      \
+    } while (0)
+
+#endif /* CONFIG_FFT_FLOAT */
+
+#define ff_imdct_calc_c FFT_NAME(ff_imdct_calc_c)
+#define ff_imdct_half_c FFT_NAME(ff_imdct_half_c)
+#define ff_mdct_calc_c  FFT_NAME(ff_mdct_calc_c)
+
+void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
+void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input);
+void ff_mdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
+
+#endif /* AVCODEC_FFT_INTERNAL_H */

libavcodec/fft.c

@@ -30,6 +30,7 @@
 #include <string.h>
 #include "libavutil/mathematics.h"
 #include "fft.h"
+#include "fft-internal.h"
 
 /* cos(2*pi*x/n) for 0<=x<=n/4, followed by its reverse */
 #if !CONFIG_HARDCODED_TABLES
@@ -47,10 +48,21 @@ COSTABLE(16384);
 COSTABLE(32768);
 COSTABLE(65536);
 #endif
-COSTABLE_CONST FFTSample * const ff_cos_tabs[] = {
+COSTABLE_CONST FFTSample * const FFT_NAME(ff_cos_tabs)[] = {
     NULL, NULL, NULL, NULL,
-    ff_cos_16, ff_cos_32, ff_cos_64, ff_cos_128, ff_cos_256, ff_cos_512, ff_cos_1024,
-    ff_cos_2048, ff_cos_4096, ff_cos_8192, ff_cos_16384, ff_cos_32768, ff_cos_65536,
+    FFT_NAME(ff_cos_16),
+    FFT_NAME(ff_cos_32),
+    FFT_NAME(ff_cos_64),
+    FFT_NAME(ff_cos_128),
+    FFT_NAME(ff_cos_256),
+    FFT_NAME(ff_cos_512),
+    FFT_NAME(ff_cos_1024),
+    FFT_NAME(ff_cos_2048),
+    FFT_NAME(ff_cos_4096),
+    FFT_NAME(ff_cos_8192),
+    FFT_NAME(ff_cos_16384),
+    FFT_NAME(ff_cos_32768),
+    FFT_NAME(ff_cos_65536),
 };
 
 static void ff_fft_permute_c(FFTContext *s, FFTComplex *z);
@@ -73,9 +85,9 @@ av_cold void ff_init_ff_cos_tabs(int index)
     int i;
     int m = 1<<index;
     double freq = 2*M_PI/m;
-    FFTSample *tab = ff_cos_tabs[index];
+    FFTSample *tab = FFT_NAME(ff_cos_tabs)[index];
     for(i=0; i<=m/4; i++)
-        tab[i] = cos(i*freq);
+        tab[i] = FIX15(cos(i*freq));
     for(i=1; i<m/4; i++)
         tab[m/2-i] = tab[i];
 #endif
@@ -107,9 +119,11 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
     s->mdct_calc   = ff_mdct_calc_c;
 #endif
 
+#if CONFIG_FFT_FLOAT
     if (ARCH_ARM)     ff_fft_init_arm(s);
     if (HAVE_ALTIVEC) ff_fft_init_altivec(s);
     if (HAVE_MMX)     ff_fft_init_mmx(s);
+#endif
 
     for(j=4; j<=nbits; j++) {
         ff_init_ff_cos_tabs(j);
@@ -144,13 +158,6 @@ av_cold void ff_fft_end(FFTContext *s)
     av_freep(&s->tmp_buf);
 }
 
-#define sqrthalf (float)M_SQRT1_2
-
-#define BF(x,y,a,b) {\
-    x = a - b;\
-    y = a + b;\
-}
-
 #define BUTTERFLIES(a0,a1,a2,a3) {\
     BF(t3, t5, t5, t1);\
     BF(a2.re, a0.re, a0.re, t5);\
@@ -174,10 +181,8 @@ av_cold void ff_fft_end(FFTContext *s)
 }
 
 #define TRANSFORM(a0,a1,a2,a3,wre,wim) {\
-    t1 = a2.re * wre + a2.im * wim;\
-    t2 = a2.im * wre - a2.re * wim;\
-    t5 = a3.re * wre - a3.im * wim;\
-    t6 = a3.im * wre + a3.re * wim;\
+    CMUL(t1, t2, a2.re, a2.im, wre, -wim);\
+    CMUL(t5, t6, a3.re, a3.im, wre,  wim);\
     BUTTERFLIES(a0,a1,a2,a3)\
 }
 
@@ -193,7 +198,7 @@ av_cold void ff_fft_end(FFTContext *s)
 #define PASS(name)\
 static void name(FFTComplex *z, const FFTSample *wre, unsigned int n)\
 {\
-    FFTSample t1, t2, t3, t4, t5, t6;\
+    FFTDouble t1, t2, t3, t4, t5, t6;\
     int o1 = 2*n;\
     int o2 = 4*n;\
     int o3 = 6*n;\
@@ -222,12 +227,12 @@ static void fft##n(FFTComplex *z)\
     fft##n2(z);\
     fft##n4(z+n4*2);\
     fft##n4(z+n4*3);\
-    pass(z,ff_cos_##n,n4/2);\
+    pass(z,FFT_NAME(ff_cos_##n),n4/2);\
 }
 
 static void fft4(FFTComplex *z)
 {
-    FFTSample t1, t2, t3, t4, t5, t6, t7, t8;
+    FFTDouble t1, t2, t3, t4, t5, t6, t7, t8;
 
     BF(t3, t1, z[0].re, z[1].re);
     BF(t8, t6, z[3].re, z[2].re);
@@ -241,7 +246,7 @@ static void fft4(FFTComplex *z)
 
 static void fft8(FFTComplex *z)
 {
-    FFTSample t1, t2, t3, t4, t5, t6, t7, t8;
+    FFTDouble t1, t2, t3, t4, t5, t6, t7, t8;
 
     fft4(z);
 
@@ -262,7 +267,9 @@ static void fft8(FFTComplex *z)
 #if !CONFIG_SMALL
 static void fft16(FFTComplex *z)
 {
-    FFTSample t1, t2, t3, t4, t5, t6;
+    FFTDouble t1, t2, t3, t4, t5, t6;
+    FFTSample cos_16_1 = FFT_NAME(ff_cos_16)[1];
+    FFTSample cos_16_3 = FFT_NAME(ff_cos_16)[3];
 
     fft8(z);
     fft4(z+8);
@@ -270,8 +277,8 @@ static void fft16(FFTComplex *z)
 
     TRANSFORM_ZERO(z[0],z[4],z[8],z[12]);
     TRANSFORM(z[2],z[6],z[10],z[14],sqrthalf,sqrthalf);
-    TRANSFORM(z[1],z[5],z[9],z[13],ff_cos_16[1],ff_cos_16[3]);
-    TRANSFORM(z[3],z[7],z[11],z[15],ff_cos_16[3],ff_cos_16[1]);
+    TRANSFORM(z[1],z[5],z[9],z[13],cos_16_1,cos_16_3);
+    TRANSFORM(z[3],z[7],z[11],z[15],cos_16_3,cos_16_1);
 }
 #else
 DECL_FFT(16,8,4)

libavcodec/fft.h

@@ -22,11 +22,37 @@
 #ifndef AVCODEC_FFT_H
 #define AVCODEC_FFT_H
 
+#ifndef CONFIG_FFT_FLOAT
+#define CONFIG_FFT_FLOAT 1
+#endif
+
 #include <stdint.h>
 #include "config.h"
 #include "libavutil/mem.h"
+
+#if CONFIG_FFT_FLOAT
+
 #include "avfft.h"
 
+#define FFT_NAME(x) x
+
+typedef float FFTDouble;
+
+#else
+
+#define FFT_NAME(x) x ## _fixed
+
+typedef int16_t FFTSample;
+typedef int     FFTDouble;
+
+typedef struct FFTComplex {
+    int16_t re, im;
+} FFTComplex;
+
+typedef struct FFTContext FFTContext;
+
+#endif /* CONFIG_FFT_FLOAT */
+
 /* FFT computation */
 
 struct FFTContext {
@@ -66,7 +92,7 @@ struct FFTContext {
 #endif
 
 #define COSTABLE(size) \
-    COSTABLE_CONST DECLARE_ALIGNED(16, FFTSample, ff_cos_##size)[size/2]
+    COSTABLE_CONST DECLARE_ALIGNED(16, FFTSample, FFT_NAME(ff_cos_##size))[size/2]
 
 extern COSTABLE(16);
 extern COSTABLE(32);
@@ -81,7 +107,9 @@ extern COSTABLE(8192);
 extern COSTABLE(16384);
 extern COSTABLE(32768);
 extern COSTABLE(65536);
-extern COSTABLE_CONST FFTSample* const ff_cos_tabs[17];
+extern COSTABLE_CONST FFTSample* const FFT_NAME(ff_cos_tabs)[17];
+
+#define ff_init_ff_cos_tabs FFT_NAME(ff_init_ff_cos_tabs)
 
 /**
  * Initialize the cosine table in ff_cos_tabs[index]
@@ -89,6 +117,9 @@ extern COSTABLE_CONST FFTSample* const ff_cos_tabs[17];
  */
 void ff_init_ff_cos_tabs(int index);
 
+#define ff_fft_init FFT_NAME(ff_fft_init)
+#define ff_fft_end  FFT_NAME(ff_fft_end)
+
 /**
  * Set up a complex FFT.
  * @param nbits           log2 of the length of the input array
@@ -102,10 +133,10 @@ void ff_fft_init_arm(FFTContext *s);
 
 void ff_fft_end(FFTContext *s);
 
+#define ff_mdct_init FFT_NAME(ff_mdct_init)
+#define ff_mdct_end  FFT_NAME(ff_mdct_end)
+
 int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale);
-void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
-void ff_imdct_half_c(FFTContext *s, FFTSample *output, const FFTSample *input);
-void ff_mdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input);
 void ff_mdct_end(FFTContext *s);
 
 #endif /* AVCODEC_FFT_H */

libavcodec/fft_fixed.c

@@ -0,0 +1,20 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define CONFIG_FFT_FLOAT 0
+#include "fft.c"

libavcodec/fft_float.c

@@ -0,0 +1,20 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define CONFIG_FFT_FLOAT 1
+#include "fft.c"

libavcodec/mdct.c

@@ -24,12 +24,19 @@
 #include "libavutil/common.h"
 #include "libavutil/mathematics.h"
 #include "fft.h"
+#include "fft-internal.h"
 
 /**
  * @file
  * MDCT/IMDCT transforms.
  */
 
+#if CONFIG_FFT_FLOAT
+#   define RSCALE(x) (x)
+#else
+#   define RSCALE(x) ((x) >> 1)
+#endif
+
 /**
  * init MDCT or IMDCT computation.
  */
@@ -70,8 +77,8 @@ av_cold int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale)
     scale = sqrt(fabs(scale));
     for(i=0;i<n4;i++) {
         alpha = 2 * M_PI * (i + theta) / n;
-        s->tcos[i*tstep] = -cos(alpha) * scale;
-        s->tsin[i*tstep] = -sin(alpha) * scale;
+        s->tcos[i*tstep] = FIX15(-cos(alpha) * scale);
+        s->tsin[i*tstep] = FIX15(-sin(alpha) * scale);
     }
     return 0;
  fail:
@@ -79,17 +86,6 @@ av_cold int ff_mdct_init(FFTContext *s, int nbits, int inverse, double scale)
     return -1;
 }
 
-/* complex multiplication: p = a * b */
-#define CMUL(pre, pim, are, aim, bre, bim) \
-{\
-    FFTSample _are = (are);\
-    FFTSample _aim = (aim);\
-    FFTSample _bre = (bre);\
-    FFTSample _bim = (bim);\
-    (pre) = _are * _bre - _aim * _bim;\
-    (pim) = _are * _bim + _aim * _bre;\
-}
-
 /**
  * Compute the middle half of the inverse MDCT of size N = 2^nbits,
  * thus excluding the parts that can be derived by symmetry
@@ -161,7 +157,7 @@ void ff_imdct_calc_c(FFTContext *s, FFTSample *output, const FFTSample *input)
 void ff_mdct_calc_c(FFTContext *s, FFTSample *out, const FFTSample *input)
 {
     int i, j, n, n8, n4, n2, n3;
-    FFTSample re, im;
+    FFTDouble re, im;
     const uint16_t *revtab = s->revtab;
     const FFTSample *tcos = s->tcos;
     const FFTSample *tsin = s->tsin;
@@ -175,13 +171,13 @@ void ff_mdct_calc_c(FFTContext *s, FFTSample *out, const FFTSample *input)
 
     /* pre rotation */
     for(i=0;i<n8;i++) {
-        re = -input[2*i+n3] - input[n3-1-2*i];
-        im = -input[n4+2*i] + input[n4-1-2*i];
+        re = RSCALE(-input[2*i+n3] - input[n3-1-2*i]);
+        im = RSCALE(-input[n4+2*i] + input[n4-1-2*i]);
         j = revtab[i];
         CMUL(x[j].re, x[j].im, re, im, -tcos[i], tsin[i]);
 
-        re = input[2*i] - input[n2-1-2*i];
-        im = -(input[n2+2*i] + input[n-1-2*i]);
+        re = RSCALE( input[2*i]    - input[n2-1-2*i]);
+        im = RSCALE(-input[n2+2*i] - input[ n-1-2*i]);
         j = revtab[n8 + i];
         CMUL(x[j].re, x[j].im, re, im, -tcos[n8 + i], tsin[n8 + i]);
     }

libavcodec/mdct_fixed.c

@@ -0,0 +1,20 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define CONFIG_FFT_FLOAT 0
+#include "mdct.c"

libavcodec/mdct_float.c

@@ -0,0 +1,20 @@
+/*
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * Libav is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#define CONFIG_FFT_FLOAT 1
+#include "mdct.c"