3DNow! & Extended 3DNow! versions of FFT

Patch by Zuxy Meng, zuxy <<dot>> meng >>at<< gmail <<dot>> com
Minor non-functional diff-related fixes by me.

Originally committed as revision 5125 to svn://svn.ffmpeg.org/ffmpeg/trunk

configure

@@ -1110,6 +1110,19 @@ if $cc -msse -o $TMPO $TMPC 2> /dev/null ; then
   builtin_vector=yes
 fi
 
+# test for mm3dnow.h
+cat > $TMPC << EOF
+#include <mm3dnow.h>
+int main(void) {
+return 0;
+}
+EOF
+
+mm3dnow=no
+if $cc -march=athlon -o $TMPO $TMPC 2> /dev/null ; then
+  mm3dnow=yes
+fi
+
 # Probe for -Wdeclaration-after-statement
 if test "$cc" = "gcc"; then
   cat > $TMPC << EOF
@@ -1328,6 +1341,7 @@ echo "broken inttypes.h $emu_fast_int"
 if test $cpu = "x86" -o $cpu = "x86_64"; then
 echo "MMX enabled      $mmx"
 echo "Vector Builtins  $builtin_vector"
+echo "3DNow! Builtins  $mm3dnow"
 fi
 if test $cpu = "armv4l"; then
 echo "IWMMXT enabled   $iwmmxt"
@@ -1508,6 +1522,10 @@ if test "$builtin_vector" = "yes" ; then
   echo "TARGET_BUILTIN_VECTOR=yes" >> config.mak
   echo "#define HAVE_BUILTIN_VECTOR 1" >> $TMPH
 fi
+if test "$mm3dnow" = "yes" ; then
+  echo "TARGET_BUILTIN_3DNOW=yes" >> config.mak
+  echo "#define HAVE_MM3DNOW 1" >> $TMPH
+fi
 if test "$iwmmxt" = "yes" ; then
   echo "TARGET_IWMMXT=yes" >> config.mak
   echo "#define HAVE_IWMMXT 1" >> $TMPH

libavcodec/Makefile

@@ -330,7 +330,7 @@ OBJS += i386/fdct_mmx.o i386/cputest.o \
 	i386/dsputil_mmx.o i386/mpegvideo_mmx.o \
 	i386/idct_mmx.o i386/motion_est_mmx.o \
 	i386/simple_idct_mmx.o i386/fft_sse.o i386/vp3dsp_mmx.o \
-	i386/vp3dsp_sse2.o
+	i386/vp3dsp_sse2.o i386/fft_3dn.o i386/fft_3dn2.o
 ifeq ($(CONFIG_GPL),yes)
 OBJS += i386/idct_mmx_xvid.o
 endif
@@ -338,6 +338,10 @@ ifdef TARGET_BUILTIN_VECTOR
 i386/fft_sse.o: CFLAGS+= -msse
 depend: CFLAGS+= -msse
 endif
+ifdef TARGET_BUILTIN_3DNOW
+i386/fft_3dn.o: CFLAGS+= -m3dnow
+i386/fft_3dn2.o: CFLAGS+= -march=athlon
+endif
 endif
 
 # armv4l specific stuff

libavcodec/dsputil.h

@@ -564,6 +564,8 @@ int ff_fft_init(FFTContext *s, int nbits, int inverse);
 void ff_fft_permute(FFTContext *s, FFTComplex *z);
 void ff_fft_calc_c(FFTContext *s, FFTComplex *z);
 void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
+void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z);
+void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z);
 void ff_fft_calc_altivec(FFTContext *s, FFTComplex *z);
 
 static inline void ff_fft_calc(FFTContext *s, FFTComplex *z)

libavcodec/fft.c

@@ -57,12 +57,12 @@ int ff_fft_init(FFTContext *s, int nbits, int inverse)
     s->exptab1 = NULL;
 
     /* compute constant table for HAVE_SSE version */
-#if (defined(HAVE_MMX) && defined(HAVE_BUILTIN_VECTOR)) || defined(HAVE_ALTIVEC)
+#if (defined(HAVE_MMX) && (defined(HAVE_BUILTIN_VECTOR) || defined(HAVE_MM3DNOW))) || defined(HAVE_ALTIVEC)
     {
         int has_vectors = 0;
 
 #if defined(HAVE_MMX)
-        has_vectors = mm_support() & MM_SSE;
+        has_vectors = mm_support() & (MM_3DNOW | MM_3DNOWEXT | MM_SSE | MM_SSE2);
 #endif
 #if defined(HAVE_ALTIVEC) && !defined(ALTIVEC_USE_REFERENCE_C_CODE)
         has_vectors = mm_support() & MM_ALTIVEC;
@@ -94,8 +94,24 @@ int ff_fft_init(FFTContext *s, int nbits, int inverse)
             } while (nblocks != 0);
             av_freep(&s->exptab);
 #if defined(HAVE_MMX)
-            s->fft_calc = ff_fft_calc_sse;
+#ifdef HAVE_MM3DNOW
-#else
+            if (has_vectors & MM_3DNOWEXT)
+                /* 3DNowEx for Athlon(XP) */
+                s->fft_calc = ff_fft_calc_3dn2;
+            else if (has_vectors & MM_3DNOW)
+                /* 3DNow! for K6-2/3 */
+                s->fft_calc = ff_fft_calc_3dn;
+#endif
+#ifdef HAVE_BUILTIN_VECTOR
+            if (has_vectors & MM_SSE2)
+                /* SSE for P4/K8 */
+                s->fft_calc = ff_fft_calc_sse;
+            else if ((has_vectors & MM_SSE) &&
+                     s->fft_calc == ff_fft_calc_c)
+                /* SSE for P3 */
+                s->fft_calc = ff_fft_calc_sse;
+#endif
+#else /* HAVE_MMX */
             s->fft_calc = ff_fft_calc_altivec;
 #endif
         }

libavcodec/i386/fft_3dn.c

@@ -0,0 +1,136 @@
+/*
+ * FFT/MDCT transform with 3DNow! optimizations
+ * Copyright (c) 2006 Zuxy MENG Jie.
+ * Based on fft_sse.c copyright (c) 2002 Fabrice Bellard.
+ *
+ * This library 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 of the License, or (at your option) any later version.
+ *
+ * This library 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 this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+#include "../dsputil.h"
+#include <math.h>
+
+#ifdef HAVE_MM3DNOW
+
+#include <mm3dnow.h>
+
+static const int p1m1[2] __attribute__((aligned(8))) =
+    { 0, 1 << 31 };
+
+static const int m1p1[2] __attribute__((aligned(8))) =
+    { 1 << 31, 0 };
+
+void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z)
+{
+    int ln = s->nbits;
+    int j, np, np2;
+    int nblocks, nloops;
+    register FFTComplex *p, *q;
+    FFTComplex *cptr, *cptr1;
+    int k;
+
+    np = 1 << ln;
+    /* FEMMS not a must here but recommended by AMD */
+    _m_femms();
+
+    {
+        __m64 *r, a0, a1, b0, b1, tmp, c;
+
+        r = (__m64 *)&z[0];
+        if (s->inverse)
+            c = *(__m64 *)m1p1;
+        else
+            c = *(__m64 *)p1m1;
+
+        j = (np >> 2);
+        do {
+            /* do the pass 0 butterfly */
+            a0 = _m_pfadd(r[0], r[1]);
+            a1 = _m_pfsub(r[0], r[1]);
+
+            /* do the pass 0 butterfly */
+            b0 = _m_pfadd(r[2], r[3]);
+            b1 = _m_pfsub(r[2], r[3]);
+
+            /* multiply third by -i */
+            tmp = _m_punpckhdq(b1, b1);
+            b1 = _m_punpckldq(b1, b1);
+            b1 = _m_punpckldq(tmp, b1);
+            b1 = _m_pxor(b1, c);
+
+            /* do the pass 1 butterfly */
+            r[0] = _m_pfadd(a0, b0);
+            r[1] = _m_pfadd(a1, b1);
+            r[2] = _m_pfsub(a0, b0);
+            r[3] = _m_pfsub(a1, b1);
+            r += 4;
+        } while (--j != 0);
+    }
+    /* pass 2 .. ln-1 */
+
+    nblocks = np >> 3;
+    nloops = 1 << 2;
+    np2 = np >> 1;
+
+    cptr1 = s->exptab1;
+    do {
+        p = z;
+        q = z + nloops;
+        j = nblocks;
+        do {
+            cptr = cptr1;
+            k = nloops >> 1;
+            do {
+                __m64 a0, a1, b0, b1, c0, c1, t10, t11, t20, t21;
+
+                a0 = *(__m64 *)&p[0];
+                a1 = *(__m64 *)&p[1];
+                b0 = *(__m64 *)&q[0];
+                b1 = *(__m64 *)&q[1];
+
+                /* complex mul */
+                c0 = *(__m64 *)&cptr[0];
+                c1 = *(__m64 *)&cptr[1];
+                /*  cre*re cim*re */
+                t10 = _m_pfmul(c0, _m_punpckldq(b0, b0));
+                t11 = _m_pfmul(c1, _m_punpckldq(b1, b1));
+                c0 = *(__m64 *)&cptr[2];
+                c1 = *(__m64 *)&cptr[3];
+                /*  -cim*im cre*im */
+                t20 = _m_pfmul(c0, _m_punpckhdq(b0, b0));
+                t21 = _m_pfmul(c1, _m_punpckhdq(b1, b1));
+                b0 = _m_pfadd(t10, t20);
+                b1 = _m_pfadd(t11, t21);
+
+                /* butterfly */
+                *(__m64 *)&p[0] = _m_pfadd(a0, b0);
+                *(__m64 *)&p[1] = _m_pfadd(a1, b1);
+                *(__m64 *)&q[0] = _m_pfsub(a0, b0);
+                *(__m64 *)&q[1] = _m_pfsub(a1, b1);
+
+                p += 2;
+                q += 2;
+                cptr += 4;
+            } while (--k);
+
+            p += nloops;
+            q += nloops;
+        } while (--j);
+        cptr1 += nloops * 2;
+        nblocks = nblocks >> 1;
+        nloops = nloops << 1;
+    } while (nblocks != 0);
+    _m_femms();
+}
+
+#endif

libavcodec/i386/fft_3dn2.c

@@ -0,0 +1,136 @@
+/*
+ * FFT/MDCT transform with Extended 3DNow! optimizations
+ * Copyright (c) 2006 Zuxy MENG Jie.
+ * Based on fft_sse.c copyright (c) 2002 Fabrice Bellard.
+ *
+ * This library 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 of the License, or (at your option) any later version.
+ *
+ * This library 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 this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+#include "../dsputil.h"
+#include <math.h>
+
+#ifdef HAVE_MM3DNOW
+
+#include <mm3dnow.h>
+
+static const int p1m1[2] __attribute__((aligned(8))) =
+    { 0, 1 << 31 };
+
+static const int m1p1[2] __attribute__((aligned(8))) =
+    { 1 << 31, 0 };
+
+void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z)
+{
+    int ln = s->nbits;
+    int j, np, np2;
+    int nblocks, nloops;
+    register FFTComplex *p, *q;
+    FFTComplex *cptr, *cptr1;
+    int k;
+
+    np = 1 << ln;
+    /* FEMMS is not a must here but recommended by AMD */
+    _m_femms();
+
+    {
+        __m64 *r, a0, a1, b0, b1, c;
+
+        r = (__m64 *)&z[0];
+        if (s->inverse)
+            c = *(__m64 *)m1p1;
+        else
+            c = *(__m64 *)p1m1;
+
+        j = (np >> 2);
+        do {
+            /* do the pass 0 butterfly */
+            a0 = _m_pfadd(r[0], r[1]);
+            a1 = _m_pfsub(r[0], r[1]);
+
+            /* do the pass 0 butterfly */
+            b0 = _m_pfadd(r[2], r[3]);
+            b1 = _m_pfsub(r[2], r[3]);
+
+            /* multiply third by -i */
+            b1 = _m_pswapd(b1);
+            b1 = _m_pxor(b1, c);
+
+            r[0] = _m_pfadd(a0, b0);
+            r[1] = _m_pfadd(a1, b1);
+            r[2] = _m_pfsub(a0, b0);
+            r[3] = _m_pfsub(a1, b1);
+            r += 4;
+        } while (--j != 0);
+    }
+    /* pass 2 .. ln-1 */
+
+    nblocks = np >> 3;
+    nloops = 1 << 2;
+    np2 = np >> 1;
+
+    cptr1 = s->exptab1;
+    do {
+        p = z;
+        q = z + nloops;
+        j = nblocks;
+        do {
+            cptr = cptr1;
+            k = nloops >> 1;
+            do {
+                __m64 a0, a1, b0, b1, c0, c1, t10, t11, t20, t21;
+
+                a0 = *(__m64 *)&p[0];
+                a1 = *(__m64 *)&p[1];
+                b0 = *(__m64 *)&q[0];
+                b1 = *(__m64 *)&q[1];
+
+                /* complex mul */
+                c0 = *(__m64 *)&cptr[0];
+                c1 = *(__m64 *)&cptr[1];
+                /* cre*re cim*im */
+                t10 = _m_pfmul(c0, b0);
+                t11 = _m_pfmul(c1, b1);
+                /* no need to access cptr[2] & cptr[3] */
+                c0 = _m_pswapd(c0);
+                c1 = _m_pswapd(c1);
+                /* cim*re cre*im */
+                t20 = _m_pfmul(c0, b0);
+                t21 = _m_pfmul(c1, b1);
+
+                /* cre*re-cim*im cim*re+cre*im */
+                b0 = _m_pfpnacc(t10, t20);
+                b1 = _m_pfpnacc(t11, t21);
+
+                /* butterfly */
+                *(__m64 *)&p[0] = _m_pfadd(a0, b0);
+                *(__m64 *)&p[1] = _m_pfadd(a1, b1);
+                *(__m64 *)&q[0] = _m_pfsub(a0, b0);
+                *(__m64 *)&q[1] = _m_pfsub(a1, b1);
+
+                p += 2;
+                q += 2;
+                cptr += 4;
+            } while (--k);
+
+            p += nloops;
+            q += nloops;
+        } while (--j);
+        cptr1 += nloops * 2;
+        nblocks = nblocks >> 1;
+        nloops = nloops << 1;
+    } while (nblocks != 0);
+    _m_femms();
+}
+
+#endif