From 5f232025efa9d525fb709f6373d685db60e43316 Mon Sep 17 00:00:00 2001
From: Renaud Dartus <reno@videolan.org>
Date: Wed, 16 May 2001 14:51:29 +0000
Subject: [PATCH] * Add 3D Now! imdct * Remove kmudge for ac3 on MacOS X
---
Makefile | 7 +-
configure | 2 +-
configure.in | 2 +-
plugins/imdct/Makefile | 18 +-
plugins/imdct/ac3_imdct_3dn.c | 559 +++++++++++++++++++++++++++++++
plugins/imdct/ac3_imdct_c.c | 62 +---
plugins/imdct/ac3_imdct_common.c | 65 +---
plugins/imdct/ac3_imdct_common.h | 3 +-
plugins/imdct/ac3_imdct_sse.c | 58 +---
plugins/imdct/ac3_retables.h | 83 +++++
plugins/imdct/ac3_srfft_3dn.c | 344 +++++++++++++++++++
plugins/imdct/ac3_srfft_sse.c | 18 +-
plugins/imdct/imdct3dn.c | 152 +++++++++
src/ac3_decoder/ac3_imdct.c | 4 +-
14 files changed, 1180 insertions(+), 197 deletions(-)
create mode 100644 plugins/imdct/ac3_imdct_3dn.c
create mode 100644 plugins/imdct/ac3_retables.h
create mode 100644 plugins/imdct/ac3_srfft_3dn.c
create mode 100644 plugins/imdct/imdct3dn.c
diff --git a/Makefile b/Makefile
index 9a868def9f..4aa0e32453 100644
--- a/Makefile
+++ b/Makefile
@@ -26,7 +26,7 @@ PLUGINS_TARGETS := alsa/alsa beos/beos darwin/darwin dsp/dsp dummy/dummy \
dvd/dvd esd/esd fb/fb ggi/ggi glide/glide gnome/gnome gtk/gtk \
downmix/downmix downmix/downmixsse downmix/downmix3dn \
idct/idct idct/idctclassic idct/idctmmx idct/idctmmxext \
- imdct/imdct imdct/imdctsse \
+ imdct/imdct imdct/imdct3dn imdct/imdctsse \
macosx/macosx mga/mga \
motion/motion motion/motionmmx motion/motionmmxext \
mpeg/es mpeg/ps mpeg/ts null/null qt/qt sdl/sdl \
@@ -317,12 +317,7 @@ endif
$(C_OBJ): %.o: Makefile.opts Makefile.dep Makefile
$(C_OBJ): %.o: .dep/%.d
$(C_OBJ): %.o: %.c
-ifneq (,$(findstring darwin,$(SYS)))
-#this is uglier of all
- @if test "src/ac3_decoder/ac3_imdct.c" = "$<"; then $(CC) `echo $(CFLAGS) | sed -e 's/-O3/-O/'` -c -o $@ $<; echo "(CC) `echo $(CFLAGS) | sed -e 's/-O3/-O/'` -c -o $@ $<"; else $(CC) $(CFLAGS) -c -o $@ $<; echo "$(CC) $(CFLAGS) -c -o $@ $<"; fi
-else
$(CC) $(CFLAGS) -c -o $@ $<
-endif
$(CPP_OBJ): %.o: Makefile.opts Makefile.dep Makefile
$(CPP_OBJ): %.o: .dep/%.dpp
diff --git a/configure b/configure
index b75b5741a5..a03a4d4cad 100755
--- a/configure
+++ b/configure
@@ -3285,7 +3285,7 @@ int main() {
EOF
if { (eval echo configure:3287: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then
rm -rf conftest*
- ACCEL_PLUGINS="${ACCEL_PLUGINS} idctmmxext motionmmxext imdctsse downmix3dn downmixsse"
+ ACCEL_PLUGINS="${ACCEL_PLUGINS} idctmmxext motionmmxext imdct3dn imdctsse downmix3dn downmixsse"
echo "$ac_t""yes" 1>&6
else
echo "configure: failed program was:" >&5
diff --git a/configure.in b/configure.in
index 4ff3bade95..23766c7b0b 100644
--- a/configure.in
+++ b/configure.in
@@ -162,7 +162,7 @@ AC_TRY_COMPILE([void quux(){void *p;asm("packuswb %%mm1,%%mm2"::"r"(p));}],,
AC_MSG_CHECKING([if \$CC groks MMX EXT or SSE inline assembly])
AC_TRY_COMPILE([void quux(){void *p;asm("maskmovq %%mm1,%%mm2"::"r"(p));}],,
- ACCEL_PLUGINS="${ACCEL_PLUGINS} idctmmxext motionmmxext imdctsse downmix3dn downmixsse"
+ ACCEL_PLUGINS="${ACCEL_PLUGINS} idctmmxext motionmmxext imdct3dn imdctsse downmix3dn downmixsse"
AC_MSG_RESULT(yes), AC_MSG_RESULT(no))
dnl
diff --git a/plugins/imdct/Makefile b/plugins/imdct/Makefile
index 330287c9b6..b4dad20ac9 100644
--- a/plugins/imdct/Makefile
+++ b/plugins/imdct/Makefile
@@ -9,15 +9,18 @@
PLUGIN_IMDCT = imdct.o ac3_imdct_c.o ac3_srfft_c.o
PLUGIN_IMDCTSSE = imdctsse.o ac3_imdct_sse.o ac3_srfft_sse.o
+PLUGIN_IMDCT3DN = imdct3dn.o ac3_imdct_3dn.o ac3_srfft_3dn.o
PLUGIN_IMDCTCOMMON = ac3_imdct_common.o
BUILTIN_IMDCT = $(PLUGIN_IMDCT:%.o=BUILTIN_IMDCT_%.o) \
$(PLUGIN_IMDCTCOMMON:%.o=BUILTIN_IMDCT_%.o)
BUILTIN_IMDCTSSE = $(PLUGIN_IMDCTSSE:%.o=BUILTIN_IMDCTSSE_%.o) \
$(PLUGIN_IMDCTCOMMON:%.o=BUILTIN_IMDCTSSE_%.o)
+BUILTIN_IMDCT3DN = $(PLUGIN_IMDCT3DN:%.o=BUILTIN_IMDCT3DN_%.o) \
+ $(PLUGIN_IMDCTCOMMON:%.o=BUILTIN_IMDCT3DN_%.o)
-PLUGIN_C = $(PLUGIN_IMDCT) $(PLUGIN_IMDCTSSE) $(PLUGIN_IMDCTCOMMON)
-ALL_OBJ = $(PLUGIN_C) $(BUILTIN_IMDCT) $(BUILTIN_IMDCTSSE)
+PLUGIN_C = $(PLUGIN_IMDCT) $(PLUGIN_IMDCTSSE) $(PLUGIN_IMDCT3DN) $(PLUGIN_IMDCTCOMMON)
+ALL_OBJ = $(PLUGIN_C) $(BUILTIN_IMDCT) $(BUILTIN_IMDCTSSE) $(BUILTIN_IMDCT3DN)
#
# Virtual targets
@@ -33,6 +36,10 @@ $(BUILTIN_IMDCTSSE): BUILTIN_IMDCTSSE_%.o: .dep/%.d
$(BUILTIN_IMDCTSSE): BUILTIN_IMDCTSSE_%.o: %.c
$(CC) $(CFLAGS) -DBUILTIN -DMODULE_NAME=imdctsse -c -o $@ $<
+$(BUILTIN_IMDCT3DN): BUILTIN_IMDCT3DN_%.o: .dep/%.d
+$(BUILTIN_IMDCT3DN): BUILTIN_IMDCT3DN_%.o: %.c
+ $(CC) $(CFLAGS) -DBUILTIN -DMODULE_NAME=imdct3dn -c -o $@ $<
+
#
# Real targets
#
@@ -51,3 +58,10 @@ $(BUILTIN_IMDCTSSE): BUILTIN_IMDCTSSE_%.o: %.c
ar r $@ $^
$(RANLIB) $@
+../../lib/imdct3dn.so: $(PLUGIN_IMDCT3DN) $(PLUGIN_IMDCTCOMMON)
+ $(CC) $(PCFLAGS) -o $@ $^ $(PLCFLAGS)
+
+../../lib/imdct3dn.a: $(BUILTIN_IMDCT3DN)
+ ar r $@ $^
+ $(RANLIB) $@
+
diff --git a/plugins/imdct/ac3_imdct_3dn.c b/plugins/imdct/ac3_imdct_3dn.c
new file mode 100644
index 0000000000..5bb5e5d9a9
--- /dev/null
+++ b/plugins/imdct/ac3_imdct_3dn.c
@@ -0,0 +1,559 @@
+/*****************************************************************************
+ * ac3_imdct_3dn.c: accelerated 3D Now! ac3 DCT
+ *****************************************************************************
+ * Copyright (C) 1999, 2000 VideoLAN
+ * $Id: ac3_imdct_3dn.c,v 1.1 2001/05/16 14:51:29 reno Exp $
+ *
+ * Authors: Renaud Dartus <reno@videolan.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
+ *****************************************************************************/
+
+#define MODULE_NAME imdct3dn
+#include "modules_inner.h"
+
+/*****************************************************************************
+ * Preamble
+ *****************************************************************************/
+#include "defs.h"
+
+#include <math.h>
+#include <stdio.h>
+
+#include "config.h"
+#include "common.h"
+#include "threads.h"
+#include "mtime.h"
+
+#include "ac3_imdct.h"
+#include "ac3_imdct_common.h"
+#include "ac3_retables.h"
+
+void _M( fft_64p ) ( complex_t *x );
+void _M( fft_128p ) ( complex_t *a );
+
+static void imdct512_pre_ifft_twiddle_3dn (const int *pmt, complex_t *buf, float *data, float *xcos_sin_sse);
+static void imdct512_post_ifft_twiddle_3dn (complex_t *buf, float *xcos_sin_sse);
+static void imdct512_window_delay_3dn (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt);
+static void imdct512_window_delay_nol_3dn (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt);
+
+
+void _M( imdct_init ) (imdct_t * p_imdct)
+{
+ int i;
+ float scale = 181.019;
+
+ fprintf(stderr,"imct_init\n");
+ for (i=0; i < 128; i++)
+ {
+ float xcos_i = cos(2.0f * M_PI * (8*i+1)/(8*N)) * scale;
+ float xsin_i = sin(2.0f * M_PI * (8*i+1)/(8*N)) * scale;
+ p_imdct->xcos_sin_sse[i * 4] = xcos_i;
+ p_imdct->xcos_sin_sse[i * 4 + 1] = -xsin_i;
+ p_imdct->xcos_sin_sse[i * 4 + 2] = -xsin_i;
+ p_imdct->xcos_sin_sse[i * 4 + 3] = -xcos_i;
+ }
+ fprintf(stderr,"done imct_init\n");
+}
+
+void _M( imdct_do_512 ) (imdct_t * p_imdct, float data[], float delay[])
+{
+ imdct512_pre_ifft_twiddle_3dn (pm128, p_imdct->buf, data, p_imdct->xcos_sin_sse);
+ _M( fft_128p ) (p_imdct->buf);
+ imdct512_post_ifft_twiddle_3dn (p_imdct->buf, p_imdct->xcos_sin_sse);
+ imdct512_window_delay_3dn (p_imdct->buf, data, window, delay);
+}
+
+void _M( imdct_do_512_nol ) (imdct_t * p_imdct, float data[], float delay[])
+{
+ imdct512_pre_ifft_twiddle_3dn (pm128, p_imdct->buf, data, p_imdct->xcos_sin_sse);
+ _M( fft_128p ) (p_imdct->buf);
+ imdct512_post_ifft_twiddle_3dn (p_imdct->buf, p_imdct->xcos_sin_sse);
+ imdct512_window_delay_nol_3dn (p_imdct->buf, data, window, delay);
+}
+
+static void imdct512_pre_ifft_twiddle_3dn (const int *pmt, complex_t *buf, float *data, float *xcos_sin_sse)
+{
+ __asm__ __volatile__ (
+ "pushl %%ebp\n"
+ "movl %%esp, %%ebp\n"
+ "addl $-4, %%esp\n" /* local variable, loop counter */
+
+ "pushl %%eax\n"
+ "pushl %%ebx\n"
+ "pushl %%ecx\n"
+ "pushl %%edx\n"
+ "pushl %%edi\n"
+ "pushl %%esi\n"
+
+ "movl 8(%%ebp), %%eax\n" /* pmt */
+ "movl 12(%%ebp), %%ebx\n" /* buf */
+ "movl 16(%%ebp), %%ecx\n" /* data */
+ "movl 20(%%ebp), %%edx\n" /* xcos_sin_sse */
+ "movl $128, -4(%%ebp)\n"
+
+".loop:\n"
+ "movl (%%eax), %%esi\n"
+ "movd (%%ecx, %%esi, 8), %%mm1\n" /* 2j */
+ "punpckldq %%mm1, %%mm1\n" /* 2j | 2j */
+
+ "shll $1, %%esi\n"
+
+ "movq (%%edx, %%esi, 8), %%mm0\n" /* -s_j | c_j */
+ "movq 8(%%edx, %%esi, 8), %%mm2\n" /* -c_j | -s_j */
+
+ "negl %%esi\n"
+
+ "movd 1020(%%ecx, %%esi, 4), %%mm4\n" /* 255-2j */
+ "punpckldq %%mm4, %%mm4\n" /* 255-2j | 255-2j */
+ "addl $4, %%eax\n"
+
+ "pfmul %%mm4, %%mm0\n" /* 255-2j * -s_j | 255-2j * c_j */
+ "pfmul %%mm1, %%mm2\n" /* 2j * -c_j | 2j * -s_j */
+ "addl $8, %%ebx\n"
+ "pfadd %%mm2, %%mm0\n" /* 2j * -c_j + 255-2j * -s_j | 2j * -s_j + 255-2j * c_j */
+
+ "movq %%mm0, -8(%%ebx)\n"
+ "decl -4(%%ebp)\n"
+ "jnz .loop\n"
+
+ "popl %%esi\n"
+ "popl %%edi\n"
+ "popl %%edx\n"
+ "popl %%ecx\n"
+ "popl %%ebx\n"
+ "popl %%eax\n"
+
+ "addl $4, %%esp\n"
+ "popl %%ebp\n"
+ "femms\n"
+ ::);
+}
+
+static void imdct512_post_ifft_twiddle_3dn (complex_t *buf, float *xcos_sin_sse)
+{
+ __asm__ __volatile__ (
+ "pushl %%ebx\n"
+ "movl $64, %%ebx\n" /* loop counter */
+
+".loop1:\n"
+ "movq (%%eax), %%mm0\n" /* im0 | re0 */
+ "movq %%mm0, %%mm1\n" /* im0 | re0 */
+ "punpckldq %%mm0, %%mm0\n" /* re0 | re0 */
+ "punpckhdq %%mm1, %%mm1\n" /* im0 | im0 */
+
+ "movq (%%ecx), %%mm2\n" /* -s | c */
+ "movq 8(%%ecx), %%mm3\n" /* -c | -s */
+ "movq %%mm3, %%mm4\n"
+
+ "punpckhdq %%mm2,%%mm3\n" /* -s | -c */
+ "punpckldq %%mm2,%%mm4\n" /* c | -s */
+
+ "movq 8(%%eax), %%mm2\n" /* im1 | re1 */
+ "movq %%mm2, %%mm5\n" /* im1 | re1 */
+ "punpckldq %%mm2, %%mm2\n" /* re1 | re1 */
+ "punpckhdq %%mm5, %%mm5\n" /* im1 | im1 */
+
+ "pfmul %%mm3, %%mm0\n" /* -s * re0 | -c * re0 */
+ "pfmul %%mm4, %%mm1\n" /* c * im0 | -s * im0 */
+
+ "movq 16(%%ecx), %%mm6\n" /* -s1 | c1 */
+ "movq 24(%%ecx), %%mm7\n" /* -c1 | -s1 */
+ "movq %%mm7, %%mm4\n"
+
+ "punpckhdq %%mm6,%%mm7\n" /* -s1 | -c1 */
+ "punpckldq %%mm6,%%mm4\n" /* c1 | -s1 */
+
+ "pfmul %%mm7, %%mm2\n" /* -s1*re1 | -c1*re1 */
+ "pfmul %%mm4, %%mm5\n" /* c1*im1 | -s1*im1 */
+
+ "pfadd %%mm1, %%mm0\n" /* -s * re0 + c * im0 | -c * re0 - s * im0 */
+ "pfadd %%mm5, %%mm2\n" /* -s1 * re1 + c1 * im1 | -c1 * re1 - s1 * im1 */
+
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm2, 8(%%eax)\n"
+ "addl $32, %%ecx\n"
+ "addl $16, %%eax\n"
+ "decl %%ebx\n"
+ "jnz .loop1\n"
+
+ "popl %%ebx\n"
+ "femms\n"
+ : "=a" (buf)
+ : "a" (buf), "c" (xcos_sin_sse) );
+}
+
+static void imdct512_window_delay_3dn (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt)
+{
+ __asm__ __volatile__ (
+ "pushl %%ebp\n"
+ "movl %%esp, %%ebp\n"
+
+ "pushl %%eax\n"
+ "pushl %%ebx\n"
+ "pushl %%ecx\n"
+ "pushl %%edx\n"
+ "pushl %%esi\n"
+ "pushl %%edi\n"
+
+ "movl 20(%%ebp), %%ebx\n" /* delay */
+ "movl 16(%%ebp), %%edx\n" /* window */
+
+ "movl 8(%%ebp), %%eax\n" /* buf */
+ "movl $32, %%ecx\n" /* loop count */
+ "leal 516(%%eax), %%esi\n" /* buf[64].im */
+ "leal 504(%%eax), %%edi\n" /* buf[63].re */
+ "movl 12(%%ebp), %%eax\n" /* data */
+
+".first_128_samples:\n"
+ "movd (%%esi), %%mm0\n" /* im0 */
+ "movd 8(%%esi), %%mm2\n" /* im1 */
+ "movd (%%edi), %%mm1\n" /* re0 */
+ "movd -8(%%edi), %%mm3\n" /* re1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm0, %%mm4\n" /* -im0 */
+ "pfsub %%mm2, %%mm5\n" /* -im1 */
+
+ "punpckldq %%mm1, %%mm4\n" /* re0 | -im0 */
+ "punpckldq %%mm3, %%mm5\n" /* re1 | -im1 */
+
+ "movq (%%edx), %%mm0\n" /* w1 | w0 */
+ "movq 8(%%edx), %%mm1\n" /* w3 | w2 */
+ "movq (%%ebx), %%mm2\n" /* d1 | d0 */
+ "movq 8(%%ebx), %%mm3\n" /* d3 | d2 */
+
+ "pfmul %%mm4, %%mm0\n" /* w1*re0 | -w0*im0 */
+ "pfmul %%mm5, %%mm1\n" /* w3*re1 | -w2*im1 */
+
+ "pfadd %%mm2, %%mm0\n" /* w1*re0+d1 | -w0*im0+d0 */
+ "pfadd %%mm3, %%mm1\n" /* w3*re1+d3 | -w2*im1+d2 */
+
+ "addl $16, %%edx\n"
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm1, 8(%%eax)\n"
+ "addl $16, %%ebx\n"
+ "addl $16, %%esi\n"
+ "addl $16, %%eax\n"
+ "addl $-16, %%edi\n"
+ "decl %%ecx\n"
+ "jnz .first_128_samples\n"
+
+ "movl 8(%%ebp), %%esi\n" /* buf[0].re */
+ "leal 1020(%%esi), %%edi\n" /* buf[127].im */
+ "movl $32, %%ecx\n" /* loop count */
+
+".second_128_samples:\n"
+ "movd (%%esi), %%mm0\n" /* buf[i].re */
+ "movd 8(%%esi), %%mm2\n" /* re1 */
+ "movd (%%edi), %%mm1\n" /* buf[127-i].im */
+ "movd -8(%%edi), %%mm3\n" /* im1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm0, %%mm4\n" /* -re0 */
+ "pfsub %%mm2, %%mm5\n" /* -re1 */
+
+ "punpckldq %%mm1, %%mm4\n" /* im0 | -re0 */
+ "punpckldq %%mm3, %%mm5\n" /* im1 | -re1 */
+
+ "movq (%%edx), %%mm0\n" /* w1 | w0 */
+ "movq 8(%%edx), %%mm1\n" /* w3 | w2 */
+ "movq (%%ebx), %%mm2\n" /* d1 | d0 */
+ "movq 8(%%ebx), %%mm3\n" /* d3 | d2 */
+
+ "addl $16, %%esi\n"
+
+ "pfmul %%mm4, %%mm0\n" /* w1*im0 | -w0*re0 */
+ "pfmul %%mm5, %%mm1\n" /* w3*im1 | -w2*re1 */
+
+ "pfadd %%mm2, %%mm0\n" /* w1*im0+d1 | -w0*re0+d0 */
+ "pfadd %%mm3, %%mm1\n" /* w3*im1+d3 | -w2*re1+d2 */
+
+ "addl $-16, %%edi\n"
+
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm1, 8(%%eax)\n"
+
+ "addl $16, %%edx\n"
+ "addl $16, %%eax\n"
+ "addl $16, %%ebx\n"
+ "decl %%ecx\n"
+ "jnz .second_128_samples\n"
+
+ "movl 8(%%ebp), %%eax\n"
+ "leal 512(%%eax), %%esi\n" /* buf[64].re */
+ "leal 508(%%eax), %%edi\n" /* buf[63].im */
+ "movl $32, %%ecx\n" /* loop count */
+ "movl 20(%%ebp), %%eax\n" /* delay */
+
+".first_128_delay:\n"
+ "movd (%%esi), %%mm0\n" /* re0 */
+ "movd 8(%%esi), %%mm2\n" /* re1 */
+ "movd (%%edi), %%mm1\n" /* im0 */
+ "movd -8(%%edi), %%mm3\n" /* im1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm0, %%mm4\n" /* -re0 */
+ "pfsub %%mm2, %%mm5\n" /* -re1 */
+
+ "punpckldq %%mm1, %%mm4\n" /* im0 | -re0 */
+ "punpckldq %%mm3, %%mm5\n" /* im1 | -re1 */
+
+
+ "movq -16(%%edx), %%mm1\n" /* w3 | w2 */
+ "movq -8(%%edx), %%mm0\n" /* w1 | w0 */
+
+ "addl $-16, %%edx\n"
+
+ "pfmul %%mm4, %%mm0\n" /* w1*im0 | -w0*re0 */
+ "pfmul %%mm5, %%mm1\n" /* w3*im1 | -w2*re1 */
+
+
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm1, 8(%%eax)\n"
+ "addl $16, %%esi\n"
+ "addl $-16, %%edi\n"
+ "addl $16, %%eax\n"
+ "decl %%ecx\n"
+ "jnz .first_128_delay\n"
+
+ "movl 8(%%ebp), %%ebx\n"
+ "leal 4(%%ebx), %%esi\n" /* buf[0].im */
+ "leal 1016(%%ebx), %%edi\n" /* buf[127].re */
+ "movl $32, %%ecx\n" /* loop count */
+
+".second_128_delay:\n"
+ "movd (%%esi), %%mm0\n" /* im0 */
+ "movd 8(%%esi), %%mm2\n" /* im1 */
+ "movd (%%edi), %%mm1\n" /* re0 */
+ "movd -8(%%edi), %%mm3\n" /* re1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm1, %%mm4\n" /* -re0 */
+ "pfsub %%mm3, %%mm5\n" /* -re1 */
+
+ "punpckldq %%mm4, %%mm0\n" /* -re0 | im0 */
+ "punpckldq %%mm5, %%mm2\n" /* -re1 | im1 */
+
+
+ "movq -16(%%edx), %%mm1\n" /* w3 | w2 */
+ "movq -8(%%edx), %%mm3\n" /* w1 | w0 */
+
+ "addl $-16, %%edx\n"
+
+ "pfmul %%mm0, %%mm1\n" /* -w1*re0 | w0*im0 */
+ "pfmul %%mm2, %%mm3\n" /* -w3*re1 | w2*im1 */
+
+
+ "movq %%mm1, (%%eax)\n"
+ "movq %%mm3, 8(%%eax)\n"
+ "addl $16, %%esi\n"
+ "addl $-16, %%edi\n"
+ "addl $16, %%eax\n"
+ "decl %%ecx\n"
+ "jnz .second_128_delay\n"
+
+ "popl %%edi\n"
+ "popl %%esi\n"
+ "popl %%edx\n"
+ "popl %%ecx\n"
+ "popl %%ebx\n"
+ "popl %%eax\n"
+
+ "leave\n"
+ "femms\n"
+ ::);
+}
+
+static void imdct512_window_delay_nol_3dn (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt)
+{
+ __asm__ __volatile__ (
+ "pushl %%ebp\n"
+ "movl %%esp, %%ebp\n"
+
+ "pushl %%eax\n"
+ "pushl %%ebx\n"
+ "pushl %%ecx\n"
+ "pushl %%edx\n"
+ "pushl %%esi\n"
+ "pushl %%edi\n"
+
+ "movl 20(%%ebp), %%ebx\n" /* delay */
+ "movl 16(%%ebp), %%edx\n" /* window */
+
+ "movl 8(%%ebp), %%eax\n" /* buf */
+ "movl $32, %%ecx\n" /* loop count */
+ "leal 516(%%eax), %%esi\n" /* buf[64].im */
+ "leal 504(%%eax), %%edi\n" /* buf[63].re */
+ "movl 12(%%ebp), %%eax\n" /* data */
+
+".first_128_samples2:\n"
+ "movd (%%esi), %%mm0\n" /* im0 */
+ "movd 8(%%esi), %%mm2\n" /* im1 */
+ "movd (%%edi), %%mm1\n" /* re0 */
+ "movd -8(%%edi), %%mm3\n" /* re1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm0, %%mm4\n" /* -im0 */
+ "pfsub %%mm2, %%mm5\n" /* -im1 */
+
+ "punpckldq %%mm1, %%mm4\n" /* re0 | -im0 */
+ "punpckldq %%mm3, %%mm5\n" /* re1 | -im1 */
+
+ "movq (%%edx), %%mm0\n" /* w1 | w0 */
+ "movq 8(%%edx), %%mm1\n" /* w3 | w2 */
+
+ "pfmul %%mm4, %%mm0\n" /* w1*re0 | -w0*im0 */
+ "pfmul %%mm5, %%mm1\n" /* w3*re1 | -w2*im1 */
+
+ "addl $16, %%edx\n"
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm1, 8(%%eax)\n"
+ "addl $16, %%ebx\n"
+ "addl $16, %%esi\n"
+ "addl $16, %%eax\n"
+ "addl $-16, %%edi\n"
+ "decl %%ecx\n"
+ "jnz .first_128_samples2\n"
+
+ "movl 8(%%ebp), %%esi\n" /* buf[0].re */
+ "leal 1020(%%esi), %%edi\n" /* buf[127].im */
+ "movl $32, %%ecx\n" /* loop count */
+
+".second_128_samples2:\n"
+ "movd (%%esi), %%mm0\n" /* buf[i].re */
+ "movd 8(%%esi), %%mm2\n" /* re1 */
+ "movd (%%edi), %%mm1\n" /* buf[127-i].im */
+ "movd -8(%%edi), %%mm3\n" /* im1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm0, %%mm4\n" /* -re0 */
+ "pfsub %%mm2, %%mm5\n" /* -re1 */
+
+ "punpckldq %%mm1, %%mm4\n" /* im0 | -re0 */
+ "punpckldq %%mm3, %%mm5\n" /* im1 | -re1 */
+
+ "movq (%%edx), %%mm0\n" /* w1 | w0 */
+ "movq 8(%%edx), %%mm1\n" /* w3 | w2 */
+
+ "addl $16, %%esi\n"
+
+ "pfmul %%mm4, %%mm0\n" /* w1*im0 | -w0*re0 */
+ "pfmul %%mm5, %%mm1\n" /* w3*im1 | -w2*re1 */
+
+ "addl $-16, %%edi\n"
+
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm1, 8(%%eax)\n"
+
+ "addl $16, %%edx\n"
+ "addl $16, %%eax\n"
+ "addl $16, %%ebx\n"
+ "decl %%ecx\n"
+ "jnz .second_128_samples2\n"
+
+ "movl 8(%%ebp), %%eax\n"
+ "leal 512(%%eax), %%esi\n" /* buf[64].re */
+ "leal 508(%%eax), %%edi\n" /* buf[63].im */
+ "movl $32, %%ecx\n" /* loop count */
+ "movl 20(%%ebp), %%eax\n" /* delay */
+
+".first_128_delays:\n"
+ "movd (%%esi), %%mm0\n" /* re0 */
+ "movd 8(%%esi), %%mm2\n" /* re1 */
+ "movd (%%edi), %%mm1\n" /* im0 */
+ "movd -8(%%edi), %%mm3\n" /* im1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm0, %%mm4\n" /* -re0 */
+ "pfsub %%mm2, %%mm5\n" /* -re1 */
+
+ "punpckldq %%mm1, %%mm4\n" /* im0 | -re0 */
+ "punpckldq %%mm3, %%mm5\n" /* im1 | -re1 */
+
+
+ "movq -16(%%edx), %%mm1\n" /* w3 | w2 */
+ "movq -8(%%edx), %%mm0\n" /* w1 | w0 */
+
+ "addl $-16, %%edx\n"
+
+ "pfmul %%mm4, %%mm0\n" /* w1*im0 | -w0*re0 */
+ "pfmul %%mm5, %%mm1\n" /* w3*im1 | -w2*re1 */
+
+
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm1, 8(%%eax)\n"
+ "addl $16, %%esi\n"
+ "addl $-16, %%edi\n"
+ "addl $16, %%eax\n"
+ "decl %%ecx\n"
+ "jnz .first_128_delays\n"
+
+ "movl 8(%%ebp), %%ebx\n"
+ "leal 4(%%ebx), %%esi\n" /* buf[0].im */
+ "leal 1016(%%ebx), %%edi\n" /* buf[127].re */
+ "movl $32, %%ecx\n" /* loop count */
+
+".second_128_delays:\n"
+ "movd (%%esi), %%mm0\n" /* im0 */
+ "movd 8(%%esi), %%mm2\n" /* im1 */
+ "movd (%%edi), %%mm1\n" /* re0 */
+ "movd -8(%%edi), %%mm3\n" /* re1 */
+
+ "pxor %%mm4, %%mm4\n"
+ "pxor %%mm5, %%mm5\n"
+ "pfsub %%mm1, %%mm4\n" /* -re0 */
+ "pfsub %%mm3, %%mm5\n" /* -re1 */
+
+ "punpckldq %%mm4, %%mm0\n" /* -re0 | im0 */
+ "punpckldq %%mm5, %%mm2\n" /* -re1 | im1 */
+
+
+ "movq -16(%%edx), %%mm1\n" /* w3 | w2 */
+ "movq -8(%%edx), %%mm3\n" /* w1 | w0 */
+
+ "addl $-16, %%edx\n"
+
+ "pfmul %%mm0, %%mm1\n" /* -w1*re0 | w0*im0 */
+ "pfmul %%mm2, %%mm3\n" /* -w3*re1 | w2*im1 */
+
+
+ "movq %%mm1, (%%eax)\n"
+ "movq %%mm3, 8(%%eax)\n"
+ "addl $16, %%esi\n"
+ "addl $-16, %%edi\n"
+ "addl $16, %%eax\n"
+ "decl %%ecx\n"
+ "jnz .second_128_delays\n"
+
+ "popl %%edi\n"
+ "popl %%esi\n"
+ "popl %%edx\n"
+ "popl %%ecx\n"
+ "popl %%ebx\n"
+ "popl %%eax\n"
+
+ "leave\n"
+ "femms\n"
+ ::);
+}
+
diff --git a/plugins/imdct/ac3_imdct_c.c b/plugins/imdct/ac3_imdct_c.c
index 3ebf16c9b3..106d7faaaf 100644
--- a/plugins/imdct/ac3_imdct_c.c
+++ b/plugins/imdct/ac3_imdct_c.c
@@ -2,7 +2,7 @@
* ac3_imdct_c.c: ac3 DCT in C
*****************************************************************************
* Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_imdct_c.c,v 1.1 2001/05/15 16:19:42 sam Exp $
+ * $Id: ac3_imdct_c.c,v 1.2 2001/05/16 14:51:29 reno Exp $
*
* Authors: Renaud Dartus <reno@videolan.org>
* Aaron Holtzman <aholtzma@engr.uvic.ca>
@@ -42,6 +42,7 @@
#include "ac3_imdct.h"
#include "ac3_imdct_common.h"
+#include "ac3_retables.h"
#ifndef M_PI
# define M_PI 3.14159265358979323846
@@ -50,65 +51,6 @@
void _M( fft_64p ) ( complex_t *x );
void _M( fft_128p ) ( complex_t *x );
-static float window[] = {
- 0.00014, 0.00024, 0.00037, 0.00051, 0.00067, 0.00086, 0.00107, 0.00130,
- 0.00157, 0.00187, 0.00220, 0.00256, 0.00297, 0.00341, 0.00390, 0.00443,
- 0.00501, 0.00564, 0.00632, 0.00706, 0.00785, 0.00871, 0.00962, 0.01061,
- 0.01166, 0.01279, 0.01399, 0.01526, 0.01662, 0.01806, 0.01959, 0.02121,
- 0.02292, 0.02472, 0.02662, 0.02863, 0.03073, 0.03294, 0.03527, 0.03770,
- 0.04025, 0.04292, 0.04571, 0.04862, 0.05165, 0.05481, 0.05810, 0.06153,
- 0.06508, 0.06878, 0.07261, 0.07658, 0.08069, 0.08495, 0.08935, 0.09389,
- 0.09859, 0.10343, 0.10842, 0.11356, 0.11885, 0.12429, 0.12988, 0.13563,
- 0.14152, 0.14757, 0.15376, 0.16011, 0.16661, 0.17325, 0.18005, 0.18699,
- 0.19407, 0.20130, 0.20867, 0.21618, 0.22382, 0.23161, 0.23952, 0.24757,
- 0.25574, 0.26404, 0.27246, 0.28100, 0.28965, 0.29841, 0.30729, 0.31626,
- 0.32533, 0.33450, 0.34376, 0.35311, 0.36253, 0.37204, 0.38161, 0.39126,
- 0.40096, 0.41072, 0.42054, 0.43040, 0.44030, 0.45023, 0.46020, 0.47019,
- 0.48020, 0.49022, 0.50025, 0.51028, 0.52031, 0.53033, 0.54033, 0.55031,
- 0.56026, 0.57019, 0.58007, 0.58991, 0.59970, 0.60944, 0.61912, 0.62873,
- 0.63827, 0.64774, 0.65713, 0.66643, 0.67564, 0.68476, 0.69377, 0.70269,
- 0.71150, 0.72019, 0.72877, 0.73723, 0.74557, 0.75378, 0.76186, 0.76981,
- 0.77762, 0.78530, 0.79283, 0.80022, 0.80747, 0.81457, 0.82151, 0.82831,
- 0.83496, 0.84145, 0.84779, 0.85398, 0.86001, 0.86588, 0.87160, 0.87716,
- 0.88257, 0.88782, 0.89291, 0.89785, 0.90264, 0.90728, 0.91176, 0.91610,
- 0.92028, 0.92432, 0.92822, 0.93197, 0.93558, 0.93906, 0.94240, 0.94560,
- 0.94867, 0.95162, 0.95444, 0.95713, 0.95971, 0.96217, 0.96451, 0.96674,
- 0.96887, 0.97089, 0.97281, 0.97463, 0.97635, 0.97799, 0.97953, 0.98099,
- 0.98236, 0.98366, 0.98488, 0.98602, 0.98710, 0.98811, 0.98905, 0.98994,
- 0.99076, 0.99153, 0.99225, 0.99291, 0.99353, 0.99411, 0.99464, 0.99513,
- 0.99558, 0.99600, 0.99639, 0.99674, 0.99706, 0.99736, 0.99763, 0.99788,
- 0.99811, 0.99831, 0.99850, 0.99867, 0.99882, 0.99895, 0.99908, 0.99919,
- 0.99929, 0.99938, 0.99946, 0.99953, 0.99959, 0.99965, 0.99969, 0.99974,
- 0.99978, 0.99981, 0.99984, 0.99986, 0.99988, 0.99990, 0.99992, 0.99993,
- 0.99994, 0.99995, 0.99996, 0.99997, 0.99998, 0.99998, 0.99998, 0.99999,
- 0.99999, 0.99999, 0.99999, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000,
- 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000
-};
-
-static const int pm128[128] =
-{
- 0, 16, 32, 48, 64, 80, 96, 112, 8, 40, 72, 104, 24, 56, 88, 120,
- 4, 20, 36, 52, 68, 84, 100, 116, 12, 28, 44, 60, 76, 92, 108, 124,
- 2, 18, 34, 50, 66, 82, 98, 114, 10, 42, 74, 106, 26, 58, 90, 122,
- 6, 22, 38, 54, 70, 86, 102, 118, 14, 46, 78, 110, 30, 62, 94, 126,
- 1, 17, 33, 49, 65, 81, 97, 113, 9, 41, 73, 105, 25, 57, 89, 121,
- 5, 21, 37, 53, 69, 85, 101, 117, 13, 29, 45, 61, 77, 93, 109, 125,
- 3, 19, 35, 51, 67, 83, 99, 115, 11, 43, 75, 107, 27, 59, 91, 123,
- 7, 23, 39, 55, 71, 87, 103, 119, 15, 31, 47, 63, 79, 95, 111, 127
-};
-
-static const int pm64[64] =
-{
- 0, 8, 16, 24, 32, 40, 48, 56,
- 4, 20, 36, 52, 12, 28, 44, 60,
- 2, 10, 18, 26, 34, 42, 50, 58,
- 6, 14, 22, 30, 38, 46, 54, 62,
- 1, 9, 17, 25, 33, 41, 49, 57,
- 5, 21, 37, 53, 13, 29, 45, 61,
- 3, 11, 19, 27, 35, 43, 51, 59,
- 7, 23, 39, 55, 15, 31, 47, 63
-};
-
void _M( imdct_init ) (imdct_t * p_imdct)
{
int i;
diff --git a/plugins/imdct/ac3_imdct_common.c b/plugins/imdct/ac3_imdct_common.c
index 493eca2d12..b5bdc6d64b 100644
--- a/plugins/imdct/ac3_imdct_common.c
+++ b/plugins/imdct/ac3_imdct_common.c
@@ -2,7 +2,7 @@
* ac3_imdct_common.c: common ac3 DCT functions
*****************************************************************************
* Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_imdct_common.c,v 1.2 2001/05/15 19:36:27 sam Exp $
+ * $Id: ac3_imdct_common.c,v 1.3 2001/05/16 14:51:29 reno Exp $
*
* Authors: Renaud Dartus <reno@videolan.org>
* Aaron Holtzman <aholtzma@engr.uvic.ca>
@@ -45,71 +45,13 @@
#include "mtime.h"
#include "ac3_imdct.h"
+#include "ac3_retables.h"
#ifndef M_PI
# define M_PI 3.14159265358979323846
#endif
-static float window[] = {
- 0.00014, 0.00024, 0.00037, 0.00051, 0.00067, 0.00086, 0.00107, 0.00130,
- 0.00157, 0.00187, 0.00220, 0.00256, 0.00297, 0.00341, 0.00390, 0.00443,
- 0.00501, 0.00564, 0.00632, 0.00706, 0.00785, 0.00871, 0.00962, 0.01061,
- 0.01166, 0.01279, 0.01399, 0.01526, 0.01662, 0.01806, 0.01959, 0.02121,
- 0.02292, 0.02472, 0.02662, 0.02863, 0.03073, 0.03294, 0.03527, 0.03770,
- 0.04025, 0.04292, 0.04571, 0.04862, 0.05165, 0.05481, 0.05810, 0.06153,
- 0.06508, 0.06878, 0.07261, 0.07658, 0.08069, 0.08495, 0.08935, 0.09389,
- 0.09859, 0.10343, 0.10842, 0.11356, 0.11885, 0.12429, 0.12988, 0.13563,
- 0.14152, 0.14757, 0.15376, 0.16011, 0.16661, 0.17325, 0.18005, 0.18699,
- 0.19407, 0.20130, 0.20867, 0.21618, 0.22382, 0.23161, 0.23952, 0.24757,
- 0.25574, 0.26404, 0.27246, 0.28100, 0.28965, 0.29841, 0.30729, 0.31626,
- 0.32533, 0.33450, 0.34376, 0.35311, 0.36253, 0.37204, 0.38161, 0.39126,
- 0.40096, 0.41072, 0.42054, 0.43040, 0.44030, 0.45023, 0.46020, 0.47019,
- 0.48020, 0.49022, 0.50025, 0.51028, 0.52031, 0.53033, 0.54033, 0.55031,
- 0.56026, 0.57019, 0.58007, 0.58991, 0.59970, 0.60944, 0.61912, 0.62873,
- 0.63827, 0.64774, 0.65713, 0.66643, 0.67564, 0.68476, 0.69377, 0.70269,
- 0.71150, 0.72019, 0.72877, 0.73723, 0.74557, 0.75378, 0.76186, 0.76981,
- 0.77762, 0.78530, 0.79283, 0.80022, 0.80747, 0.81457, 0.82151, 0.82831,
- 0.83496, 0.84145, 0.84779, 0.85398, 0.86001, 0.86588, 0.87160, 0.87716,
- 0.88257, 0.88782, 0.89291, 0.89785, 0.90264, 0.90728, 0.91176, 0.91610,
- 0.92028, 0.92432, 0.92822, 0.93197, 0.93558, 0.93906, 0.94240, 0.94560,
- 0.94867, 0.95162, 0.95444, 0.95713, 0.95971, 0.96217, 0.96451, 0.96674,
- 0.96887, 0.97089, 0.97281, 0.97463, 0.97635, 0.97799, 0.97953, 0.98099,
- 0.98236, 0.98366, 0.98488, 0.98602, 0.98710, 0.98811, 0.98905, 0.98994,
- 0.99076, 0.99153, 0.99225, 0.99291, 0.99353, 0.99411, 0.99464, 0.99513,
- 0.99558, 0.99600, 0.99639, 0.99674, 0.99706, 0.99736, 0.99763, 0.99788,
- 0.99811, 0.99831, 0.99850, 0.99867, 0.99882, 0.99895, 0.99908, 0.99919,
- 0.99929, 0.99938, 0.99946, 0.99953, 0.99959, 0.99965, 0.99969, 0.99974,
- 0.99978, 0.99981, 0.99984, 0.99986, 0.99988, 0.99990, 0.99992, 0.99993,
- 0.99994, 0.99995, 0.99996, 0.99997, 0.99998, 0.99998, 0.99998, 0.99999,
- 0.99999, 0.99999, 0.99999, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000,
- 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000
-};
-
-static const int pm128[128] =
-{
- 0, 16, 32, 48, 64, 80, 96, 112, 8, 40, 72, 104, 24, 56, 88, 120,
- 4, 20, 36, 52, 68, 84, 100, 116, 12, 28, 44, 60, 76, 92, 108, 124,
- 2, 18, 34, 50, 66, 82, 98, 114, 10, 42, 74, 106, 26, 58, 90, 122,
- 6, 22, 38, 54, 70, 86, 102, 118, 14, 46, 78, 110, 30, 62, 94, 126,
- 1, 17, 33, 49, 65, 81, 97, 113, 9, 41, 73, 105, 25, 57, 89, 121,
- 5, 21, 37, 53, 69, 85, 101, 117, 13, 29, 45, 61, 77, 93, 109, 125,
- 3, 19, 35, 51, 67, 83, 99, 115, 11, 43, 75, 107, 27, 59, 91, 123,
- 7, 23, 39, 55, 71, 87, 103, 119, 15, 31, 47, 63, 79, 95, 111, 127
-};
-
-static const int pm64[64] =
-{
- 0, 8, 16, 24, 32, 40, 48, 56,
- 4, 20, 36, 52, 12, 28, 44, 60,
- 2, 10, 18, 26, 34, 42, 50, 58,
- 6, 14, 22, 30, 38, 46, 54, 62,
- 1, 9, 17, 25, 33, 41, 49, 57,
- 5, 21, 37, 53, 13, 29, 45, 61,
- 3, 11, 19, 27, 35, 43, 51, 59,
- 7, 23, 39, 55, 15, 31, 47, 63
-};
-
-void _M( fft_64p ) ( complex_t *a );
+void _M( fft_64p ) ( complex_t *x );
void _M( imdct_do_256 ) (imdct_t * p_imdct, float data[],float delay[])
{
@@ -266,4 +208,3 @@ void _M( imdct_do_256_nol ) (imdct_t * p_imdct, float data[], float delay[])
*delay_ptr++ = -buf2[64-i-1].real * *--window_ptr;
}
}
-
diff --git a/plugins/imdct/ac3_imdct_common.h b/plugins/imdct/ac3_imdct_common.h
index ce0a7ab6d8..3977b2a3ba 100644
--- a/plugins/imdct/ac3_imdct_common.h
+++ b/plugins/imdct/ac3_imdct_common.h
@@ -2,7 +2,7 @@
* ac3_imdct_common.h: common ac3 DCT headers
*****************************************************************************
* Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_imdct_common.h,v 1.1 2001/05/15 16:19:42 sam Exp $
+ * $Id: ac3_imdct_common.h,v 1.2 2001/05/16 14:51:29 reno Exp $
*
* Authors: Renaud Dartus <reno@videolan.org>
* Aaron Holtzman <aholtzma@engr.uvic.ca>
@@ -27,4 +27,3 @@ void _M( imdct_do_256 ) ( imdct_t * p_imdct, float data[], float delay[] );
void _M( imdct_do_256_nol ) ( imdct_t * p_imdct, float data[], float delay[] );
void _M( imdct_do_512 ) ( imdct_t * p_imdct, float data[], float delay[] );
void _M( imdct_do_512_nol ) ( imdct_t * p_imdct, float data[], float delay[] );
-
diff --git a/plugins/imdct/ac3_imdct_sse.c b/plugins/imdct/ac3_imdct_sse.c
index d426f55a66..a9dad29166 100644
--- a/plugins/imdct/ac3_imdct_sse.c
+++ b/plugins/imdct/ac3_imdct_sse.c
@@ -2,7 +2,7 @@
* ac3_imdct_sse.c: accelerated SSE ac3 DCT
*****************************************************************************
* Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_imdct_sse.c,v 1.1 2001/05/15 16:19:42 sam Exp $
+ * $Id: ac3_imdct_sse.c,v 1.2 2001/05/16 14:51:29 reno Exp $
*
* Authors: Renaud Dartus <reno@videolan.org>
* Aaron Holtzman <aholtzma@engr.uvic.ca>
@@ -40,61 +40,15 @@
#include "ac3_imdct.h"
#include "ac3_imdct_common.h"
-
-static const float window[] = {
- 0.00014, 0.00024, 0.00037, 0.00051, 0.00067, 0.00086, 0.00107, 0.00130,
- 0.00157, 0.00187, 0.00220, 0.00256, 0.00297, 0.00341, 0.00390, 0.00443,
- 0.00501, 0.00564, 0.00632, 0.00706, 0.00785, 0.00871, 0.00962, 0.01061,
- 0.01166, 0.01279, 0.01399, 0.01526, 0.01662, 0.01806, 0.01959, 0.02121,
- 0.02292, 0.02472, 0.02662, 0.02863, 0.03073, 0.03294, 0.03527, 0.03770,
- 0.04025, 0.04292, 0.04571, 0.04862, 0.05165, 0.05481, 0.05810, 0.06153,
- 0.06508, 0.06878, 0.07261, 0.07658, 0.08069, 0.08495, 0.08935, 0.09389,
- 0.09859, 0.10343, 0.10842, 0.11356, 0.11885, 0.12429, 0.12988, 0.13563,
- 0.14152, 0.14757, 0.15376, 0.16011, 0.16661, 0.17325, 0.18005, 0.18699,
- 0.19407, 0.20130, 0.20867, 0.21618, 0.22382, 0.23161, 0.23952, 0.24757,
- 0.25574, 0.26404, 0.27246, 0.28100, 0.28965, 0.29841, 0.30729, 0.31626,
- 0.32533, 0.33450, 0.34376, 0.35311, 0.36253, 0.37204, 0.38161, 0.39126,
- 0.40096, 0.41072, 0.42054, 0.43040, 0.44030, 0.45023, 0.46020, 0.47019,
- 0.48020, 0.49022, 0.50025, 0.51028, 0.52031, 0.53033, 0.54033, 0.55031,
- 0.56026, 0.57019, 0.58007, 0.58991, 0.59970, 0.60944, 0.61912, 0.62873,
- 0.63827, 0.64774, 0.65713, 0.66643, 0.67564, 0.68476, 0.69377, 0.70269,
- 0.71150, 0.72019, 0.72877, 0.73723, 0.74557, 0.75378, 0.76186, 0.76981,
- 0.77762, 0.78530, 0.79283, 0.80022, 0.80747, 0.81457, 0.82151, 0.82831,
- 0.83496, 0.84145, 0.84779, 0.85398, 0.86001, 0.86588, 0.87160, 0.87716,
- 0.88257, 0.88782, 0.89291, 0.89785, 0.90264, 0.90728, 0.91176, 0.91610,
- 0.92028, 0.92432, 0.92822, 0.93197, 0.93558, 0.93906, 0.94240, 0.94560,
- 0.94867, 0.95162, 0.95444, 0.95713, 0.95971, 0.96217, 0.96451, 0.96674,
- 0.96887, 0.97089, 0.97281, 0.97463, 0.97635, 0.97799, 0.97953, 0.98099,
- 0.98236, 0.98366, 0.98488, 0.98602, 0.98710, 0.98811, 0.98905, 0.98994,
- 0.99076, 0.99153, 0.99225, 0.99291, 0.99353, 0.99411, 0.99464, 0.99513,
- 0.99558, 0.99600, 0.99639, 0.99674, 0.99706, 0.99736, 0.99763, 0.99788,
- 0.99811, 0.99831, 0.99850, 0.99867, 0.99882, 0.99895, 0.99908, 0.99919,
- 0.99929, 0.99938, 0.99946, 0.99953, 0.99959, 0.99965, 0.99969, 0.99974,
- 0.99978, 0.99981, 0.99984, 0.99986, 0.99988, 0.99990, 0.99992, 0.99993,
- 0.99994, 0.99995, 0.99996, 0.99997, 0.99998, 0.99998, 0.99998, 0.99999,
- 0.99999, 0.99999, 0.99999, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000,
- 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000
-};
-
-static const int pm128[128] =
-{
- 0, 16, 32, 48, 64, 80, 96, 112, 8, 40, 72, 104, 24, 56, 88, 120,
- 4, 20, 36, 52, 68, 84, 100, 116, 12, 28, 44, 60, 76, 92, 108, 124,
- 2, 18, 34, 50, 66, 82, 98, 114, 10, 42, 74, 106, 26, 58, 90, 122,
- 6, 22, 38, 54, 70, 86, 102, 118, 14, 46, 78, 110, 30, 62, 94, 126,
- 1, 17, 33, 49, 65, 81, 97, 113, 9, 41, 73, 105, 25, 57, 89, 121,
- 5, 21, 37, 53, 69, 85, 101, 117, 13, 29, 45, 61, 77, 93, 109, 125,
- 3, 19, 35, 51, 67, 83, 99, 115, 11, 43, 75, 107, 27, 59, 91, 123,
- 7, 23, 39, 55, 71, 87, 103, 119, 15, 31, 47, 63, 79, 95, 111, 127
-};
+#include "ac3_retables.h"
void _M( fft_64p ) ( complex_t *x );
void _M( fft_128p ) ( complex_t *a );
static void imdct512_pre_ifft_twiddle_sse (const int *pmt, complex_t *buf, float *data, float *xcos_sin_sse);
static void imdct512_post_ifft_twiddle_sse (complex_t *buf, float *xcos_sin_sse);
-static void imdct512_window_delay_sse (complex_t *buf, float *data_ptr, const float *window_prt, float *delay_prt);
-static void imdct512_window_delay_nol_sse (complex_t *buf, float *data_ptr, const float *window_prt, float *delay_prt);
+static void imdct512_window_delay_sse (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt);
+static void imdct512_window_delay_nol_sse (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt);
void _M( imdct_init ) (imdct_t * p_imdct)
@@ -260,7 +214,7 @@ static void imdct512_post_ifft_twiddle_sse (complex_t *buf, float *xcos_sin_sse)
: "a" (buf), "c" (xcos_sin_sse) );
}
-static void imdct512_window_delay_sse (complex_t *buf, float *data_ptr, const float *window_prt, float *delay_prt)
+static void imdct512_window_delay_sse (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt)
{
__asm__ __volatile__ (
"pushl %%ebp\n"
@@ -448,7 +402,7 @@ static void imdct512_window_delay_sse (complex_t *buf, float *data_ptr, const fl
::);
}
-static void imdct512_window_delay_nol_sse (complex_t *buf, float *data_ptr, const float *window_prt, float *delay_prt)
+static void imdct512_window_delay_nol_sse (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt)
{
__asm__ __volatile__ (
"pushl %%ebp\n"
diff --git a/plugins/imdct/ac3_retables.h b/plugins/imdct/ac3_retables.h
new file mode 100644
index 0000000000..50e1d1c55d
--- /dev/null
+++ b/plugins/imdct/ac3_retables.h
@@ -0,0 +1,83 @@
+/*****************************************************************************
+ * ac3_retables.h: ac3 DCT tables
+ *****************************************************************************
+ * Copyright (C) 1999, 2000 VideoLAN
+ * $Id: ac3_retables.h,v 1.1 2001/05/16 14:51:29 reno Exp $
+ *
+ * Authors: Renaud Dartus <reno@videolan.org>
+ * Aaron Holtzman <aholtzma@engr.uvic.ca>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
+ *****************************************************************************/
+
+static float window[] = {
+ 0.00014, 0.00024, 0.00037, 0.00051, 0.00067, 0.00086, 0.00107, 0.00130,
+ 0.00157, 0.00187, 0.00220, 0.00256, 0.00297, 0.00341, 0.00390, 0.00443,
+ 0.00501, 0.00564, 0.00632, 0.00706, 0.00785, 0.00871, 0.00962, 0.01061,
+ 0.01166, 0.01279, 0.01399, 0.01526, 0.01662, 0.01806, 0.01959, 0.02121,
+ 0.02292, 0.02472, 0.02662, 0.02863, 0.03073, 0.03294, 0.03527, 0.03770,
+ 0.04025, 0.04292, 0.04571, 0.04862, 0.05165, 0.05481, 0.05810, 0.06153,
+ 0.06508, 0.06878, 0.07261, 0.07658, 0.08069, 0.08495, 0.08935, 0.09389,
+ 0.09859, 0.10343, 0.10842, 0.11356, 0.11885, 0.12429, 0.12988, 0.13563,
+ 0.14152, 0.14757, 0.15376, 0.16011, 0.16661, 0.17325, 0.18005, 0.18699,
+ 0.19407, 0.20130, 0.20867, 0.21618, 0.22382, 0.23161, 0.23952, 0.24757,
+ 0.25574, 0.26404, 0.27246, 0.28100, 0.28965, 0.29841, 0.30729, 0.31626,
+ 0.32533, 0.33450, 0.34376, 0.35311, 0.36253, 0.37204, 0.38161, 0.39126,
+ 0.40096, 0.41072, 0.42054, 0.43040, 0.44030, 0.45023, 0.46020, 0.47019,
+ 0.48020, 0.49022, 0.50025, 0.51028, 0.52031, 0.53033, 0.54033, 0.55031,
+ 0.56026, 0.57019, 0.58007, 0.58991, 0.59970, 0.60944, 0.61912, 0.62873,
+ 0.63827, 0.64774, 0.65713, 0.66643, 0.67564, 0.68476, 0.69377, 0.70269,
+ 0.71150, 0.72019, 0.72877, 0.73723, 0.74557, 0.75378, 0.76186, 0.76981,
+ 0.77762, 0.78530, 0.79283, 0.80022, 0.80747, 0.81457, 0.82151, 0.82831,
+ 0.83496, 0.84145, 0.84779, 0.85398, 0.86001, 0.86588, 0.87160, 0.87716,
+ 0.88257, 0.88782, 0.89291, 0.89785, 0.90264, 0.90728, 0.91176, 0.91610,
+ 0.92028, 0.92432, 0.92822, 0.93197, 0.93558, 0.93906, 0.94240, 0.94560,
+ 0.94867, 0.95162, 0.95444, 0.95713, 0.95971, 0.96217, 0.96451, 0.96674,
+ 0.96887, 0.97089, 0.97281, 0.97463, 0.97635, 0.97799, 0.97953, 0.98099,
+ 0.98236, 0.98366, 0.98488, 0.98602, 0.98710, 0.98811, 0.98905, 0.98994,
+ 0.99076, 0.99153, 0.99225, 0.99291, 0.99353, 0.99411, 0.99464, 0.99513,
+ 0.99558, 0.99600, 0.99639, 0.99674, 0.99706, 0.99736, 0.99763, 0.99788,
+ 0.99811, 0.99831, 0.99850, 0.99867, 0.99882, 0.99895, 0.99908, 0.99919,
+ 0.99929, 0.99938, 0.99946, 0.99953, 0.99959, 0.99965, 0.99969, 0.99974,
+ 0.99978, 0.99981, 0.99984, 0.99986, 0.99988, 0.99990, 0.99992, 0.99993,
+ 0.99994, 0.99995, 0.99996, 0.99997, 0.99998, 0.99998, 0.99998, 0.99999,
+ 0.99999, 0.99999, 0.99999, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000,
+ 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000, 1.00000
+};
+
+static const int pm128[128] =
+{
+ 0, 16, 32, 48, 64, 80, 96, 112, 8, 40, 72, 104, 24, 56, 88, 120,
+ 4, 20, 36, 52, 68, 84, 100, 116, 12, 28, 44, 60, 76, 92, 108, 124,
+ 2, 18, 34, 50, 66, 82, 98, 114, 10, 42, 74, 106, 26, 58, 90, 122,
+ 6, 22, 38, 54, 70, 86, 102, 118, 14, 46, 78, 110, 30, 62, 94, 126,
+ 1, 17, 33, 49, 65, 81, 97, 113, 9, 41, 73, 105, 25, 57, 89, 121,
+ 5, 21, 37, 53, 69, 85, 101, 117, 13, 29, 45, 61, 77, 93, 109, 125,
+ 3, 19, 35, 51, 67, 83, 99, 115, 11, 43, 75, 107, 27, 59, 91, 123,
+ 7, 23, 39, 55, 71, 87, 103, 119, 15, 31, 47, 63, 79, 95, 111, 127
+};
+
+static const int pm64[64] =
+{
+ 0, 8, 16, 24, 32, 40, 48, 56,
+ 4, 20, 36, 52, 12, 28, 44, 60,
+ 2, 10, 18, 26, 34, 42, 50, 58,
+ 6, 14, 22, 30, 38, 46, 54, 62,
+ 1, 9, 17, 25, 33, 41, 49, 57,
+ 5, 21, 37, 53, 13, 29, 45, 61,
+ 3, 11, 19, 27, 35, 43, 51, 59,
+ 7, 23, 39, 55, 15, 31, 47, 63
+};
+
diff --git a/plugins/imdct/ac3_srfft_3dn.c b/plugins/imdct/ac3_srfft_3dn.c
new file mode 100644
index 0000000000..729f0981c8
--- /dev/null
+++ b/plugins/imdct/ac3_srfft_3dn.c
@@ -0,0 +1,344 @@
+/*****************************************************************************
+ * ac3_srfft_3dn.c: accelerated 3D Now! ac3 fft functions
+ *****************************************************************************
+ * Copyright (C) 1999, 2000, 2001 VideoLAN
+ * $Id: ac3_srfft_3dn.c,v 1.1 2001/05/16 14:51:29 reno Exp $
+ *
+ * Authors: Renaud Dartus <reno@videolan.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
+ *****************************************************************************/
+
+#define MODULE_NAME imdct3dn
+#include "modules_inner.h"
+
+/*****************************************************************************
+ * * Preamble
+ * *****************************************************************************/
+#include <stdio.h>
+
+#include "defs.h"
+
+#include <math.h>
+#include <stdio.h>
+
+#include "config.h"
+#include "common.h"
+#include "threads.h"
+#include "mtime.h"
+
+#include "ac3_imdct.h"
+#include "ac3_srfft.h"
+
+void hsqrt2_3dn (void);
+void C_1_3dn (void);
+static void fft_4_3dn (complex_t *x);
+static void fft_8_3dn (complex_t *x);
+static void fft_asmb_3dn (int k, complex_t *x, complex_t *wTB,
+ const complex_t *d, const complex_t *d_3);
+
+void _M( fft_64p ) ( complex_t *a )
+{
+ fft_8_3dn(&a[0]); fft_4_3dn(&a[8]); fft_4_3dn(&a[12]);
+ fft_asmb_3dn(2, &a[0], &a[8], &delta16[0], &delta16_3[0]);
+
+ fft_8_3dn(&a[16]), fft_8_3dn(&a[24]);
+ fft_asmb_3dn(4, &a[0], &a[16],&delta32[0], &delta32_3[0]);
+
+ fft_8_3dn(&a[32]); fft_4_3dn(&a[40]); fft_4_3dn(&a[44]);
+ fft_asmb_3dn(2, &a[32], &a[40], &delta16[0], &delta16_3[0]);
+
+ fft_8_3dn(&a[48]); fft_4_3dn(&a[56]); fft_4_3dn(&a[60]);
+ fft_asmb_3dn(2, &a[48], &a[56], &delta16[0], &delta16_3[0]);
+
+ fft_asmb_3dn(8, &a[0], &a[32],&delta64[0], &delta64_3[0]);
+}
+
+void _M( fft_128p ) ( complex_t *a )
+{
+ fft_8_3dn(&a[0]); fft_4_3dn(&a[8]); fft_4_3dn(&a[12]);
+ fft_asmb_3dn(2, &a[0], &a[8], &delta16[0], &delta16_3[0]);
+
+ fft_8_3dn(&a[16]), fft_8_3dn(&a[24]);
+ fft_asmb_3dn(4, &a[0], &a[16],&delta32[0], &delta32_3[0]);
+
+ fft_8_3dn(&a[32]); fft_4_3dn(&a[40]); fft_4_3dn(&a[44]);
+ fft_asmb_3dn(2, &a[32], &a[40], &delta16[0], &delta16_3[0]);
+
+ fft_8_3dn(&a[48]); fft_4_3dn(&a[56]); fft_4_3dn(&a[60]);
+ fft_asmb_3dn(2, &a[48], &a[56], &delta16[0], &delta16_3[0]);
+
+ fft_asmb_3dn(8, &a[0], &a[32],&delta64[0], &delta64_3[0]);
+
+ fft_8_3dn(&a[64]); fft_4_3dn(&a[72]); fft_4_3dn(&a[76]);
+ /* fft_16(&a[64]); */
+ fft_asmb_3dn(2, &a[64], &a[72], &delta16[0], &delta16_3[0]);
+
+ fft_8_3dn(&a[80]); fft_8_3dn(&a[88]);
+
+ /* fft_32(&a[64]); */
+ fft_asmb_3dn(4, &a[64], &a[80],&delta32[0], &delta32_3[0]);
+
+ fft_8_3dn(&a[96]); fft_4_3dn(&a[104]), fft_4_3dn(&a[108]);
+ /* fft_16(&a[96]); */
+ fft_asmb_3dn(2, &a[96], &a[104], &delta16[0], &delta16_3[0]);
+
+ fft_8_3dn(&a[112]), fft_8_3dn(&a[120]);
+ /* fft_32(&a[96]); */
+ fft_asmb_3dn(4, &a[96], &a[112], &delta32[0], &delta32_3[0]);
+
+ /* fft_128(&a[0]); */
+ fft_asmb_3dn(16, &a[0], &a[64], &delta128[0], &delta128_3[0]);
+}
+
+void hsqrt2_3dn (void)
+{
+ __asm__ (
+ ".float 0f0.707106781188\n"
+ ".float 0f0.707106781188\n"
+ ".float 0f-0.707106781188\n"
+ ".float 0f-0.707106781188\n"
+ );
+}
+
+void C_1_3dn (void)
+{
+ __asm__ (
+ ".float 0f-1.0\n"
+ ".float 0f1.0\n"
+ ".float 0f-1.0\n"
+ ".float 0f1.0\n"
+ );
+}
+
+static void fft_4_3dn (complex_t *x)
+{
+ __asm__ __volatile__ (
+ "movq (%%eax), %%mm0\n" /* x[0] */
+ "movq 8(%%eax), %%mm1\n" /* x[1] */
+ "movq 16(%%eax), %%mm2\n" /* x[2] */
+ "movq 24(%%eax), %%mm3\n" /* x[3] */
+ "movq %%mm0, %%mm4\n" /* x[1] */
+ "movq %%mm1, %%mm5\n" /* x[1] */
+ "movq %%mm0, %%mm6\n" /* x[0] */
+ "pfadd %%mm2, %%mm0\n" /* x[0] + x[2] */
+ "pfadd %%mm3, %%mm1\n" /* x[1] + x[3] */
+ "pfsub %%mm2, %%mm4\n" /* x[0] - x[2] */
+ "pfsub %%mm3, %%mm5\n" /* x[1] - x[3] */
+
+ "pfadd %%mm1, %%mm0\n" /* x[0] + x[2] + x[1] + x[3] */
+ "pfsub %%mm1, %%mm6\n" /* x[0] + x[2] - x[1] - x[3] */
+
+ "movq %%mm0, (%%eax)\n"
+ "movq %%mm6, 16(%%eax)\n"
+
+ "pxor %%mm6, %%mm6\n"
+ "movq %%mm5, %%mm2\n" /* x[1] - x[3] */
+ "movq %%mm4, %%mm3\n" /* x[0] - x[2] */
+ "pfsub %%mm5, %%mm6\n" /* x[3] - x[1] */
+
+ "punpckhdq %%mm2,%%mm2\n" /* x[1] - x[3].im */
+ "punpckldq %%mm6,%%mm6\n" /* x[3] - x[1].re */
+ "punpckhdq %%mm6,%%mm2\n" /* x[3] - x[1].re, x[1] - x[3].im */
+
+ "pfsub %%mm2, %%mm4\n" /* x0i-x2i-x3r+x1.r,x0r-x2r-x1i+x3i */
+ "pfadd %%mm3, %%mm2\n" /* x0i-x2i+x3r-x1.r, x0r-x2r+x1i-x3.i */
+
+ "movq %%mm2, 8(%%eax)\n" /* mm4_2 + mm6_1, mm4_1 + mm5_2 */
+ "movq %%mm4, 24(%%eax)\n" /* mm4_2 - mm6_1, mm4_1 - mm5_2 */
+ "femms\n"
+ : "=a" (x)
+ : "a" (x) );
+}
+
+static void fft_8_3dn (complex_t *x)
+{
+ register float wT1_r, wT1_i, wB1_r, wB1_i, wT2_r, wT2_i, wB2_r, wB2_i;
+
+ wT1_r = x[1].real;
+ wT1_i = x[1].imag;
+ wB1_r = x[3].real;
+ wB1_i = x[3].imag;
+
+ x[1] = x[2];
+ x[2] = x[4];
+ x[3] = x[6];
+ { /* fft_4 */
+ register float yt_r, yt_i, yb_r, yb_i, u_r, u_i, vi_r, vi_i;
+
+ yt_r = x[0].real;
+ yb_r = yt_r - x[2].real;
+ yt_r += x[2].real;
+
+ u_r = x[1].real;
+ vi_i = x[3].real - u_r;
+ u_r += x[3].real;
+
+ u_i = x[1].imag;
+ vi_r = u_i - x[3].imag;
+ u_i += x[3].imag;
+
+ yt_i = yt_r;
+ yt_i += u_r;
+ x[0].real = yt_i;
+ yt_r -= u_r;
+ x[2].real = yt_r;
+ yt_i = yb_r;
+ yt_i += vi_r;
+ x[1].real = yt_i;
+ yb_r -= vi_r;
+ x[3].real = yb_r;
+
+ yt_i = x[0].imag;
+ yb_i = yt_i - x[2].imag;
+ yt_i += x[2].imag;
+
+ yt_r = yt_i;
+ yt_r += u_i;
+ x[0].imag = yt_r;
+ yt_i -= u_i;
+ x[2].imag = yt_i;
+ yt_r = yb_i;
+ yt_r += vi_i;
+ x[1].imag = yt_r;
+ yb_i -= vi_i;
+ x[3].imag = yb_i;
+ }
+
+ /* x[0] x[4] */
+ wT2_r = x[5].real;
+ wT2_r += x[7].real;
+ wT2_r += wT1_r;
+ wT2_r += wB1_r;
+ wT2_i = wT2_r;
+ wT2_r += x[0].real;
+ wT2_i = x[0].real - wT2_i;
+ x[0].real = wT2_r;
+ x[4].real = wT2_i;
+
+ wT2_i = x[5].imag;
+ wT2_i += x[7].imag;
+ wT2_i += wT1_i;
+ wT2_i += wB1_i;
+ wT2_r = wT2_i;
+ wT2_r += x[0].imag;
+ wT2_i = x[0].imag - wT2_i;
+ x[0].imag = wT2_r;
+ x[4].imag = wT2_i;
+
+ /* x[2] x[6] */
+ wT2_r = x[5].imag;
+ wT2_r -= x[7].imag;
+ wT2_r += wT1_i;
+ wT2_r -= wB1_i;
+ wT2_i = wT2_r;
+ wT2_r += x[2].real;
+ wT2_i = x[2].real - wT2_i;
+ x[2].real = wT2_r;
+ x[6].real = wT2_i;
+
+ wT2_i = x[5].real;
+ wT2_i -= x[7].real;
+ wT2_i += wT1_r;
+ wT2_i -= wB1_r;
+ wT2_r = wT2_i;
+ wT2_r += x[2].imag;
+ wT2_i = x[2].imag - wT2_i;
+ x[2].imag = wT2_i;
+ x[6].imag = wT2_r;
+
+ /* x[1] x[5] */
+ wT2_r = wT1_r;
+ wT2_r += wB1_i;
+ wT2_r -= x[5].real;
+ wT2_r -= x[7].imag;
+ wT2_i = wT1_i;
+ wT2_i -= wB1_r;
+ wT2_i -= x[5].imag;
+ wT2_i += x[7].real;
+
+ wB2_r = wT2_r;
+ wB2_r += wT2_i;
+ wT2_i -= wT2_r;
+ wB2_r *= HSQRT2;
+ wT2_i *= HSQRT2;
+ wT2_r = wB2_r;
+ wB2_r += x[1].real;
+ wT2_r = x[1].real - wT2_r;
+
+ wB2_i = x[5].real;
+ x[1].real = wB2_r;
+ x[5].real = wT2_r;
+
+ wT2_r = wT2_i;
+ wT2_r += x[1].imag;
+ wT2_i = x[1].imag - wT2_i;
+ wB2_r = x[5].imag;
+ x[1].imag = wT2_r;
+ x[5].imag = wT2_i;
+
+ /* x[3] x[7] */
+ wT1_r -= wB1_i;
+ wT1_i += wB1_r;
+ wB1_r = wB2_i - x[7].imag;
+ wB1_i = wB2_r + x[7].real;
+ wT1_r -= wB1_r;
+ wT1_i -= wB1_i;
+ wB1_r = wT1_r + wT1_i;
+ wB1_r *= HSQRT2;
+ wT1_i -= wT1_r;
+ wT1_i *= HSQRT2;
+ wB2_r = x[3].real;
+ wB2_i = wB2_r + wT1_i;
+ wB2_r -= wT1_i;
+ x[3].real = wB2_i;
+ x[7].real = wB2_r;
+ wB2_i = x[3].imag;
+ wB2_r = wB2_i + wB1_r;
+ wB2_i -= wB1_r;
+ x[3].imag = wB2_i;
+ x[7].imag = wB2_r;
+}
+
+
+static void fft_asmb_3dn (int k, complex_t *x, complex_t *wTB,
+ const complex_t *d, const complex_t *d_3)
+{
+ register complex_t *x2k, *x3k, *x4k, *wB;
+ register float a_r, a_i, a1_r, a1_i, u_r, u_i, v_r, v_i;
+
+ x2k = x + 2 * k;
+ x3k = x2k + 2 * k;
+ x4k = x3k + 2 * k;
+ wB = wTB + 2 * k;
+
+ TRANSZERO(x[0],x2k[0],x3k[0],x4k[0]);
+ TRANS(x[1],x2k[1],x3k[1],x4k[1],wTB[1],wB[1],d[1],d_3[1]);
+
+ --k;
+ for(;;) {
+ TRANS(x[2],x2k[2],x3k[2],x4k[2],wTB[2],wB[2],d[2],d_3[2]);
+ TRANS(x[3],x2k[3],x3k[3],x4k[3],wTB[3],wB[3],d[3],d_3[3]);
+ if (!--k) break;
+ x += 2;
+ x2k += 2;
+ x3k += 2;
+ x4k += 2;
+ d += 2;
+ d_3 += 2;
+ wTB += 2;
+ wB += 2;
+ }
+}
diff --git a/plugins/imdct/ac3_srfft_sse.c b/plugins/imdct/ac3_srfft_sse.c
index 2de563b57b..741b2a255d 100644
--- a/plugins/imdct/ac3_srfft_sse.c
+++ b/plugins/imdct/ac3_srfft_sse.c
@@ -2,7 +2,7 @@
* ac3_srfft_sse.c: accelerated SSE ac3 fft functions
*****************************************************************************
* Copyright (C) 1999, 2000, 2001 VideoLAN
- * $Id: ac3_srfft_sse.c,v 1.1 2001/05/15 16:19:42 sam Exp $
+ * $Id: ac3_srfft_sse.c,v 1.2 2001/05/16 14:51:29 reno Exp $
*
* Authors: Renaud Dartus <reno@videolan.org>
* Aaron Holtzman <aholtzma@engr.uvic.ca>
@@ -43,8 +43,8 @@
#include "ac3_imdct.h"
#include "ac3_srfft.h"
-void hsqrt2 (void);
-void C_1 (void);
+void hsqrt2_sse (void);
+void C_1_sse (void);
static void fft_4_sse (complex_t *x);
static void fft_8_sse (complex_t *x);
static void fft_asmb_sse (int k, complex_t *x, complex_t *wTB,
@@ -104,7 +104,7 @@ void _M( fft_128p ) ( complex_t *a )
fft_asmb_sse(16, &a[0], &a[64], &delta128[0], &delta128_3[0]);
}
-void hsqrt2 (void)
+void hsqrt2_sse (void)
{
__asm__ (
".float 0f0.707106781188\n"
@@ -114,7 +114,7 @@ void hsqrt2 (void)
);
}
-void C_1 (void)
+void C_1_sse (void)
{
__asm__ (
".float 0f-1.0\n"
@@ -174,7 +174,7 @@ static void fft_8_sse (complex_t *x)
"subps %%xmm5, %%xmm7\n" /* yb = i*(x6-x2)+x0-x4 | -x6-x2+x0+x4 */
"movhps 24(%%eax), %%xmm1\n" /* x[3] | x[1] */
- "movl $hsqrt2, %%ebx\n"
+ "movl $hsqrt2_sse, %%ebx\n"
"movlps 40(%%eax), %%xmm2\n" /* x[5] */
"movhps 56(%%eax), %%xmm2\n" /* x[7] | x[5] */
"movups %%xmm1, %%xmm3\n" /* x[3] | x[1] */
@@ -191,7 +191,7 @@ static void fft_8_sse (complex_t *x)
"movlhps %%xmm6, %%xmm1\n" /* (1+i)/sqrt2 * (x[1]-x[5]) | x[1]+x[5] */
"shufps $0xe4, %%xmm6, %%xmm5\n" /* (-1-i)/sqrt2 * (x[3]-x[7]) | x[3]+x[7] */
"movups %%xmm1, %%xmm3\n" /* (1-i)/sqrt2 * (x[1]-x[5]) | x[1]+x[5] */
- "movl $C_1, %%ebx\n"
+ "movl $C_1_sse, %%ebx\n"
"addps %%xmm5, %%xmm1\n" /* u */
"subps %%xmm5, %%xmm3\n" /* v */
"movups %%xmm0, %%xmm2\n" /* yb */
@@ -258,7 +258,7 @@ static void fft_asmb_sse (int k, complex_t *x, complex_t *wTB,
"movhlps %%xmm5, %%xmm7\n" /* wT[1].im * d[1].im | wT[1].re * d[1].im */
"movlhps %%xmm6, %%xmm5\n" /* wB[1].im * d3[1].re | wB[1].re * d3[1].re | wT[1].im * d[1].re | wT[1].re * d[1].re */
"shufps $0xb1, %%xmm6, %%xmm7\n" /* wB[1].re * d3[1].im | wB[i].im * d3[1].im | wT[1].re * d[1].im | wT[1].im * d[1].im */
- "movl $C_1, %%edi\n"
+ "movl $C_1_sse, %%edi\n"
"movups (%%edi), %%xmm4\n"
"mulps %%xmm4, %%xmm7\n"
"addps %%xmm7, %%xmm5\n" /* wB[1] * d3[1] | wT[1] * d[1] */
@@ -318,7 +318,7 @@ static void fft_asmb_sse (int k, complex_t *x, complex_t *wTB,
"mulps %%xmm5, %%xmm4\n" /* wB[0].im * d3[0].im | wB[0].re * d3[0].im | wB[0].im * d3[0].re | wB[0].re * d3[0].re */
"mulps %%xmm7, %%xmm6\n" /* wB[1].im * d3[1].im | wB[1].re * d3[1].im | wB[1].im * d3[1].re | wB[1].re * d3[1].re */
"shufps $0xb1, %%xmm2, %%xmm1\n" /* d[1].im * wT[1].re | d[1].im * wT[1].im | d[0].im * wT[0].re | d[0].im * wT[0].im */
- "movl $C_1, %%edi\n"
+ "movl $C_1_sse, %%edi\n"
"movups (%%edi), %%xmm3\n" /* 1.0 | -1.0 | 1.0 | -1.0 */
"movhlps %%xmm4, %%xmm5\n" /* wB[0].im * d3[0].im | wB[0].re * d3[0].im */
diff --git a/plugins/imdct/imdct3dn.c b/plugins/imdct/imdct3dn.c
new file mode 100644
index 0000000000..7432b7164c
--- /dev/null
+++ b/plugins/imdct/imdct3dn.c
@@ -0,0 +1,152 @@
+/*****************************************************************************
+ * imdct3dn.c : accelerated 3D Now! IMDCT module
+ *****************************************************************************
+ * Copyright (C) 1999, 2000 VideoLAN
+ * $Id: imdct3dn.c,v 1.1 2001/05/16 14:51:29 reno Exp $
+ *
+ * Authors: Ga�l Hendryckx <jimmy@via.ecp.fr>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program 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 General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111, USA.
+ *****************************************************************************/
+
+#define MODULE_NAME imdct3dn
+#include "modules_inner.h"
+
+/*****************************************************************************
+ * Preamble
+ *****************************************************************************/
+#include "defs.h"
+
+#include <stdlib.h>
+
+#include "config.h"
+#include "common.h"
+#include "threads.h"
+#include "mtime.h"
+#include "tests.h"
+
+#include "ac3_imdct.h"
+#include "ac3_imdct_common.h"
+
+#include "modules.h"
+
+/*****************************************************************************
+ * Local and extern prototypes.
+ *****************************************************************************/
+static void imdct_getfunctions( function_list_t * p_function_list );
+static int imdct_Probe ( probedata_t *p_data );
+
+/*****************************************************************************
+ * Build configuration tree.
+ *****************************************************************************/
+MODULE_CONFIG_START
+ADD_WINDOW( "Configuration for IMDCT module" )
+ ADD_COMMENT( "Ha, ha -- nothing to configure yet" )
+MODULE_CONFIG_END
+
+/*****************************************************************************
+ * InitModule: get the module structure and configuration.
+ *****************************************************************************
+ * We have to fill psz_name, psz_longname and psz_version. These variables
+ * will be strdup()ed later by the main application because the module can
+ * be unloaded later to save memory, and we want to be able to access this
+ * data even after the module has been unloaded.
+ *****************************************************************************/
+MODULE_INIT
+{
+ p_module->psz_name = MODULE_STRING;
+ p_module->psz_longname = "AC3 IMDCT module";
+ p_module->psz_version = VERSION;
+
+ p_module->i_capabilities = MODULE_CAPABILITY_NULL
+ | MODULE_CAPABILITY_IMDCT;
+
+ return( 0 );
+}
+
+/*****************************************************************************
+ * ActivateModule: set the module to an usable state.
+ *****************************************************************************
+ * This function fills the capability functions and the configuration
+ * structure. Once ActivateModule() has been called, the i_usage can
+ * be set to 0 and calls to NeedModule() be made to increment it. To unload
+ * the module, one has to wait until i_usage == 0 and call DeactivateModule().
+ *****************************************************************************/
+MODULE_ACTIVATE
+{
+ p_module->p_functions = malloc( sizeof( module_functions_t ) );
+ if( p_module->p_functions == NULL )
+ {
+ return( -1 );
+ }
+
+ imdct_getfunctions( &p_module->p_functions->imdct );
+
+ p_module->p_config = p_config;
+
+ return( 0 );
+}
+
+/*****************************************************************************
+ * DeactivateModule: make sure the module can be unloaded.
+ *****************************************************************************
+ * This function must only be called when i_usage == 0. If it successfully
+ * returns, i_usage can be set to -1 and the module unloaded. Be careful to
+ * lock usage_lock during the whole process.
+ *****************************************************************************/
+MODULE_DEACTIVATE
+{
+ free( p_module->p_functions );
+
+ return( 0 );
+}
+
+/* Following functions are local */
+
+/*****************************************************************************
+ * Functions exported as capabilities. They are declared as static so that
+ * we don't pollute the namespace too much.
+ *****************************************************************************/
+static void imdct_getfunctions( function_list_t * p_function_list )
+{
+ p_function_list->pf_probe = imdct_Probe;
+#define F p_function_list->functions.imdct
+ F.pf_imdct_init = _M( imdct_init );
+ F.pf_imdct_256 = _M( imdct_do_256 );
+ F.pf_imdct_256_nol = _M( imdct_do_256_nol );
+ F.pf_imdct_512 = _M( imdct_do_512 );
+ F.pf_imdct_512_nol = _M( imdct_do_512_nol );
+#undef F
+}
+
+/*****************************************************************************
+ * imdct_Probe: returns a preference score
+ *****************************************************************************/
+static int imdct_Probe( probedata_t *p_data )
+{
+ if( !TestCPU( CPU_CAPABILITY_3DNOW ) )
+ {
+ return( 0 );
+ }
+
+ if( TestMethod( DOWNMIX_METHOD_VAR, "imdct3dn" ) )
+ {
+ return( 999 );
+ }
+
+ /* This plugin always works */
+ return( 200 );
+}
+
diff --git a/src/ac3_decoder/ac3_imdct.c b/src/ac3_decoder/ac3_imdct.c
index 639ac439f3..6ad5fbc724 100644
--- a/src/ac3_decoder/ac3_imdct.c
+++ b/src/ac3_decoder/ac3_imdct.c
@@ -2,7 +2,7 @@
* ac3_imdct.c: ac3 DCT
*****************************************************************************
* Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_imdct.c,v 1.20 2001/05/15 16:19:42 sam Exp $
+ * $Id: ac3_imdct.c,v 1.21 2001/05/16 14:51:29 reno Exp $
*
* Authors: Michel Kaempf <maxx@via.ecp.fr>
* Aaron Holtzman <aholtzma@engr.uvic.ca>
@@ -54,7 +54,7 @@ void imdct_init(imdct_t * p_imdct)
int i;
float scale = 181.019;
- p_imdct->pf_imdct_init( p_imdct );
+ p_imdct->pf_imdct_init( p_imdct );
/* More twiddle factors to turn IFFT into IMDCT */
for (i=0; i < 64; i++) {
--
2.25.4