comparison liba52/liba52_changes.diff @ 19249:9960f9ef96fd

cosmetics: Make patch apply cleanly.

19248:dd30cb46819d

--- liba52-0.7.4/a52.h	2006-06-12 15:04:57.000000000 +0200
+++ liba52/a52.h	2006-06-05 02:23:02.000000000 +0200
@@ -59,4 +63,9 @@
 int a52_block (a52_state_t * state);
 void a52_free (a52_state_t * state);
 

+extern int (* a52_resample) (float * _f, int16_t * s16);
+
+uint16_t crc16_block(uint8_t *data,uint32_t num_bytes);
+
 #endif /* A52_H */
--- liba52-0.7.4/a52_internal.h	2006-06-12 15:05:07.000000000 +0200
+++ liba52/a52_internal.h	2006-06-05 02:23:02.000000000 +0200
@@ -103,18 +107,34 @@
 #define DELTA_BIT_NONE (2)
 #define DELTA_BIT_RESERVED (3)
 

 void a52_imdct_init (uint32_t mm_accel);
 void a52_imdct_256 (sample_t * data, sample_t * delay, sample_t bias);
-void a52_imdct_512 (sample_t * data, sample_t * delay, sample_t bias);
+extern void (*a52_imdct_512) (sample_t * data, sample_t * delay, sample_t bias);
+void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias);
--- liba52-0.7.4/bitstream.c	2006-06-12 15:05:07.000000000 +0200
+++ liba52/bitstream.c	2006-06-05 02:23:02.000000000 +0200
@@ -31,6 +35,10 @@
 
 #define BUFFER_SIZE 4096
 

+    indx=0;
+#endif
     bitstream_get (state, align * 8);
 }
 
--- liba52-0.7.4/bitstream.h	2006-06-12 15:05:07.000000000 +0200
+++ liba52/bitstream.h	2006-06-05 02:23:02.000000000 +0200
@@ -21,6 +25,48 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 

     }
 
     return a52_bitstream_get_bh_2 (state, num_bits);
+#endif
 }
--- liba52-0.7.4/downmix.c	2006-06-12 15:17:53.000000000 +0200
+++ liba52/downmix.c	2006-06-05 02:23:02.000000000 +0200
@@ -23,18 +23,47 @@
  * 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  02111-1307  USA
+ *
+ * SSE optimizations from Michael Niedermayer ([e-mail])

+}
+ 
 int a52_downmix_init (int input, int flags, sample_t * level,
 		      sample_t clev, sample_t slev)
 {
@@ -451,7 +480,7 @@
 	samples[i] = 0;
 }
 
-void a52_downmix (sample_t * samples, int acmod, int output, sample_t bias,
+void downmix_C (sample_t * samples, int acmod, int output, sample_t bias,
 		  sample_t clev, sample_t slev)
 {
     switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
@@ -563,7 +592,7 @@
 	break;
 
     case CONVERT (A52_3F2R, A52_2F1R):
-	mix3to2 (samples, bias);
+	mix3to2 (samples, bias); //FIXME possible bug? (output doesnt seem to be used)
 	move2to1 (samples + 768, samples + 512, bias);
 	break;
 
@@ -587,12 +616,12 @@
 	break;
 
     case CONVERT (A52_3F1R, A52_3F2R):
-	memcpy (samples + 1027, samples + 768, 256 * sizeof (sample_t));
+	memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));

-void a52_upmix (sample_t * samples, int acmod, int output)
+void upmix_C (sample_t * samples, int acmod, int output)
 {
     switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
 
@@ -657,3 +686,1137 @@
 	goto mix_31to21;
     }
 }
+
+#if defined(ARCH_X86) || defined(ARCH_X86_64)

+    }
+    __asm __volatile("femms":::"memory");
+}
+
+#endif // ARCH_X86 || ARCH_X86_64
--- liba52-0.7.4/imdct.c	2006-06-12 15:18:27.000000000 +0200
+++ liba52/imdct.c	2006-06-12 19:18:39.000000000 +0200
@@ -26,9 +26,15 @@
  * 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  02111-1307  USA
+ *
+ * SSE optimizations from Michael Niedermayer ([e-mail])

 #include "config.h"
+#include "asmalign.h"
 
 #include <math.h>
 #include <stdio.h>
@@ -43,12 +49,49 @@
 #include "a52.h"
 #include "a52_internal.h"
 #include "mm_accel.h"
+#include "mangle.h"
+

+	0x0f, 0x4f, 0x2f, 0x6f, 0x1f, 0x5f, 0x3f, 0x7f};
+
 static uint8_t fftorder[] = {
       0,128, 64,192, 32,160,224, 96, 16,144, 80,208,240,112, 48,176,
       8,136, 72,200, 40,168,232,104,248,120, 56,184, 24,152,216, 88,
@@ -60,6 +103,40 @@
       6,134, 70,198, 38,166,230,102,246,118, 54,182, 22,150,214, 86
 };
 
+static complex_t __attribute__((aligned(16))) buf[128];
+

+#endif
+
 /* Root values for IFFT */
 static sample_t roots16[3];
 static sample_t roots32[7];
@@ -245,7 +322,7 @@
     ifft_pass (buf, roots128 - 32, 32);
 }
 
-void a52_imdct_512 (sample_t * data, sample_t * delay, sample_t bias)
+void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias)
 {
     int i, k;
     sample_t t_r, t_i, a_r, a_i, b_r, b_i, w_1, w_2;
@@ -289,6 +366,714 @@
     }
 }
 
+#ifdef HAVE_ALTIVEC
+

+#endif // ARCH_X86 || ARCH_X86_64
+
 void a52_imdct_256(sample_t * data, sample_t * delay, sample_t bias)
 {
     int i, k;
@@ -368,7 +1153,7 @@
 
 void a52_imdct_init (uint32_t mm_accel)
 {
-    int i, k;
+    int i, j, k;
     double sum;
 
     /* compute imdct window - kaiser-bessel derived window, alpha = 5.0 */
@@ -420,6 +1205,99 @@
 	post2[i].real = cos ((M_PI / 128) * (i + 0.5));
 	post2[i].imag = sin ((M_PI / 128) * (i + 0.5));
     }
+    for (i = 0; i < 128; i++) {
+	xcos1[i] = -cos ((M_PI / 2048) * (8 * i + 1));

+	else
+#endif
 
 #ifdef LIBA52_DJBFFT
     if (mm_accel & MM_ACCEL_DJBFFT) {
@@ -430,7 +1308,5 @@
 #endif
     {
 	fprintf (stderr, "No accelerated IMDCT transform found\n");
-	ifft128 = ifft128_c;
-	ifft64 = ifft64_c;
     }
 }
--- liba52-0.7.4/mm_accel.h	2006-06-12 15:05:00.000000000 +0200
+++ liba52/mm_accel.h	2006-06-05 02:23:04.000000000 +0200
@@ -30,7 +34,12 @@
 /* x86 accelerations */
 #define MM_ACCEL_X86_MMX	0x80000000
 #define MM_ACCEL_X86_3DNOW	0x40000000

+/* PPC accelerations */
+#define MM_ACCEL_PPC_ALTIVEC	0x00010000
 
 uint32_t mm_accel (void);
 
--- liba52-0.7.4/parse.c	2006-06-12 15:05:07.000000000 +0200
+++ liba52/parse.c	2006-06-12 14:51:33.000000000 +0200
@@ -24,6 +28,7 @@
 #include "config.h"
 
 #include <stdlib.h>

19249:9960f9ef96fd

--- include/a52.h	2006-06-12 15:04:57.000000000 +0200
+++ liba52/a52.h	2006-06-05 02:23:02.000000000 +0200
@@ -59,4 +63,9 @@
 int a52_block (a52_state_t * state);
 void a52_free (a52_state_t * state);
 

+extern int (* a52_resample) (float * _f, int16_t * s16);
+
+uint16_t crc16_block(uint8_t *data,uint32_t num_bytes);
+
 #endif /* A52_H */
--- liba52/a52_internal.h	2006-06-12 15:05:07.000000000 +0200
+++ liba52/a52_internal.h	2006-06-05 02:23:02.000000000 +0200
@@ -103,18 +107,34 @@
 #define DELTA_BIT_NONE (2)
 #define DELTA_BIT_RESERVED (3)
 

 void a52_imdct_init (uint32_t mm_accel);
 void a52_imdct_256 (sample_t * data, sample_t * delay, sample_t bias);
-void a52_imdct_512 (sample_t * data, sample_t * delay, sample_t bias);
+extern void (*a52_imdct_512) (sample_t * data, sample_t * delay, sample_t bias);
+void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias);
--- liba52/bitstream.c	2006-06-12 15:05:07.000000000 +0200
+++ liba52/bitstream.c	2006-06-05 02:23:02.000000000 +0200
@@ -31,6 +35,10 @@
 
 #define BUFFER_SIZE 4096
 

+    indx=0;
+#endif
     bitstream_get (state, align * 8);
 }
 
--- liba52/bitstream.h	2006-06-12 15:05:07.000000000 +0200
+++ liba52/bitstream.h	2006-06-05 02:23:02.000000000 +0200
@@ -21,6 +25,48 @@
  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  */
 

     }
 
     return a52_bitstream_get_bh_2 (state, num_bits);
+#endif
 }
--- liba52/downmix.c	2006-06-12 15:17:53.000000000 +0200
+++ liba52/downmix.c	2006-06-05 02:23:02.000000000 +0200
@@ -19,18 +23,47 @@
  * 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  02111-1307  USA
+ *
+ * SSE optimizations from Michael Niedermayer ([e-mail])

+}
+ 
 int a52_downmix_init (int input, int flags, sample_t * level,
 		      sample_t clev, sample_t slev)
 {
@@ -447,7 +480,7 @@
 	samples[i] = 0;
 }
 
-void a52_downmix (sample_t * samples, int acmod, int output, sample_t bias,
+void downmix_C (sample_t * samples, int acmod, int output, sample_t bias,
 		  sample_t clev, sample_t slev)
 {
     switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
@@ -559,7 +592,7 @@
 	break;
 
     case CONVERT (A52_3F2R, A52_2F1R):
-	mix3to2 (samples, bias);
+	mix3to2 (samples, bias); //FIXME possible bug? (output doesnt seem to be used)
 	move2to1 (samples + 768, samples + 512, bias);
 	break;
 
@@ -583,12 +616,12 @@
 	break;
 
     case CONVERT (A52_3F1R, A52_3F2R):
-	memcpy (samples + 1027, samples + 768, 256 * sizeof (sample_t));
+	memcpy (samples + 1024, samples + 768, 256 * sizeof (sample_t));

-void a52_upmix (sample_t * samples, int acmod, int output)
+void upmix_C (sample_t * samples, int acmod, int output)
 {
     switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) {
 
@@ -653,3 +686,1137 @@
 	goto mix_31to21;
     }
 }
+
+#if defined(ARCH_X86) || defined(ARCH_X86_64)

+    }
+    __asm __volatile("femms":::"memory");
+}
+
+#endif // ARCH_X86 || ARCH_X86_64
--- liba52/imdct.c	2006-06-12 15:18:27.000000000 +0200
+++ liba52/imdct.c	2006-06-12 19:18:39.000000000 +0200
@@ -22,9 +26,15 @@
  * 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  02111-1307  USA
+ *
+ * SSE optimizations from Michael Niedermayer ([e-mail])

 #include "config.h"
+#include "asmalign.h"
 
 #include <math.h>
 #include <stdio.h>
@@ -39,12 +49,49 @@
 #include "a52.h"
 #include "a52_internal.h"
 #include "mm_accel.h"
+#include "mangle.h"
+

+	0x0f, 0x4f, 0x2f, 0x6f, 0x1f, 0x5f, 0x3f, 0x7f};
+
 static uint8_t fftorder[] = {
       0,128, 64,192, 32,160,224, 96, 16,144, 80,208,240,112, 48,176,
       8,136, 72,200, 40,168,232,104,248,120, 56,184, 24,152,216, 88,
@@ -56,6 +103,40 @@
       6,134, 70,198, 38,166,230,102,246,118, 54,182, 22,150,214, 86
 };
 
+static complex_t __attribute__((aligned(16))) buf[128];
+

+#endif
+
 /* Root values for IFFT */
 static sample_t roots16[3];
 static sample_t roots32[7];
@@ -241,7 +322,7 @@
     ifft_pass (buf, roots128 - 32, 32);
 }
 
-void a52_imdct_512 (sample_t * data, sample_t * delay, sample_t bias)
+void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias)
 {
     int i, k;
     sample_t t_r, t_i, a_r, a_i, b_r, b_i, w_1, w_2;
@@ -285,6 +366,714 @@
     }
 }
 
+#ifdef HAVE_ALTIVEC
+

+#endif // ARCH_X86 || ARCH_X86_64
+
 void a52_imdct_256(sample_t * data, sample_t * delay, sample_t bias)
 {
     int i, k;
@@ -364,7 +1153,7 @@
 
 void a52_imdct_init (uint32_t mm_accel)
 {
-    int i, k;
+    int i, j, k;
     double sum;
 
     /* compute imdct window - kaiser-bessel derived window, alpha = 5.0 */
@@ -416,6 +1205,99 @@
 	post2[i].real = cos ((M_PI / 128) * (i + 0.5));
 	post2[i].imag = sin ((M_PI / 128) * (i + 0.5));
     }
+    for (i = 0; i < 128; i++) {
+	xcos1[i] = -cos ((M_PI / 2048) * (8 * i + 1));

+	else
+#endif
 
 #ifdef LIBA52_DJBFFT
     if (mm_accel & MM_ACCEL_DJBFFT) {
@@ -426,7 +1308,5 @@
 #endif
     {
 	fprintf (stderr, "No accelerated IMDCT transform found\n");
-	ifft128 = ifft128_c;
-	ifft64 = ifft64_c;
     }
 }
--- include/mm_accel.h	2006-06-12 15:05:00.000000000 +0200
+++ liba52/mm_accel.h	2006-06-05 02:23:04.000000000 +0200
@@ -30,7 +34,12 @@
 /* x86 accelerations */
 #define MM_ACCEL_X86_MMX	0x80000000
 #define MM_ACCEL_X86_3DNOW	0x40000000

+/* PPC accelerations */
+#define MM_ACCEL_PPC_ALTIVEC	0x00010000
 
 uint32_t mm_accel (void);
 
--- liba52/parse.c	2006-06-12 15:05:07.000000000 +0200
+++ liba52/parse.c	2006-06-12 14:51:33.000000000 +0200
@@ -24,6 +28,7 @@
 #include "config.h"
 
 #include <stdlib.h>