* ac3_imdct_3dn.c: accelerated 3D Now! ac3 DCT
*****************************************************************************
* Copyright (C) 1999, 2000 VideoLAN
- * $Id: ac3_imdct_3dn.c,v 1.4 2001/06/03 12:47:21 sam Exp $
+ * $Id: ac3_imdct_3dn.c,v 1.12 2002/07/31 20:56:51 sam Exp $
*
* Authors: Renaud Dartus <reno@videolan.org>
*
* 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 <vlc/vlc.h>
#include "ac3_imdct.h"
#include "ac3_imdct_common.h"
# define M_PI 3.14159265358979323846
#endif
-void _M( fft_64p ) ( complex_t *x );
-void _M( fft_128p ) ( complex_t *a );
+void E_( fft_64p ) ( complex_t *x );
+void E_( 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_nol_3dn (complex_t *buf, float *data_ptr, float *window_prt, float *delay_prt);
-void _M( imdct_init ) (imdct_t * p_imdct)
+void E_( imdct_init ) (imdct_t * p_imdct)
{
int i;
float scale = 181.019;
}
}
-void _M( imdct_do_512 ) (imdct_t * p_imdct, float data[], float delay[])
+void E_( 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);
+ E_( 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[])
+void E_( 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);
+ E_( 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"
+ ".align 16\n"
"pushl %%ebx\n"
- "pushl %%ecx\n"
- "pushl %%edx\n"
- "pushl %%edi\n"
"pushl %%esi\n"
+
+ "movl $128, %%ebx\n" /* loop counter */
- "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"
+ ".align 16\n"
+"0:\n"
"movl (%%eax), %%esi\n"
"movd (%%ecx, %%esi, 8), %%mm1\n" /* 2j */
"punpckldq %%mm1, %%mm1\n" /* 2j | 2j */
"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"
+ "addl $8, %%edi\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"
+ "movq %%mm0, -8(%%edi)\n"
+ "decl %%ebx\n"
+ "jnz 0b\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"
- ::);
+ : "=D" (buf)
+ : "a" (pmt), "c" (data), "d" (xcos_sin_sse), "D" (buf));
}
static void imdct512_post_ifft_twiddle_3dn (complex_t *buf, float *xcos_sin_sse)
{
__asm__ __volatile__ (
+ ".align 16\n"
"pushl %%ebx\n"
"movl $64, %%ebx\n" /* loop counter */
-".loop1:\n"
+ ".align 16\n"
+"0:\n"
"movq (%%eax), %%mm0\n" /* im0 | re0 */
"movq %%mm0, %%mm1\n" /* im0 | re0 */
"punpckldq %%mm0, %%mm0\n" /* re0 | re0 */
"addl $32, %%ecx\n"
"addl $16, %%eax\n"
"decl %%ebx\n"
- "jnz .loop1\n"
+ "jnz 0b\n"
"popl %%ebx\n"
"femms\n"
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"
+ ".align 16\n"
"pushl %%eax\n"
"pushl %%ebx\n"
"pushl %%ecx\n"
"pushl %%edx\n"
- "pushl %%esi\n"
"pushl %%edi\n"
+ "pushl %%esi\n"
+ "pushl %%ebp\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 */
+ "movl %%esi, %%ebp\n" /* buf */
+ "movl $32, %%ebx\n" /* loop count */
+ "leal 516(%%ebp), %%esi\n" /* buf[64].im */
+ "leal 504(%%ebp), %%edi\n" /* buf[63].re */
+
+ ".align 16\n"
".first_128_samples:\n"
"movd (%%esi), %%mm0\n" /* im0 */
"movd 8(%%esi), %%mm2\n" /* 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 */
+ "movq (%%ecx), %%mm2\n" /* d1 | d0 */
+ "movq 8(%%ecx), %%mm3\n" /* d3 | d2 */
"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, %%ecx\n"
"addl $16, %%esi\n"
"addl $16, %%eax\n"
"addl $-16, %%edi\n"
- "decl %%ecx\n"
+ "decl %%ebx\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 */
+ "movl %%ebp, %%esi\n" /* buf[0].re */
+ "movl $32, %%ebx\n" /* loop count */
+ "leal 1020(%%ebp), %%edi\n" /* buf[127].im */
+ ".align 16\n"
".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 */
"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 */
+ "movq (%%ecx), %%mm2\n" /* d1 | d0 */
+ "movq 8(%%ecx), %%mm3\n" /* d3 | d2 */
"addl $16, %%esi\n"
"addl $16, %%edx\n"
"addl $16, %%eax\n"
- "addl $16, %%ebx\n"
- "decl %%ecx\n"
+ "addl $16, %%ecx\n"
+ "decl %%ebx\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 */
+ "leal 512(%%ebp), %%esi\n" /* buf[64].re */
+ "leal 508(%%ebp), %%edi\n" /* buf[63].im */
+ "movl $32, %%ebx\n" /* loop count */
+ "addl $-1024, %%ecx\n" /* delay */
+ ".align 16\n"
".first_128_delay:\n"
"movd (%%esi), %%mm0\n" /* re0 */
"movd 8(%%esi), %%mm2\n" /* re1 */
"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"
+ "movq %%mm0, (%%ecx)\n"
+ "movq %%mm1, 8(%%ecx)\n"
"addl $16, %%esi\n"
"addl $-16, %%edi\n"
- "addl $16, %%eax\n"
- "decl %%ecx\n"
+ "addl $16, %%ecx\n"
+ "decl %%ebx\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 */
+ "leal 4(%%ebp), %%esi\n" /* buf[0].im */
+ "leal 1016(%%ebp), %%edi\n" /* buf[127].re */
+ "movl $32, %%ebx\n" /* loop count */
+ ".align 16\n"
".second_128_delay:\n"
"movd (%%esi), %%mm0\n" /* im0 */
"movd 8(%%esi), %%mm2\n" /* im1 */
"pfmul %%mm2, %%mm3\n" /* -w3*re1 | w2*im1 */
- "movq %%mm1, (%%eax)\n"
- "movq %%mm3, 8(%%eax)\n"
+ "movq %%mm1, (%%ecx)\n"
+ "movq %%mm3, 8(%%ecx)\n"
"addl $16, %%esi\n"
"addl $-16, %%edi\n"
- "addl $16, %%eax\n"
- "decl %%ecx\n"
+ "addl $16, %%ecx\n"
+ "decl %%ebx\n"
"jnz .second_128_delay\n"
- "popl %%edi\n"
+ "popl %%ebp\n"
"popl %%esi\n"
+ "popl %%edi\n"
"popl %%edx\n"
"popl %%ecx\n"
"popl %%ebx\n"
"popl %%eax\n"
- "leave\n"
"femms\n"
- ::);
+ : "=S" (buf), "=a" (data_ptr), "=c" (delay_prt), "=d" (window_prt)
+ : "S" (buf), "a" (data_ptr), "c" (delay_prt), "d" (window_prt));
}
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"
-
+ ".align 16\n"
+
"pushl %%eax\n"
"pushl %%ebx\n"
"pushl %%ecx\n"
"pushl %%edx\n"
- "pushl %%esi\n"
"pushl %%edi\n"
+ "pushl %%esi\n"
+ "pushl %%ebp\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 */
+ "movl %%esi, %%ebp\n" /* buf */
+ "movl $32, %%ebx\n" /* loop count */
+ "leal 516(%%ebp), %%esi\n" /* buf[64].im */
+ "leal 504(%%ebp), %%edi\n" /* buf[63].re */
+ ".align 16\n"
".first_128_samples2:\n"
"movd (%%esi), %%mm0\n" /* im0 */
"movd 8(%%esi), %%mm2\n" /* im1 */
"addl $16, %%edx\n"
"movq %%mm0, (%%eax)\n"
"movq %%mm1, 8(%%eax)\n"
- "addl $16, %%ebx\n"
+ "addl $16, %%ecx\n"
"addl $16, %%esi\n"
"addl $16, %%eax\n"
"addl $-16, %%edi\n"
- "decl %%ecx\n"
+ "decl %%ebx\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 */
+ "movl %%ebp, %%esi\n" /* buf[0].re */
+ "movl $32, %%ebx\n" /* loop count */
+ "leal 1020(%%ebp), %%edi\n" /* buf[127].im */
+ ".align 16\n"
".second_128_samples2:\n"
"movd (%%esi), %%mm0\n" /* buf[i].re */
"movd 8(%%esi), %%mm2\n" /* re1 */
"addl $16, %%edx\n"
"addl $16, %%eax\n"
- "addl $16, %%ebx\n"
- "decl %%ecx\n"
+ "addl $16, %%ecx\n"
+ "decl %%ebx\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 */
+ "leal 512(%%ebp), %%esi\n" /* buf[64].re */
+ "leal 508(%%ebp), %%edi\n" /* buf[63].im */
+ "movl $32, %%ebx\n" /* loop count */
+ "addl $-1024, %%ecx\n" /* delay */
+ ".align 16\n"
".first_128_delays:\n"
"movd (%%esi), %%mm0\n" /* re0 */
"movd 8(%%esi), %%mm2\n" /* re1 */
"pfmul %%mm5, %%mm1\n" /* w3*im1 | -w2*re1 */
- "movq %%mm0, (%%eax)\n"
- "movq %%mm1, 8(%%eax)\n"
+ "movq %%mm0, (%%ecx)\n"
+ "movq %%mm1, 8(%%ecx)\n"
"addl $16, %%esi\n"
"addl $-16, %%edi\n"
- "addl $16, %%eax\n"
- "decl %%ecx\n"
+ "addl $16, %%ecx\n"
+ "decl %%ebx\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 */
+ "leal 4(%%ebp), %%esi\n" /* buf[0].im */
+ "leal 1016(%%ebp), %%edi\n" /* buf[127].re */
+ "movl $32, %%ebx\n" /* loop count */
+ ".align 16\n"
".second_128_delays:\n"
"movd (%%esi), %%mm0\n" /* im0 */
"movd 8(%%esi), %%mm2\n" /* im1 */
"pfmul %%mm2, %%mm3\n" /* -w3*re1 | w2*im1 */
- "movq %%mm1, (%%eax)\n"
- "movq %%mm3, 8(%%eax)\n"
+ "movq %%mm1, (%%ecx)\n"
+ "movq %%mm3, 8(%%ecx)\n"
"addl $16, %%esi\n"
"addl $-16, %%edi\n"
- "addl $16, %%eax\n"
- "decl %%ecx\n"
+ "addl $16, %%ecx\n"
+ "decl %%ebx\n"
"jnz .second_128_delays\n"
- "popl %%edi\n"
+ "popl %%ebp\n"
"popl %%esi\n"
+ "popl %%edi\n"
"popl %%edx\n"
"popl %%ecx\n"
"popl %%ebx\n"
"popl %%eax\n"
- "leave\n"
"femms\n"
- ::);
+ : "=S" (buf), "=a" (data_ptr), "=c" (delay_prt), "=d" (window_prt)
+ : "S" (buf), "a" (data_ptr), "c" (delay_prt), "d" (window_prt));
}
projects / vlc / blobdiff