filter-i386.h

/*
 *      filter-i386.h -- optimized filter routines
 *
 *      Copyright (C) 1996  
 *          Thomas Sailer (sailer@ife.ee.ethz.ch, hb9jnx@hb9w.che.eu)
 *
 *      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., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/* ---------------------------------------------------------------------- */

#ifndef _FILTER_I386_H
#define _FILTER_I386_H

/* ---------------------------------------------------------------------- */

#define __HAVE_ARCH_MAC
#define mac(a,b,size) \
(__builtin_constant_p(size) ? __mac_c((a),(b),(size)) : __mac_g((a),(b),(size)))

#include <stdio.h>

static inline float __mac_g(const float *a, const float *b, unsigned int size)
{
	float sum = 0;
	unsigned int i;

	for (i = 0; i < size; i++)
		sum += (*a++) * (*b++);
	return sum;
}

static inline float __mac_c(const float *a, const float *b, unsigned int size)
{
	float f;

	/*
	 * inspired from Phil Karn, KA9Q's home page
	 */
	switch (size) {
	case 9:
		asm volatile ("flds (%1);\n\t"
			      "fmuls (%2);\n\t"
			      "flds 4(%1);\n\t"
			      "fmuls 4(%2);\n\t"
			      "flds 8(%1);\n\t"
			      "fmuls 8(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 12(%1);\n\t"
			      "fmuls 12(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 16(%1);\n\t"
			      "fmuls 16(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 20(%1);\n\t"
			      "fmuls 20(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 24(%1);\n\t"
			      "fmuls 24(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 28(%1);\n\t"
			      "fmuls 28(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 32(%1);\n\t"
			      "fmuls 32(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "faddp;\n\t" :
			      "=t" (f) :
			      "r" (a),
			      "r" (b) : "memory");
		return f;

	case 18:
		asm volatile ("flds (%1);\n\t"
			      "fmuls (%2);\n\t"
			      "flds 4(%1);\n\t"
			      "fmuls 4(%2);\n\t"
			      "flds 8(%1);\n\t"
			      "fmuls 8(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 12(%1);\n\t"
			      "fmuls 12(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 16(%1);\n\t"
			      "fmuls 16(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 20(%1);\n\t"
			      "fmuls 20(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 24(%1);\n\t"
			      "fmuls 24(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 28(%1);\n\t"
			      "fmuls 28(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 32(%1);\n\t"
			      "fmuls 32(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 36(%1);\n\t"
			      "fmuls 36(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 40(%1);\n\t"
			      "fmuls 40(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 44(%1);\n\t"
			      "fmuls 44(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 48(%1);\n\t"
			      "fmuls 48(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 52(%1);\n\t"
			      "fmuls 52(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 56(%1);\n\t"
			      "fmuls 56(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 60(%1);\n\t"
			      "fmuls 60(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 64(%1);\n\t"
			      "fmuls 64(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 68(%1);\n\t"
			      "fmuls 68(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "faddp;\n\t" :
			      "=t" (f) :
			      "r" (a),
			      "r" (b) : "memory");
		return f;

	case 24:
		asm volatile ("flds (%1);\n\t"
			      "fmuls (%2);\n\t"
			      "flds 4(%1);\n\t"
			      "fmuls 4(%2);\n\t"
			      "flds 8(%1);\n\t"
			      "fmuls 8(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 12(%1);\n\t"
			      "fmuls 12(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 16(%1);\n\t"
			      "fmuls 16(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 20(%1);\n\t"
			      "fmuls 20(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 24(%1);\n\t"
			      "fmuls 24(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 28(%1);\n\t"
			      "fmuls 28(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 32(%1);\n\t"
			      "fmuls 32(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 36(%1);\n\t"
			      "fmuls 36(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 40(%1);\n\t"
			      "fmuls 40(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 44(%1);\n\t"
			      "fmuls 44(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 48(%1);\n\t"
			      "fmuls 48(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 52(%1);\n\t"
			      "fmuls 52(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 56(%1);\n\t"
			      "fmuls 56(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 60(%1);\n\t"
			      "fmuls 60(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 64(%1);\n\t"
			      "fmuls 64(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 68(%1);\n\t"
			      "fmuls 68(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 72(%1);\n\t"
			      "fmuls 72(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 76(%1);\n\t"
			      "fmuls 76(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 80(%1);\n\t"
			      "fmuls 80(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 84(%1);\n\t"
			      "fmuls 84(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 88(%1);\n\t"
			      "fmuls 88(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "flds 92(%1);\n\t"
			      "fmuls 92(%2);\n\t"
			      "fxch %%st(2);\n\t"
			      "faddp;\n\t"
			      "faddp;\n\t" :
			      "=t" (f) :
			      "r" (a),
			      "r" (b) : "memory");
		return f;

	default:
		printf("Warning: optimize __mac_c(..., ..., %d)\n", size);
		return __mac_g(a, b, size);
	}
}

/* ---------------------------------------------------------------------- */
#endif /* _FILTER_I386_H */