Update.

1998-04-03 23:38  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/libm-ieee754/e_acos.c: Optimize by splitting large
	expressions and using array variables.
	* sysdeps/libm-ieee754/e_asin.c: Likewise.
	* sysdeps/libm-ieee754/e_j0.c: Likewise.
	* sysdeps/libm-ieee754/e_j1.c: Likewise.
	* sysdeps/libm-ieee754/e_log.c: Likewise.
	* sysdeps/libm-ieee754/e_pow.c: Likewise.
	* sysdeps/libm-ieee754/k_cos.c: Likewise.
	* sysdeps/libm-ieee754/k_sin.c: Likewise.
	* sysdeps/libm-ieee754/k_tan.c: Likewise.
	* sysdeps/libm-ieee754/s_atan.c: Likewise.
	* sysdeps/libm-ieee754/s_erf.c: Likewise.
	* sysdeps/libm-ieee754/s_expm1.c: Likewise.
	* sysdeps/libm-ieee754/s_log1p.c: Likewise.
	Patch by Naohiko Shimizu <nshimizu@et.u-tokai.ac.jp>.

1 files changed, 33 insertions, 22 deletions

diff --git a/sysdeps/libm-ieee754/k_cos.c b/sysdeps/libm-ieee754/k_cos.c
index d8740b3508..7e38ef7915 100644
--- a/sysdeps/libm-ieee754/k_cos.c
+++ b/sysdeps/libm-ieee754/k_cos.c
@@ -5,10 +5,13 @@
  *
  * Developed at SunPro, a Sun Microsystems, Inc. business.
  * Permission to use, copy, modify, and distribute this
- * software is freely granted, provided that this notice 
+ * software is freely granted, provided that this notice
  * is preserved.
  * ====================================================
  */
+/* Modified by Naohiko Shimizu/Tokai University, Japan 1997/08/25,
+   for performance improvement on pipelined processors.
+*/
 
 #if defined(LIBM_SCCS) && !defined(lint)
 static char rcsid[] = "$NetBSD: k_cos.c,v 1.8 1995/05/10 20:46:22 jtc Exp $";
@@ -18,7 +21,7 @@ static char rcsid[] = "$NetBSD: k_cos.c,v 1.8 1995/05/10 20:46:22 jtc Exp $";
  * __kernel_cos( x,  y )
  * kernel cos function on [-pi/4, pi/4], pi/4 ~ 0.785398164
  * Input x is assumed to be bounded by ~pi/4 in magnitude.
- * Input y is the tail of x. 
+ * Input y is the tail of x.
  *
  * Algorithm
  *	1. Since cos(-x) = cos(x), we need only to consider positive x.
@@ -28,15 +31,15 @@ static char rcsid[] = "$NetBSD: k_cos.c,v 1.8 1995/05/10 20:46:22 jtc Exp $";
  *		  	                 4            14
  *	   	cos(x) ~ 1 - x*x/2 + C1*x + ... + C6*x
  *	   where the remez error is
- *	
+ *
  * 	|              2     4     6     8     10    12     14 |     -58
  * 	|cos(x)-(1-.5*x +C1*x +C2*x +C3*x +C4*x +C5*x  +C6*x  )| <= 2
- * 	|    					               | 
- * 
- * 	               4     6     8     10    12     14 
+ * 	|    					               |
+ *
+ * 	               4     6     8     10    12     14
  *	4. let r = C1*x +C2*x +C3*x +C4*x +C5*x  +C6*x  , then
  *	       cos(x) = 1 - x*x/2 + r
- *	   since cos(x+y) ~ cos(x) - sin(x)*y 
+ *	   since cos(x+y) ~ cos(x) - sin(x)*y
  *			  ~ cos(x) - x*y,
  *	   a correction term is necessary in cos(x) and hence
  *		cos(x+y) = 1 - (x*x/2 - (r - x*y))
@@ -53,17 +56,18 @@ static char rcsid[] = "$NetBSD: k_cos.c,v 1.8 1995/05/10 20:46:22 jtc Exp $";
 #include "math_private.h"
 
 #ifdef __STDC__
-static const double 
+static const double
 #else
-static double 
+static double
 #endif
-one =  1.00000000000000000000e+00, /* 0x3FF00000, 0x00000000 */
-C1  =  4.16666666666666019037e-02, /* 0x3FA55555, 0x5555554C */
-C2  = -1.38888888888741095749e-03, /* 0xBF56C16C, 0x16C15177 */
-C3  =  2.48015872894767294178e-05, /* 0x3EFA01A0, 0x19CB1590 */
-C4  = -2.75573143513906633035e-07, /* 0xBE927E4F, 0x809C52AD */
-C5  =  2.08757232129817482790e-09, /* 0x3E21EE9E, 0xBDB4B1C4 */
-C6  = -1.13596475577881948265e-11; /* 0xBDA8FAE9, 0xBE8838D4 */
+C[] = {
+  1.00000000000000000000e+00, /* 0x3FF00000, 0x00000000 */
+  4.16666666666666019037e-02, /* 0x3FA55555, 0x5555554C */
+ -1.38888888888741095749e-03, /* 0xBF56C16C, 0x16C15177 */
+  2.48015872894767294178e-05, /* 0x3EFA01A0, 0x19CB1590 */
+ -2.75573143513906633035e-07, /* 0xBE927E4F, 0x809C52AD */
+  2.08757232129817482790e-09, /* 0x3E21EE9E, 0xBDB4B1C4 */
+ -1.13596475577881948265e-11}; /* 0xBDA8FAE9, 0xBE8838D4 */
 
 #ifdef __STDC__
 	double __kernel_cos(double x, double y)
@@ -72,17 +76,24 @@ C6  = -1.13596475577881948265e-11; /* 0xBDA8FAE9, 0xBE8838D4 */
 	double x,y;
 #endif
 {
-	double a,hz,z,r,qx;
+	double a,hz,z,r,qx,r1,r2,r3,z1,z2,z3;
 	int32_t ix;
+	z  = x*x;
 	GET_HIGH_WORD(ix,x);
 	ix &= 0x7fffffff;			/* ix = |x|'s high word*/
 	if(ix<0x3e400000) {			/* if x < 2**27 */
-	    if(((int)x)==0) return one;		/* generate inexact */
+	    if(((int)x)==0) return C[0];	/* generate inexact */
 	}
-	z  = x*x;
+#ifdef DO_NOT_USE_THIS
 	r  = z*(C1+z*(C2+z*(C3+z*(C4+z*(C5+z*C6)))));
-	if(ix < 0x3FD33333) 			/* if |x| < 0.3 */ 
-	    return one - (0.5*z - (z*r - x*y));
+#else
+	r1=z*C[6];r1=r1+C[5];z1=z*z;
+	r2=z*C[4];r2=r2+C[3];z2=z1*z;
+	r3=z*C[2];r3=r3+C[1];z3=z2*z1;
+	r=z3*r1+z2*r2+z*r3;
+#endif
+	if(ix < 0x3FD33333) 			/* if |x| < 0.3 */
+	    return C[0] - (0.5*z - (z*r - x*y));
 	else {
 	    if(ix > 0x3fe90000) {		/* x > 0.78125 */
 		qx = 0.28125;
@@ -90,7 +101,7 @@ C6  = -1.13596475577881948265e-11; /* 0xBDA8FAE9, 0xBE8838D4 */
 	        INSERT_WORDS(qx,ix-0x00200000,0);	/* x/4 */
 	    }
 	    hz = 0.5*z-qx;
-	    a  = one-qx;
+	    a  = C[0]-qx;
 	    return a - (hz - (z*r-x*y));
 	}
 }