[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Commits] r15585 - in /trunk: ./ libc/ libc/math/ libc/sysdeps/i386/fpu/ libc/sysdeps/ieee754/dbl-64/ libc/sysdeps/ieee754/flt-32/ lib...



Author: joseph
Date: Wed Oct 26 16:46:37 2011
New Revision: 15585

Log:
Merge changes between r15557 and r15584 from /fsf/trunk.

Added:
    trunk/libc/sysdeps/x86_64/fpu/multiarch/brandred-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/brandred-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/doasin-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/doasin-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/dosincos-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/dosincos-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_asin-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/e_asin-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_atan2-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/e_atan2-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_exp-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/e_exp-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_log-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/e_log-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/mpa-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/mpa-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/mpatan-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/mpatan-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/mpatan2-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/mpatan2-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/mpexp-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/mpexp-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/mplog-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/mplog-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/mpsqrt-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/mpsqrt-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/mptan-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/mptan-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/s_atan-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/s_atan-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/s_sin-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/s_sin-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/s_tan-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/s_tan-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/sincos32-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/sincos32-avx.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/slowexp-avx.c
      - copied unchanged from r15584, fsf/trunk/libc/sysdeps/x86_64/fpu/multiarch/slowexp-avx.c
    trunk/ports/sysdeps/m68k/m680x0/fpu/math_private.h
      - copied unchanged from r15584, fsf/trunk/ports/sysdeps/m68k/m680x0/fpu/math_private.h
    trunk/ports/sysdeps/m68k/m680x0/fpu/sincostab.c
      - copied unchanged from r15584, fsf/trunk/ports/sysdeps/m68k/m680x0/fpu/sincostab.c
Modified:
    trunk/   (props changed)
    trunk/libc/ChangeLog
    trunk/libc/math/math_private.h
    trunk/libc/sysdeps/i386/fpu/math_private.h
    trunk/libc/sysdeps/ieee754/dbl-64/e_rem_pio2.c
    trunk/libc/sysdeps/ieee754/dbl-64/s_round.c
    trunk/libc/sysdeps/ieee754/flt-32/e_j0f.c
    trunk/libc/sysdeps/ieee754/flt-32/s_roundf.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/Makefile
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_asin.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_atan2.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_exp.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/e_log.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/s_atan.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/s_sin.c
    trunk/libc/sysdeps/x86_64/fpu/multiarch/s_tan.c
    trunk/libc/sysdeps/x86_64/multiarch/init-arch.h
    trunk/libc/sysdeps/x86_64/multiarch/strcmp-sse42.S
    trunk/ports/ChangeLog.arm
    trunk/ports/ChangeLog.m68k
    trunk/ports/sysdeps/arm/dl-machine.h
    trunk/ports/sysdeps/unix/arm/sysdep.S

Propchange: trunk/
------------------------------------------------------------------------------
--- svn:mergeinfo (original)
+++ svn:mergeinfo Wed Oct 26 16:46:37 2011
@@ -1,1 +1,1 @@
-/fsf/trunk:15224-15557
+/fsf/trunk:15224-15584

Modified: trunk/libc/ChangeLog
==============================================================================
--- trunk/libc/ChangeLog (original)
+++ trunk/libc/ChangeLog Wed Oct 26 16:46:37 2011
@@ -1,3 +1,53 @@
+2011-10-26  Andreas Schwab  <schwab@xxxxxxxxxx>
+
+	* sysdeps/ieee754/flt-32/e_j0f.c: Fix use of math_force_eval.
+	* sysdeps/ieee754/dbl-64/s_round.c: Likewise.
+	* sysdeps/ieee754/flt-32/s_roundf.c: Likewise.
+
+	* math/math_private.h (math_force_eval): Allow non-addressable
+	arguments.
+	* sysdeps/i386/fpu/math_private.h (math_force_eval): Likewise.
+
+2011-10-25  Ulrich Drepper  <drepper@xxxxxxxxx>
+
+	* sysdeps/ieee754/dbl-64/e_rem_pio2.c: Comment everything out, the
+	file is not needed.
+
+	* sysdeps/x86_64/fpu/multiarch/e_asin.c: Support AVX variants.
+	* sysdeps/x86_64/fpu/multiarch/e_atan2.c: Likewise.
+	* sysdeps/x86_64/fpu/multiarch/e_exp.c: Likewise.
+	* sysdeps/x86_64/fpu/multiarch/e_log.c: Likewise.
+	* sysdeps/x86_64/fpu/multiarch/s_atan.c: Likewise.
+	* sysdeps/x86_64/fpu/multiarch/s_sin.c: Likewise.
+	* sysdeps/x86_64/fpu/multiarch/s_tan.c: Likewise.
+	* sysdeps/x86_64/fpu/multiarch/Makefile: Fix some CFLAGS-* variables.
+	Add AVX variants.
+	* sysdeps/x86_64/fpu/multiarch/brandred-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/doasin-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/dosincos-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/e_asin-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/e_atan2-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/e_exp-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/e_log-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/mpa-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/mpatan-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/mpatan2-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/mpexp-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/mplog-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/mpsqrt-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/mptan-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/s_atan-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/s_sin-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/s_tan-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/sincos32-avx.c: New file.
+	* sysdeps/x86_64/fpu/multiarch/slowexp-avx.c: New file.
+
+	* sysdeps/x86_64/multiarch/init-arch.h: Make bit_* macros available
+	all the time.  Define bit_AVX.  Define HAS_* macros using bit_* macros.
+
+	* sysdeps/x86_64/multiarch/strcmp-sse42.S: Move common code to earlier
+	place.  Use VEX encoding when compiling for AVX.
+
 2011-10-25  Andreas Schwab  <schwab@xxxxxxxxxx>
 
 	* wcsmbs/wcscmp.c (WCSCMP): Compare as wchar_t, not wint_t.
@@ -8,7 +58,7 @@
 2011-10-25  Ulrich Drepper  <drepper@xxxxxxxxx>
 
 	* sysdeps/ieee754/dbl-64/e_atanh.c: Use math_force_eval instead of a
-	useful if() expression.
+	useless if() expression.
 	* sysdeps/ieee754/dbl-64/e_j0.c: Likewise.
 	* sysdeps/ieee754/dbl-64/s_ceil.c: Likewise.
 	* sysdeps/ieee754/dbl-64/s_expm1.c: Likewise.

Modified: trunk/libc/math/math_private.h
==============================================================================
--- trunk/libc/math/math_private.h (original)
+++ trunk/libc/math/math_private.h Wed Oct 26 16:46:37 2011
@@ -354,8 +354,9 @@
 
 #ifndef math_opt_barrier
 #define math_opt_barrier(x) \
-({ __typeof (x) __x = x; __asm ("" : "+m" (__x)); __x; })
-#define math_force_eval(x) __asm __volatile ("" : : "m" (x))
+({ __typeof (x) __x = (x); __asm ("" : "+m" (__x)); __x; })
+#define math_force_eval(x) \
+({ __typeof (x) __x = (x); __asm __volatile ("" : : "m" (__x)); })
 #endif
 
 

Modified: trunk/libc/sysdeps/i386/fpu/math_private.h
==============================================================================
--- trunk/libc/sysdeps/i386/fpu/math_private.h (original)
+++ trunk/libc/sysdeps/i386/fpu/math_private.h Wed Oct 26 16:46:37 2011
@@ -1,16 +1,17 @@
 #ifndef _MATH_PRIVATE_H
 
 #define math_opt_barrier(x) \
-({ __typeof(x) __x;					\
+({ __typeof (x) __x;					\
    __asm ("" : "=t" (__x) : "0" (x));			\
    __x; })
 #define math_force_eval(x) \
 do							\
   {							\
+    __typeof (x) __x = (x);				\
     if (sizeof (x) <= sizeof (double))			\
-      __asm __volatile ("" : : "m" (x));		\
+      __asm __volatile ("" : : "m" (__x));		\
     else						\
-      __asm __volatile ("" : : "f" (x));		\
+      __asm __volatile ("" : : "f" (__x));		\
   }							\
 while (0)
 

Modified: trunk/libc/sysdeps/ieee754/dbl-64/e_rem_pio2.c
==============================================================================
--- trunk/libc/sysdeps/ieee754/dbl-64/e_rem_pio2.c (original)
+++ trunk/libc/sysdeps/ieee754/dbl-64/e_rem_pio2.c Wed Oct 26 16:46:37 2011
@@ -1,22 +1,19 @@
-/* @(#)e_rem_pio2.c 5.1 93/09/24 */
+#ifdef NOT_NEEDED_ANYMORE
+
 /*
  * ====================================================
  * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
  *
  * 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.
  * ====================================================
  */
 
-#if defined(LIBM_SCCS) && !defined(lint)
-static char rcsid[] = "$NetBSD: e_rem_pio2.c,v 1.8 1995/05/10 20:46:02 jtc Exp $";
-#endif
-
 /* __ieee754_rem_pio2(x,y)
- * 
- * return the remainder of x rem pi/2 in y[0]+y[1] 
+ *
+ * return the remainder of x rem pi/2 in y[0]+y[1]
  * use __kernel_rem_pio2()
  */
 
@@ -24,31 +21,23 @@
 #include "math_private.h"
 
 /*
- * Table of constants for 2/pi, 396 Hex digits (476 decimal) of 2/pi 
+ * Table of constants for 2/pi, 396 Hex digits (476 decimal) of 2/pi
  */
-#ifdef __STDC__
 static const int32_t two_over_pi[] = {
-#else
-static int32_t two_over_pi[] = {
-#endif
-0xA2F983, 0x6E4E44, 0x1529FC, 0x2757D1, 0xF534DD, 0xC0DB62, 
-0x95993C, 0x439041, 0xFE5163, 0xABDEBB, 0xC561B7, 0x246E3A, 
-0x424DD2, 0xE00649, 0x2EEA09, 0xD1921C, 0xFE1DEB, 0x1CB129, 
-0xA73EE8, 0x8235F5, 0x2EBB44, 0x84E99C, 0x7026B4, 0x5F7E41, 
-0x3991D6, 0x398353, 0x39F49C, 0x845F8B, 0xBDF928, 0x3B1FF8, 
-0x97FFDE, 0x05980F, 0xEF2F11, 0x8B5A0A, 0x6D1F6D, 0x367ECF, 
-0x27CB09, 0xB74F46, 0x3F669E, 0x5FEA2D, 0x7527BA, 0xC7EBE5, 
-0xF17B3D, 0x0739F7, 0x8A5292, 0xEA6BFB, 0x5FB11F, 0x8D5D08, 
-0x560330, 0x46FC7B, 0x6BABF0, 0xCFBC20, 0x9AF436, 0x1DA9E3, 
-0x91615E, 0xE61B08, 0x659985, 0x5F14A0, 0x68408D, 0xFFD880, 
-0x4D7327, 0x310606, 0x1556CA, 0x73A8C9, 0x60E27B, 0xC08C6B, 
+0xA2F983, 0x6E4E44, 0x1529FC, 0x2757D1, 0xF534DD, 0xC0DB62,
+0x95993C, 0x439041, 0xFE5163, 0xABDEBB, 0xC561B7, 0x246E3A,
+0x424DD2, 0xE00649, 0x2EEA09, 0xD1921C, 0xFE1DEB, 0x1CB129,
+0xA73EE8, 0x8235F5, 0x2EBB44, 0x84E99C, 0x7026B4, 0x5F7E41,
+0x3991D6, 0x398353, 0x39F49C, 0x845F8B, 0xBDF928, 0x3B1FF8,
+0x97FFDE, 0x05980F, 0xEF2F11, 0x8B5A0A, 0x6D1F6D, 0x367ECF,
+0x27CB09, 0xB74F46, 0x3F669E, 0x5FEA2D, 0x7527BA, 0xC7EBE5,
+0xF17B3D, 0x0739F7, 0x8A5292, 0xEA6BFB, 0x5FB11F, 0x8D5D08,
+0x560330, 0x46FC7B, 0x6BABF0, 0xCFBC20, 0x9AF436, 0x1DA9E3,
+0x91615E, 0xE61B08, 0x659985, 0x5F14A0, 0x68408D, 0xFFD880,
+0x4D7327, 0x310606, 0x1556CA, 0x73A8C9, 0x60E27B, 0xC08C6B,
 };
 
-#ifdef __STDC__
 static const int32_t npio2_hw[] = {
-#else
-static int32_t npio2_hw[] = {
-#endif
 0x3FF921FB, 0x400921FB, 0x4012D97C, 0x401921FB, 0x401F6A7A, 0x4022D97C,
 0x4025FDBB, 0x402921FB, 0x402C463A, 0x402F6A7A, 0x4031475C, 0x4032D97C,
 0x40346B9C, 0x4035FDBB, 0x40378FDB, 0x403921FB, 0x403AB41B, 0x403C463A,
@@ -67,11 +56,7 @@
  * pio2_3t:  pi/2 - (pio2_1+pio2_2+pio2_3)
  */
 
-#ifdef __STDC__
-static const double 
-#else
-static double 
-#endif
+static const double
 zero =  0.00000000000000000000e+00, /* 0x00000000, 0x00000000 */
 half =  5.00000000000000000000e-01, /* 0x3FE00000, 0x00000000 */
 two24 =  1.67772160000000000000e+07, /* 0x41700000, 0x00000000 */
@@ -83,12 +68,8 @@
 pio2_3  =  2.02226624871116645580e-21, /* 0x3BA3198A, 0x2E000000 */
 pio2_3t =  8.47842766036889956997e-32; /* 0x397B839A, 0x252049C1 */
 
-#ifdef __STDC__
-	int32_t __ieee754_rem_pio2(double x, double *y)
-#else
-	int32_t __ieee754_rem_pio2(x,y)
-	double x,y[];
-#endif
+int32_t
+__ieee754_rem_pio2(double x, double *y)
 {
 	double z,w,t,r,fn;
 	double tx[3];
@@ -100,9 +81,9 @@
 	if(ix<=0x3fe921fb)   /* |x| ~<= pi/4 , no need for reduction */
 	    {y[0] = x; y[1] = 0; return 0;}
 	if(ix<0x4002d97c) {  /* |x| < 3pi/4, special case with n=+-1 */
-	    if(hx>0) { 
+	    if(hx>0) {
 		z = x - pio2_1;
-		if(ix!=0x3ff921fb) { 	/* 33+53 bit pi is good enough */
+		if(ix!=0x3ff921fb) {	/* 33+53 bit pi is good enough */
 		    y[0] = z - pio2_1t;
 		    y[1] = (z-y[0])-pio2_1t;
 		} else {		/* near pi/2, use 33+33+53 bit pi */
@@ -113,7 +94,7 @@
 		return 1;
 	    } else {	/* negative x */
 		z = x + pio2_1;
-		if(ix!=0x3ff921fb) { 	/* 33+53 bit pi is good enough */
+		if(ix!=0x3ff921fb) {	/* 33+53 bit pi is good enough */
 		    y[0] = z + pio2_1t;
 		    y[1] = (z-y[0])+pio2_1t;
 		} else {		/* near pi/2, use 33+33+53 bit pi */
@@ -130,36 +111,36 @@
 	    fn = (double)n;
 	    r  = t-fn*pio2_1;
 	    w  = fn*pio2_1t;	/* 1st round good to 85 bit */
-	    if(n<32&&ix!=npio2_hw[n-1]) {	
+	    if(n<32&&ix!=npio2_hw[n-1]) {
 		y[0] = r-w;	/* quick check no cancellation */
 	    } else {
-	        u_int32_t high;
-	        j  = ix>>20;
-	        y[0] = r-w; 
+		u_int32_t high;
+		j  = ix>>20;
+		y[0] = r-w;
 		GET_HIGH_WORD(high,y[0]);
-	        i = j-((high>>20)&0x7ff);
-	        if(i>16) {  /* 2nd iteration needed, good to 118 */
+		i = j-((high>>20)&0x7ff);
+		if(i>16) {  /* 2nd iteration needed, good to 118 */
 		    t  = r;
-		    w  = fn*pio2_2;	
+		    w  = fn*pio2_2;
 		    r  = t-w;
-		    w  = fn*pio2_2t-((t-r)-w);	
+		    w  = fn*pio2_2t-((t-r)-w);
 		    y[0] = r-w;
 		    GET_HIGH_WORD(high,y[0]);
 		    i = j-((high>>20)&0x7ff);
 		    if(i>49)  {	/* 3rd iteration need, 151 bits acc */
-		    	t  = r;	/* will cover all possible cases */
-		    	w  = fn*pio2_3;	
-		    	r  = t-w;
-		    	w  = fn*pio2_3t-((t-r)-w);	
-		    	y[0] = r-w;
+			t  = r;	/* will cover all possible cases */
+			w  = fn*pio2_3;
+			r  = t-w;
+			w  = fn*pio2_3t-((t-r)-w);
+			y[0] = r-w;
 		    }
 		}
 	    }
 	    y[1] = (r-y[0])-w;
-	    if(hx<0) 	{y[0] = -y[0]; y[1] = -y[1]; return -n;}
+	    if(hx<0)	{y[0] = -y[0]; y[1] = -y[1]; return -n;}
 	    else	 return n;
 	}
-    /* 
+    /*
      * all other (large) arguments
      */
 	if(ix>=0x7ff00000) {		/* x is inf or NaN */
@@ -168,7 +149,7 @@
     /* set z = scalbn(|x|,ilogb(x)-23) */
 	GET_LOW_WORD(low,x);
 	SET_LOW_WORD(z,low);
-	e0 	= (ix>>20)-1046;	/* e0 = ilogb(z)-23; */
+	e0	= (ix>>20)-1046;	/* e0 = ilogb(z)-23; */
 	SET_HIGH_WORD(z, ix - ((int32_t)(e0<<20)));
 	for(i=0;i<2;i++) {
 		tx[i] = (double)((int32_t)(z));
@@ -181,3 +162,5 @@
 	if(hx<0) {y[0] = -y[0]; y[1] = -y[1]; return -n;}
 	return n;
 }
+
+#endif

Modified: trunk/libc/sysdeps/ieee754/dbl-64/s_round.c
==============================================================================
--- trunk/libc/sysdeps/ieee754/dbl-64/s_round.c (original)
+++ trunk/libc/sysdeps/ieee754/dbl-64/s_round.c Wed Oct 26 16:46:37 2011
@@ -38,7 +38,7 @@
     {
       if (j0 < 0)
 	{
-	  math_force_eval (huge + x > 0.0);
+	  math_force_eval (huge + x);
 
 	  i0 &= 0x80000000;
 	  if (j0 == -1)
@@ -51,7 +51,7 @@
 	  if (((i0 & i) | i1) == 0)
 	    /* X is integral.  */
 	    return x;
-	  math_force_eval (huge + x > 0.0);
+	  math_force_eval (huge + x);
 
 	  /* Raise inexact if x != 0.  */
 	  i0 += 0x00080000 >> j0;
@@ -74,7 +74,7 @@
 	/* X is integral.  */
 	return x;
 
-      math_force_eval (huge + x > 0.0);
+      math_force_eval (huge + x);
 
       /* Raise inexact if x != 0.  */
       u_int32_t j = i1 + (1 << (51 - j0));

Modified: trunk/libc/sysdeps/ieee754/flt-32/e_j0f.c
==============================================================================
--- trunk/libc/sysdeps/ieee754/flt-32/e_j0f.c (original)
+++ trunk/libc/sysdeps/ieee754/flt-32/e_j0f.c Wed Oct 26 16:46:37 2011
@@ -66,7 +66,7 @@
 		return z;
 	}
 	if(ix<0x39000000) {	/* |x| < 2**-13 */
-	    math_force_eval(huge+x>one);	/* raise inexact if x != 0 */
+	    math_force_eval(huge+x);		/* raise inexact if x != 0 */
 	    if(ix<0x32000000) return one;	/* |x|<2**-27 */
 	    else	      return one - (float)0.25*x*x;
 	}

Modified: trunk/libc/sysdeps/ieee754/flt-32/s_roundf.c
==============================================================================
--- trunk/libc/sysdeps/ieee754/flt-32/s_roundf.c (original)
+++ trunk/libc/sysdeps/ieee754/flt-32/s_roundf.c Wed Oct 26 16:46:37 2011
@@ -37,7 +37,7 @@
     {
       if (j0 < 0)
 	{
-	  math_force_eval (huge + x > 0.0F);
+	  math_force_eval (huge + x);
 
 	  i0 &= 0x80000000;
 	  if (j0 == -1)
@@ -49,7 +49,7 @@
 	  if ((i0 & i) == 0)
 	    /* X is integral.  */
 	    return x;
-	  math_force_eval (huge + x > 0.0F);
+	  math_force_eval (huge + x);
 
 	  /* Raise inexact if x != 0.  */
 	  i0 += 0x00400000 >> j0;

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/Makefile
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/Makefile (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/Makefile Wed Oct 26 16:46:37 2011
@@ -30,7 +30,36 @@
 CFLAGS-sincos32-fma4.c = -mfma4
 CFLAGS-slowexp-fma4.c = -mfma4
 CFLAGS-slowpow-fma4.c = -mfma4
-CLFAGS-s_sin-fma4.c = -mfma4
-CLFAGS-s_tan-fma4.c = -mfma4
+CFLAGS-s_sin-fma4.c = -mfma4
+CFLAGS-s_tan-fma4.c = -mfma4
+endif
+
+ifeq ($(config-cflags-avx),yes)
+libm-sysdep_routines += e_exp-avx e_log-avx s_atan-avx \
+			e_asin-avx e_atan2-avx s_sin-avx s_tan-avx \
+			mplog-avx mpa-avx slowexp-avx \
+			sincos32-avx doasin-avx dosincos-avx \
+			brandred-avx mpexp-avx \
+			mpatan2-avx mpatan-avx mpsqrt-avx mptan-avx
+
+CFLAGS-brandred-avx.c = -mavx
+CFLAGS-doasin-avx.c = -mavx
+CFLAGS-dosincos-avx.c = -mavx
+CFLAGS-e_asin-avx.c = -mavx
+CFLAGS-e_atan2-avx.c = -mavx
+CFLAGS-e_exp-avx.c = -mavx
+CFLAGS-e_log-avx.c = -mavx
+CFLAGS-mpa-avx.c = -mavx
+CFLAGS-mpatan-avx.c = -mavx
+CFLAGS-mpatan2-avx.c = -mavx
+CFLAGS-mpexp-avx.c = -mavx
+CFLAGS-mplog-avx.c = -mavx
+CFLAGS-mpsqrt-avx.c = -mavx
+CFLAGS-mptan-avx.c = -mavx
+CFLAGS-s_atan-avx.c = -mavx
+CFLAGS-s_sin-avx.c = -mavx
+CFLAGS-sincos32-avx.c = -mavx
+CFLAGS-slowexp-avx.c = -mavx
+CFLAGS-s_tan-avx.c = -mavx
 endif
 endif

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/e_asin.c
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/e_asin.c (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/e_asin.c Wed Oct 26 16:46:37 2011
@@ -1,18 +1,29 @@
-#ifdef HAVE_FMA4_SUPPORT
+#if defined HAVE_FMA4_SUPPORT || defined HAVE_AVX_SUPPORT
 # include <init-arch.h>
 # include <math_private.h>
 
 extern double __ieee754_acos_sse2 (double);
+extern double __ieee754_asin_sse2 (double);
+extern double __ieee754_acos_avx (double);
+extern double __ieee754_asin_avx (double);
+# ifdef HAVE_FMA4_SUPPORT
 extern double __ieee754_acos_fma4 (double);
-extern double __ieee754_asin_sse2 (double);
 extern double __ieee754_asin_fma4 (double);
+# else
+#  undef HAS_FMA4
+#  define HAS_FMA4 0
+#  define __ieee754_acos_fma4 ((void *) 0)
+#  define __ieee754_asin_fma4 ((void *) 0)
+# endif
 
 libm_ifunc (__ieee754_acos,
-	    HAS_FMA4 ? __ieee754_acos_fma4 : __ieee754_acos_sse2);
+	    HAS_FMA4 ? __ieee754_acos_fma4
+	    : (HAS_AVX ? __ieee754_acos_avx : __ieee754_acos_sse2));
 strong_alias (__ieee754_acos, __acos_finite)
 
 libm_ifunc (__ieee754_asin,
-	    HAS_FMA4 ? __ieee754_asin_fma4 : __ieee754_asin_sse2);
+	    HAS_FMA4 ? __ieee754_asin_fma4
+	    : (HAS_AVX ? __ieee754_asin_avx : __ieee754_asin_sse2));
 strong_alias (__ieee754_asin, __asin_finite)
 
 # define __ieee754_acos __ieee754_acos_sse2

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/e_atan2.c
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/e_atan2.c (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/e_atan2.c Wed Oct 26 16:46:37 2011
@@ -1,12 +1,20 @@
-#ifdef HAVE_FMA4_SUPPORT
+#if defined HAVE_FMA4_SUPPORT || defined HAVE_AVX_SUPPORT
 # include <init-arch.h>
 # include <math_private.h>
 
 extern double __ieee754_atan2_sse2 (double, double);
+extern double __ieee754_atan2_avx (double, double);
+# ifdef HAVE_FMA4_SUPPORT
 extern double __ieee754_atan2_fma4 (double, double);
+# else
+#  undef HAS_FMA4
+#  define HAS_FMA4 0
+#  define __ieee754_atan2_fma4 ((void *) 0)
+# endif
 
 libm_ifunc (__ieee754_atan2,
-	    HAS_FMA4 ? __ieee754_atan2_fma4 : __ieee754_atan2_sse2);
+	    HAS_FMA4 ? __ieee754_atan2_fma4
+	    : (HAS_AVX ? __ieee754_atan2_avx : __ieee754_atan2_sse2));
 strong_alias (__ieee754_atan2, __atan2_finite)
 
 # define __ieee754_atan2 __ieee754_atan2_sse2

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/e_exp.c
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/e_exp.c (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/e_exp.c Wed Oct 26 16:46:37 2011
@@ -1,11 +1,20 @@
-#ifdef HAVE_FMA4_SUPPORT
+#if defined HAVE_FMA4_SUPPORT || defined HAVE_AVX_SUPPORT
 # include <init-arch.h>
 # include <math_private.h>
 
 extern double __ieee754_exp_sse2 (double);
+extern double __ieee754_exp_avx (double);
+# ifdef HAVE_FMA4_SUPPORT
 extern double __ieee754_exp_fma4 (double);
+# else
+#  undef HAS_FMA4
+#  define HAS_FMA4 0
+#  define __ieee754_exp_fma4 ((void *) 0)
+# endif
 
-libm_ifunc (__ieee754_exp, HAS_FMA4 ? __ieee754_exp_fma4 : __ieee754_exp_sse2);
+libm_ifunc (__ieee754_exp,
+	    HAS_FMA4 ? __ieee754_exp_fma4
+	    : (HAS_AVX ? __ieee754_exp_avx : __ieee754_exp_sse2));
 strong_alias (__ieee754_exp, __exp_finite)
 
 # define __ieee754_exp __ieee754_exp_sse2

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/e_log.c
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/e_log.c (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/e_log.c Wed Oct 26 16:46:37 2011
@@ -1,11 +1,21 @@
-#ifdef HAVE_FMA4_SUPPORT
+#if defined HAVE_FMA4_SUPPORT || defined HAVE_AVX_SUPPORT
 # include <init-arch.h>
 # include <math_private.h>
 
 extern double __ieee754_log_sse2 (double);
+extern double __ieee754_log_avx (double);
+# ifdef HAVE_FMA4_SUPPORT
 extern double __ieee754_log_fma4 (double);
+# else
+#  undef HAS_FMA4
+#  define HAS_FMA4 0
+#  define __ieee754_log_fma4 ((void *) 0)
+# endif
 
-libm_ifunc (__ieee754_log, HAS_FMA4 ? __ieee754_log_fma4 : __ieee754_log_sse2);
+libm_ifunc (__ieee754_log,
+	    HAS_FMA4 ? __ieee754_log_fma4
+	    : (HAS_AVX ? __ieee754_log_avx
+	       : __ieee754_log_sse2));
 strong_alias (__ieee754_log, __log_finite)
 
 # define __ieee754_log __ieee754_log_sse2

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/s_atan.c
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/s_atan.c (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/s_atan.c Wed Oct 26 16:46:37 2011
@@ -1,11 +1,18 @@
-#ifdef HAVE_FMA4_SUPPORT
+#if defined HAVE_FMA4_SUPPORT || defined HAVE_AVX_SUPPORT
 # include <init-arch.h>
 # include <math.h>
 
 extern double __atan_sse2 (double);
+extern double __atan_avx (double);
+# ifdef HAVE_FMA4_SUPPORT
 extern double __atan_fma4 (double);
+# else
+#  undef HAS_FMA4
+#  define HAS_FMA4 0
+#  define __atan_fma4 ((void *) 0)
+# endif
 
-libm_ifunc (atan, HAS_FMA4 ? __atan_fma4 : __atan_sse2);
+libm_ifunc (atan, HAS_FMA4 ? __atan_fma4 : HAS_AVX ? __atan_avx : __atan_sse2);
 
 # define atan __atan_sse2
 #endif

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/s_sin.c
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/s_sin.c (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/s_sin.c Wed Oct 26 16:46:37 2011
@@ -1,17 +1,26 @@
-#ifdef HAVE_FMA4_SUPPORT
+#if defined HAVE_FMA4_SUPPORT || defined HAVE_AVX_SUPPORT
 # include <init-arch.h>
 # include <math.h>
 # undef NAN
 
 extern double __cos_sse2 (double);
+extern double __sin_sse2 (double);
+extern double __cos_avx (double);
+extern double __sin_avx (double);
+# ifdef HAVE_FMA4_SUPPORT
 extern double __cos_fma4 (double);
-extern double __sin_sse2 (double);
 extern double __sin_fma4 (double);
+# else
+#  undef HAS_FMA4
+#  define HAS_FMA4 0
+#  define __cos_fma4 ((void *) 0)
+#  define __sin_fma4 ((void *) 0)
+# endif
 
-libm_ifunc (__cos, HAS_FMA4 ? __cos_fma4 : __cos_sse2);
+libm_ifunc (__cos, HAS_FMA4 ? __cos_fma4 : HAS_AVX ? __cos_avx : __cos_sse2);
 weak_alias (__cos, cos)
 
-libm_ifunc (__sin, HAS_FMA4 ? __sin_fma4 : __sin_sse2);
+libm_ifunc (__sin, HAS_FMA4 ? __sin_fma4 : HAS_AVX ? __sin_avx : __sin_sse2);
 weak_alias (__sin, sin)
 
 # define __cos __cos_sse2

Modified: trunk/libc/sysdeps/x86_64/fpu/multiarch/s_tan.c
==============================================================================
--- trunk/libc/sysdeps/x86_64/fpu/multiarch/s_tan.c (original)
+++ trunk/libc/sysdeps/x86_64/fpu/multiarch/s_tan.c Wed Oct 26 16:46:37 2011
@@ -1,11 +1,18 @@
-#ifdef HAVE_FMA4_SUPPORT
+#if defined HAVE_FMA4_SUPPORT || defined HAVE_AVX_SUPPORT
 # include <init-arch.h>
 # include <math.h>
 
 extern double __tan_sse2 (double);
+extern double __tan_avx (double);
+# ifdef HAVE_FMA4_SUPPORT
 extern double __tan_fma4 (double);
+# else
+#  undef HAS_FMA4
+#  define HAS_FMA4 0
+#  define __tan_fma4 ((void *) 0)
+# endif
 
-libm_ifunc (tan, HAS_FMA4 ? __tan_fma4 : __tan_sse2);
+libm_ifunc (tan, HAS_FMA4 ? __tan_fma4 : HAS_AVX ? __tan_avx : __tan_sse2);
 
 # define tan __tan_sse2
 #endif

Modified: trunk/libc/sysdeps/x86_64/multiarch/init-arch.h
==============================================================================
--- trunk/libc/sysdeps/x86_64/multiarch/init-arch.h (original)
+++ trunk/libc/sysdeps/x86_64/multiarch/init-arch.h Wed Oct 26 16:46:37 2011
@@ -23,15 +23,18 @@
 #define bit_Fast_Unaligned_Load		(1 << 4)
 #define bit_Prefer_PMINUB_for_stringop	(1 << 5)
 
+#define bit_SSE2	(1 << 26)
+#define bit_SSSE3	(1 << 9)
+#define bit_SSE4_1	(1 << 19)
+#define bit_SSE4_2	(1 << 20)
+#define bit_AVX		(1 << 28)
+#define bit_POPCOUNT	(1 << 23)
+#define bit_FMA		(1 << 12)
+#define bit_FMA4	(1 << 16)
+
 #ifdef	__ASSEMBLER__
 
 # include <ifunc-defines.h>
-
-# define bit_SSE2	(1 << 26)
-# define bit_SSSE3	(1 << 9)
-# define bit_SSE4_1	(1 << 19)
-# define bit_SSE4_2	(1 << 20)
-# define bit_AVX	(1 << 28)
 
 # define index_SSE2	COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_EDX_OFFSET
 # define index_SSSE3	COMMON_CPUID_INDEX_1*CPUID_SIZE+CPUID_ECX_OFFSET
@@ -104,17 +107,18 @@
 # endif
 
 # define HAS_CPU_FEATURE(idx, reg, bit) \
-  ((__get_cpu_features ()->cpuid[idx].reg & (1 << (bit))) != 0)
+  ((__get_cpu_features ()->cpuid[idx].reg & (bit)) != 0)
 
 /* Following are the feature tests used throughout libc.  */
 
-# define HAS_SSE2	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, edx, 26)
-# define HAS_POPCOUNT	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, 23)
-# define HAS_SSSE3	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, 9)
-# define HAS_SSE4_1	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, 19)
-# define HAS_SSE4_2	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, 20)
-# define HAS_FMA	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, 12)
-# define HAS_FMA4	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_80000001, ecx, 16)
+# define HAS_SSE2	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, edx, bit_SSE2)
+# define HAS_POPCOUNT	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_POPCOUNT)
+# define HAS_SSSE3	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSSE3)
+# define HAS_SSE4_1	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_1)
+# define HAS_SSE4_2	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_SSE4_2)
+# define HAS_FMA	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_FMA)
+# define HAS_AVX	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_1, ecx, bit_AVX)
+# define HAS_FMA4	HAS_CPU_FEATURE (COMMON_CPUID_INDEX_80000001, ecx, bit_FMA4)
 
 # define index_Fast_Rep_String		FEATURE_INDEX_1
 # define index_Fast_Copy_Backward	FEATURE_INDEX_1

Modified: trunk/libc/sysdeps/x86_64/multiarch/strcmp-sse42.S
==============================================================================
--- trunk/libc/sysdeps/x86_64/multiarch/strcmp-sse42.S (original)
+++ trunk/libc/sysdeps/x86_64/multiarch/strcmp-sse42.S Wed Oct 26 16:46:37 2011
@@ -70,6 +70,23 @@
 	.byte	0x0f,0x1f,0x44,0x00,0x00
 END (GLABEL(__strncasecmp))
 	/* FALLTHROUGH to strncasecmp_l.  */
+#endif
+
+
+#ifdef USE_AVX
+# define movdqa vmovdqa
+# define movdqu vmovdqu
+# define pmovmskb vpmovmskb
+# define pcmpistri vpcmpistri
+# define psubb vpsubb
+# define pcmpeqb vpcmpeqb
+# define psrldq vpsrldq
+# define pslldq vpslldq
+# define palignr vpalignr
+# define pxor vpxor
+# define D(arg) arg, arg
+#else
+# define D(arg) arg
 #endif
 
 STRCMP_SSE42:
@@ -179,10 +196,10 @@
 #else
 # define TOLOWER(reg1, reg2)
 #endif
-	pxor	%xmm0, %xmm0		/* clear %xmm0 for null char checks */
-	pcmpeqb	%xmm1, %xmm0		/* Any null chars? */
-	pcmpeqb	%xmm2, %xmm1		/* compare first 16 bytes for equality */
-	psubb	%xmm0, %xmm1		/* packed sub of comparison results*/
+	pxor	%xmm0, D(%xmm0)		/* clear %xmm0 for null char checks */
+	pcmpeqb	%xmm1, D(%xmm0)		/* Any null chars? */
+	pcmpeqb	%xmm2, D(%xmm1)		/* compare first 16 bytes for equality */
+	psubb	%xmm0, D(%xmm1)		/* packed sub of comparison results*/
 	pmovmskb %xmm1, %edx
 	sub	$0xffff, %edx		/* if first 16 bytes are same, edx == 0xffff */
 	jnz	LABEL(less16bytes)/* If not, find different value or null char */
@@ -206,6 +223,7 @@
 	xor	%r8d, %r8d
 	and	$0xf, %ecx		/* offset of rsi */
 	and	$0xf, %eax		/* offset of rdi */
+	pxor	%xmm0, D(%xmm0)		/* clear %xmm0 for null char check */
 	cmp	%eax, %ecx
 	je	LABEL(ashr_0)		/* rsi and rdi relative offset same */
 	ja	LABEL(bigger)
@@ -213,10 +231,13 @@
 	xchg	%ecx, %eax
 	xchg	%rsi, %rdi
 LABEL(bigger):
+	movdqa	(%rdi), %xmm2
+	movdqa	(%rsi), %xmm1
 	lea	15(%rax), %r9
 	sub	%rcx, %r9
 	lea	LABEL(unaligned_table)(%rip), %r10
 	movslq	(%r10, %r9,4), %r9
+	pcmpeqb	%xmm1, D(%xmm0)		/* Any null chars? */
 	lea	(%r10, %r9), %r10
 	jmp	*%r10			/* jump to corresponding case */
 
@@ -229,16 +250,15 @@
 LABEL(ashr_0):
 
 	movdqa	(%rsi), %xmm1
-	pxor	%xmm0, %xmm0		/* clear %xmm0 for null char check */
-	pcmpeqb	%xmm1, %xmm0		/* Any null chars? */
-#if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
-	pcmpeqb	(%rdi), %xmm1		/* compare 16 bytes for equality */
+	pcmpeqb	%xmm1, D(%xmm0)		/* Any null chars? */
+#if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
+	pcmpeqb	(%rdi), D(%xmm1)	/* compare 16 bytes for equality */
 #else
 	movdqa	(%rdi), %xmm2
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm2, %xmm1		/* compare 16 bytes for equality */
-#endif
-	psubb	%xmm0, %xmm1		/* packed sub of comparison results*/
+	pcmpeqb	%xmm2, D(%xmm1)		/* compare 16 bytes for equality */
+#endif
+	psubb	%xmm0, D(%xmm1)		/* packed sub of comparison results*/
 	pmovmskb %xmm1, %r9d
 	shr	%cl, %edx		/* adjust 0xffff for offset */
 	shr	%cl, %r9d		/* adjust for 16-byte offset */
@@ -251,7 +271,6 @@
 	UPDATE_STRNCMP_COUNTER
 	mov	$16, %rcx
 	mov	$16, %r9
-	pxor	%xmm0, %xmm0		/* clear xmm0, may have changed above */
 
 	/*
 	 * Now both strings are aligned at 16-byte boundary. Loop over strings
@@ -319,14 +338,10 @@
  */
 	.p2align 4
 LABEL(ashr_1):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0		/* Any null chars? */
-	pslldq	$15, %xmm2		/* shift first string to align with second */
+	pslldq	$15, D(%xmm2)		/* shift first string to align with second */
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2		/* compare 16 bytes for equality */
-	psubb	%xmm0, %xmm2		/* packed sub of comparison results*/
+	pcmpeqb	%xmm1, D(%xmm2)		/* compare 16 bytes for equality */
+	psubb	%xmm0, D(%xmm2)		/* packed sub of comparison results*/
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx		/* adjust 0xffff for offset */
 	shr	%cl, %r9d		/* adjust for 16-byte offset */
@@ -335,7 +350,6 @@
 	movdqa	(%rdi), %xmm3
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx		/* index for loads*/
 	mov	$1, %r9d		/* byte position left over from less32bytes case */
 	/*
@@ -355,7 +369,7 @@
 
 LABEL(nibble_ashr_1_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $1, -16(%rdi, %rdx), %xmm0
+	palignr $1, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -374,7 +388,7 @@
 	jg	LABEL(nibble_ashr_1_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $1, -16(%rdi, %rdx), %xmm0
+	palignr $1, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -394,7 +408,7 @@
 LABEL(nibble_ashr_1_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$1, %xmm0
+	psrldq	$1, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -412,14 +426,10 @@
  */
 	.p2align 4
 LABEL(ashr_2):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$14, %xmm2
+	pslldq	$14, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -428,7 +438,6 @@
 	movdqa	(%rdi), %xmm3
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$2, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -448,7 +457,7 @@
 
 LABEL(nibble_ashr_2_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $2, -16(%rdi, %rdx), %xmm0
+	palignr $2, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -467,7 +476,7 @@
 	jg	LABEL(nibble_ashr_2_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $2, -16(%rdi, %rdx), %xmm0
+	palignr $2, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -487,7 +496,7 @@
 LABEL(nibble_ashr_2_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$2, %xmm0
+	psrldq	$2, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -505,14 +514,10 @@
  */
 	.p2align 4
 LABEL(ashr_3):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$13, %xmm2
+	pslldq	$13, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -522,7 +527,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$3, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -541,7 +545,7 @@
 
 LABEL(nibble_ashr_3_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $3, -16(%rdi, %rdx), %xmm0
+	palignr $3, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -560,7 +564,7 @@
 	jg	LABEL(nibble_ashr_3_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $3, -16(%rdi, %rdx), %xmm0
+	palignr $3, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -580,7 +584,7 @@
 LABEL(nibble_ashr_3_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$3, %xmm0
+	psrldq	$3, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -598,14 +602,10 @@
  */
 	.p2align 4
 LABEL(ashr_4):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$12, %xmm2
+	pslldq	$12, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -615,7 +615,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$4, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -635,7 +634,7 @@
 
 LABEL(nibble_ashr_4_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $4, -16(%rdi, %rdx), %xmm0
+	palignr $4, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -654,7 +653,7 @@
 	jg	LABEL(nibble_ashr_4_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $4, -16(%rdi, %rdx), %xmm0
+	palignr $4, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -674,7 +673,7 @@
 LABEL(nibble_ashr_4_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$4, %xmm0
+	psrldq	$4, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -692,14 +691,10 @@
  */
 	.p2align 4
 LABEL(ashr_5):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$11, %xmm2
+	pslldq	$11, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -709,7 +704,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$5, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -729,7 +723,7 @@
 
 LABEL(nibble_ashr_5_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $5, -16(%rdi, %rdx), %xmm0
+	palignr $5, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -749,7 +743,7 @@
 
 	movdqa	(%rdi, %rdx), %xmm0
 
-	palignr $5, -16(%rdi, %rdx), %xmm0
+	palignr $5, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -769,7 +763,7 @@
 LABEL(nibble_ashr_5_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$5, %xmm0
+	psrldq	$5, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -787,14 +781,10 @@
  */
 	.p2align 4
 LABEL(ashr_6):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$10, %xmm2
+	pslldq	$10, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -804,7 +794,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$6, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -824,7 +813,7 @@
 
 LABEL(nibble_ashr_6_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $6, -16(%rdi, %rdx), %xmm0
+	palignr $6, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -843,7 +832,7 @@
 	jg	LABEL(nibble_ashr_6_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $6, -16(%rdi, %rdx), %xmm0
+	palignr $6, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -863,7 +852,7 @@
 LABEL(nibble_ashr_6_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$6, %xmm0
+	psrldq	$6, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -881,14 +870,10 @@
  */
 	.p2align 4
 LABEL(ashr_7):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$9, %xmm2
+	pslldq	$9, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -898,7 +883,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$7, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -918,7 +902,7 @@
 
 LABEL(nibble_ashr_7_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $7, -16(%rdi, %rdx), %xmm0
+	palignr $7, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -937,7 +921,7 @@
 	jg	LABEL(nibble_ashr_7_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $7, -16(%rdi, %rdx), %xmm0
+	palignr $7, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri	$0x1a,(%rsi,%rdx), %xmm0
 #else
@@ -957,7 +941,7 @@
 LABEL(nibble_ashr_7_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$7, %xmm0
+	psrldq	$7, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -975,14 +959,10 @@
  */
 	.p2align 4
 LABEL(ashr_8):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$8, %xmm2
+	pslldq	$8, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -992,7 +972,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$8, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1012,7 +991,7 @@
 
 LABEL(nibble_ashr_8_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $8, -16(%rdi, %rdx), %xmm0
+	palignr $8, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1031,7 +1010,7 @@
 	jg	LABEL(nibble_ashr_8_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $8, -16(%rdi, %rdx), %xmm0
+	palignr $8, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1051,7 +1030,7 @@
 LABEL(nibble_ashr_8_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$8, %xmm0
+	psrldq	$8, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1069,14 +1048,10 @@
  */
 	.p2align 4
 LABEL(ashr_9):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$7, %xmm2
+	pslldq	$7, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -1086,7 +1061,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$9, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1107,7 +1081,7 @@
 LABEL(nibble_ashr_9_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
 
-	palignr $9, -16(%rdi, %rdx), %xmm0
+	palignr $9, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1126,7 +1100,7 @@
 	jg	LABEL(nibble_ashr_9_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $9, -16(%rdi, %rdx), %xmm0
+	palignr $9, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1146,7 +1120,7 @@
 LABEL(nibble_ashr_9_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$9, %xmm0
+	psrldq	$9, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1164,14 +1138,10 @@
  */
 	.p2align 4
 LABEL(ashr_10):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$6, %xmm2
+	pslldq	$6, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -1181,7 +1151,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$10, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1201,7 +1170,7 @@
 
 LABEL(nibble_ashr_10_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $10, -16(%rdi, %rdx), %xmm0
+	palignr $10, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1220,7 +1189,7 @@
 	jg	LABEL(nibble_ashr_10_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $10, -16(%rdi, %rdx), %xmm0
+	palignr $10, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1240,7 +1209,7 @@
 LABEL(nibble_ashr_10_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$10, %xmm0
+	psrldq	$10, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1258,14 +1227,10 @@
  */
 	.p2align 4
 LABEL(ashr_11):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$5, %xmm2
+	pslldq	$5, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -1275,7 +1240,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$11, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1295,7 +1259,7 @@
 
 LABEL(nibble_ashr_11_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $11, -16(%rdi, %rdx), %xmm0
+	palignr $11, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1314,7 +1278,7 @@
 	jg	LABEL(nibble_ashr_11_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $11, -16(%rdi, %rdx), %xmm0
+	palignr $11, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1334,7 +1298,7 @@
 LABEL(nibble_ashr_11_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$11, %xmm0
+	psrldq	$11, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1352,14 +1316,10 @@
  */
 	.p2align 4
 LABEL(ashr_12):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$4, %xmm2
+	pslldq	$4, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -1369,7 +1329,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$12, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1389,7 +1348,7 @@
 
 LABEL(nibble_ashr_12_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $12, -16(%rdi, %rdx), %xmm0
+	palignr $12, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1408,7 +1367,7 @@
 	jg	LABEL(nibble_ashr_12_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $12, -16(%rdi, %rdx), %xmm0
+	palignr $12, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1428,7 +1387,7 @@
 LABEL(nibble_ashr_12_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$12, %xmm0
+	psrldq	$12, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1446,14 +1405,10 @@
  */
 	.p2align 4
 LABEL(ashr_13):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$3, %xmm2
+	pslldq	$3, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -1463,7 +1418,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$13, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1484,7 +1438,7 @@
 
 LABEL(nibble_ashr_13_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $13, -16(%rdi, %rdx), %xmm0
+	palignr $13, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1503,7 +1457,7 @@
 	jg	LABEL(nibble_ashr_13_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $13, -16(%rdi, %rdx), %xmm0
+	palignr $13, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1523,7 +1477,7 @@
 LABEL(nibble_ashr_13_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$13, %xmm0
+	psrldq	$13, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1541,14 +1495,10 @@
  */
 	.p2align 4
 LABEL(ashr_14):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq  $2, %xmm2
+	pslldq  $2, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -1558,7 +1508,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$14, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1579,7 +1528,7 @@
 
 LABEL(nibble_ashr_14_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $14, -16(%rdi, %rdx), %xmm0
+	palignr $14, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1598,7 +1547,7 @@
 	jg	LABEL(nibble_ashr_14_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $14, -16(%rdi, %rdx), %xmm0
+	palignr $14, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1618,7 +1567,7 @@
 LABEL(nibble_ashr_14_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$14, %xmm0
+	psrldq	$14, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1636,14 +1585,10 @@
  */
 	.p2align 4
 LABEL(ashr_15):
-	pxor	%xmm0, %xmm0
-	movdqa	(%rdi), %xmm2
-	movdqa	(%rsi), %xmm1
-	pcmpeqb	%xmm1, %xmm0
-	pslldq	$1, %xmm2
+	pslldq	$1, D(%xmm2)
 	TOLOWER (%xmm1, %xmm2)
-	pcmpeqb	%xmm1, %xmm2
-	psubb	%xmm0, %xmm2
+	pcmpeqb	%xmm1, D(%xmm2)
+	psubb	%xmm0, D(%xmm2)
 	pmovmskb %xmm2, %r9d
 	shr	%cl, %edx
 	shr	%cl, %r9d
@@ -1654,7 +1599,6 @@
 
 	UPDATE_STRNCMP_COUNTER
 
-	pxor	%xmm0, %xmm0
 	mov	$16, %rcx	/* index for loads */
 	mov	$15, %r9d	/* byte position left over from less32bytes case */
 	/*
@@ -1676,7 +1620,7 @@
 
 LABEL(nibble_ashr_15_restart_use):
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $15, -16(%rdi, %rdx), %xmm0
+	palignr $15, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1695,7 +1639,7 @@
 	jg	LABEL(nibble_ashr_15_use)
 
 	movdqa	(%rdi, %rdx), %xmm0
-	palignr $15, -16(%rdi, %rdx), %xmm0
+	palignr $15, -16(%rdi, %rdx), D(%xmm0)
 #if !defined USE_AS_STRCASECMP_L && !defined USE_AS_STRNCASECMP_L
 	pcmpistri $0x1a, (%rsi,%rdx), %xmm0
 #else
@@ -1715,7 +1659,7 @@
 LABEL(nibble_ashr_15_use):
 	sub	$0x1000, %r10
 	movdqa	-16(%rdi, %rdx), %xmm0
-	psrldq	$15, %xmm0
+	psrldq	$15, D(%xmm0)
 	pcmpistri      $0x3a,%xmm0, %xmm0
 #if defined USE_AS_STRNCMP || defined USE_AS_STRNCASECMP_L
 	cmp	%r11, %rcx
@@ -1834,3 +1778,14 @@
 #undef LABEL
 #undef GLABEL
 #undef SECTION
+#undef movdqa
+#undef movdqu
+#undef pmovmskb
+#undef pcmpistri
+#undef psubb
+#undef pcmpeqb
+#undef psrldq
+#undef pslldq
+#undef palignr
+#undef pxor
+#undef D

Modified: trunk/ports/ChangeLog.arm
==============================================================================
--- trunk/ports/ChangeLog.arm (original)
+++ trunk/ports/ChangeLog.arm Wed Oct 26 16:46:37 2011
@@ -1,3 +1,8 @@
+2011-10-26  Joseph Myers  <joseph@xxxxxxxxxxxxxxxx>
+
+	* sysdeps/arm/dl-machine.h, sysdeps/unix/arm/sysdep.S: Restore
+	cases for use in rtld.
+
 2011-10-05  Andreas Schwab  <schwab@xxxxxxxxxx>
 
 	* sysdeps/arm/dl-machine.h (elf_machine_rel, elf_machine_rela)

Modified: trunk/ports/ChangeLog.m68k
==============================================================================
--- trunk/ports/ChangeLog.m68k (original)
+++ trunk/ports/ChangeLog.m68k Wed Oct 26 16:46:37 2011
@@ -1,3 +1,9 @@
+2011-10-26  Andreas Schwab  <schwab@xxxxxxxxxxxxxx>
+
+	* sysdeps/m68k/m680x0/fpu/math_private.h: New file.
+
+	* sysdeps/m68k/m680x0/fpu/sincostab.c: New file.
+
 2011-10-23  Andreas Schwab  <schwab@xxxxxxxxxxxxxx>
 
 	* sysdeps/m68k/coldfire/fpu/e_sqrt.c: Add __sqrt_finite alias.

Modified: trunk/ports/sysdeps/arm/dl-machine.h
==============================================================================
--- trunk/ports/sysdeps/arm/dl-machine.h (original)
+++ trunk/ports/sysdeps/arm/dl-machine.h Wed Oct 26 16:46:37 2011
@@ -242,12 +242,18 @@
    define the value.
    ELF_RTYPE_CLASS_NOCOPY iff TYPE should not be allowed to resolve to one
    of the main executable's symbols, as for a COPY reloc.  */
-#define elf_machine_type_class(type) \
+#ifndef RTLD_BOOTSTRAP
+# define elf_machine_type_class(type) \
   ((((type) == R_ARM_JUMP_SLOT || (type) == R_ARM_TLS_DTPMOD32		\
      || (type) == R_ARM_TLS_DTPOFF32 || (type) == R_ARM_TLS_TPOFF32	\
      || (type) == R_ARM_TLS_DESC)					\
     * ELF_RTYPE_CLASS_PLT)						\
    | (((type) == R_ARM_COPY) * ELF_RTYPE_CLASS_COPY))
+#else
+#define elf_machine_type_class(type) \
+  ((((type) == R_ARM_JUMP_SLOT) * ELF_RTYPE_CLASS_PLT)	\
+   | (((type) == R_ARM_COPY) * ELF_RTYPE_CLASS_COPY))
+#endif
 
 /* A reloc type used for ld.so cmdline arg lookups to reject PLT entries.  */
 #define ELF_MACHINE_JMP_SLOT	R_ARM_JUMP_SLOT

Modified: trunk/ports/sysdeps/unix/arm/sysdep.S
==============================================================================
--- trunk/ports/sysdeps/unix/arm/sysdep.S (original)
+++ trunk/ports/sysdeps/unix/arm/sysdep.S Wed Oct 26 16:46:37 2011
@@ -42,6 +42,7 @@
 	moveq r0, $EAGAIN	/* Yes; translate it to EAGAIN.  */
 #endif
 
+#ifndef IS_IN_rtld
 	mov ip, lr
 	cfi_register (lr, ip)
 	mov r1, r0
@@ -57,6 +58,16 @@
 	RETINSTR (, ip)
 
 1:	.word errno(gottpoff) + (. - 2b - 8)
+#elif RTLD_PRIVATE_ERRNO
+	ldr r1, 1f
+0:	str r0, [pc, r1]
+	mvn r0, $0
+	DO_RET(r14)
+
+1:	.word C_SYMBOL_NAME(rtld_errno) - 0b - 8
+#else
+#error "Unsupported non-TLS case"
+#endif
 
 #undef	__syscall_error
 END (__syscall_error)

_______________________________________________
Commits mailing list
Commits@xxxxxxxxxx
http://eglibc.org/cgi-bin/mailman/listinfo/commits