diff mbox

Fix yn overflow handling in non-default rounding modes (bug 16561, bug 16562)

Message ID Pine.LNX.4.64.1406232334560.28255@digraph.polyomino.org.uk
State Committed
Headers show

Commit Message

Joseph Myers June 23, 2014, 11:35 p.m. UTC
This patch fixes bugs 16561 and 16562, bad results of yn in overflow
cases in non-default rounding modes, both because an intermediate
overflow in the recurrence does not get detected if the result is not
an infinity and because an overflowing result may occur in the wrong
sign.  The fix is to set FE_TONEAREST mode internally for the parts of
the function where such overflows can occur (which includes the call
to y1 - where yn is used to compute a Bessel function of order -1,
negating the result of y1 isn't correct for overflowing results in
directed rounding modes) and then compute an overflowing value in the
original rounding mode if the to-nearest result was an infinity.

Tested x86_64 and x86 and ulps updated accordingly.  Also tested for
mips64 and powerpc32 to test the ldbl-128 and ldbl-128ibm changes.

(The tests for these bugs were added in my previous y1 patch, so the
only thing this patch has to do with the testsuite is enable yn
testing in all rounding modes.)

2014-06-23  Joseph Myers  <joseph@codesourcery.com>

	[BZ #16561]
	[BZ #16562]
	* sysdeps/ieee754/dbl-64/e_jn.c: Include <float.h>.
	(__ieee754_yn): Set FE_TONEAREST mode internally and then
	recompute overflowing results in original rounding mode.
	* sysdeps/ieee754/flt-32/e_jnf.c: Include <float.h>.
	(__ieee754_ynf): Set FE_TONEAREST mode internally and then
	recompute overflowing results in original rounding mode.
	* sysdeps/ieee754/ldbl-128/e_jnl.c: Include <float.h>.
	(__ieee754_ynl): Set FE_TONEAREST mode internally and then
	recompute overflowing results in original rounding mode.
	* sysdeps/ieee754/ldbl-128ibm/e_jnl.c: Include <float.h>.
	(__ieee754_ynl): Set FE_TONEAREST mode internally and then
	recompute overflowing results in original rounding mode.
	* sysdeps/ieee754/ldbl-96/e_jnl.c: Include <float.h>.
	(__ieee754_ynl): Set FE_TONEAREST mode internally and then
	recompute overflowing results in original rounding mode.
	* sysdeps/i386/fpu/fenv_private.h [!__SSE2_MATH__]
	(libc_feholdsetround_ctx): New macro.
	* math/libm-test.inc (yn_test): Use ALL_RM_TEST.
	* sysdeps/i386/fpu/libm-test-ulps: Update.
	* sysdeps/x86_64/fpu/libm-test-ulps : Likewise.

Comments

Joseph Myers June 27, 2014, 12:07 p.m. UTC | #1
Ping.  This patch 
<https://sourceware.org/ml/libc-alpha/2014-06/msg00660.html> is pending 
review.
Andreas Jaeger June 27, 2014, 1:38 p.m. UTC | #2
On 06/24/2014 01:35 AM, Joseph S. Myers wrote:
> This patch fixes bugs 16561 and 16562, bad results of yn in overflow
> cases in non-default rounding modes, both because an intermediate
> overflow in the recurrence does not get detected if the result is not
> an infinity and because an overflowing result may occur in the wrong
> sign.  The fix is to set FE_TONEAREST mode internally for the parts of
> the function where such overflows can occur (which includes the call
> to y1 - where yn is used to compute a Bessel function of order -1,
> negating the result of y1 isn't correct for overflowing results in
> directed rounding modes) and then compute an overflowing value in the
> original rounding mode if the to-nearest result was an infinity.
> 
> Tested x86_64 and x86 and ulps updated accordingly.  Also tested for
> mips64 and powerpc32 to test the ldbl-128 and ldbl-128ibm changes.

thanks,
Andreas
diff mbox

Patch

diff --git a/math/libm-test.inc b/math/libm-test.inc
index 45eaaef..f86a4fa 100644
--- a/math/libm-test.inc
+++ b/math/libm-test.inc
@@ -9609,9 +9609,7 @@  static const struct test_if_f_data yn_test_data[] =
 static void
 yn_test (void)
 {
-  START (yn, 0);
-  RUN_TEST_LOOP_if_f (yn, yn_test_data, );
-  END;
+  ALL_RM_TEST (yn, 0, yn_test_data, RUN_TEST_LOOP_if_f, END);
 }
 
 
diff --git a/sysdeps/i386/fpu/fenv_private.h b/sysdeps/i386/fpu/fenv_private.h
index 3998387..e20e1f1 100644
--- a/sysdeps/i386/fpu/fenv_private.h
+++ b/sysdeps/i386/fpu/fenv_private.h
@@ -482,6 +482,7 @@  libc_feupdateenv_387_ctx (struct rm_ctx *ctx)
 #else
 # define libc_feholdexcept_setround_ctx	libc_feholdexcept_setround_387_ctx
 # define libc_feupdateenv_ctx		libc_feupdateenv_387_ctx
+# define libc_feholdsetround_ctx	libc_feholdsetround_387_ctx
 # define libc_feresetround_ctx		libc_feresetround_387_ctx
 #endif /* __SSE2_MATH__ */
 
diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps
index 7cf9586..be0e7c3 100644
--- a/sysdeps/i386/fpu/libm-test-ulps
+++ b/sysdeps/i386/fpu/libm-test-ulps
@@ -1931,4 +1931,28 @@  ifloat: 3
 ildouble: 4
 ldouble: 4
 
+Function: "yn_downward":
+double: 2
+float: 2
+idouble: 2
+ifloat: 2
+ildouble: 5
+ldouble: 5
+
+Function: "yn_towardzero":
+double: 2
+float: 3
+idouble: 2
+ifloat: 3
+ildouble: 5
+ldouble: 5
+
+Function: "yn_upward":
+double: 2
+float: 3
+idouble: 2
+ifloat: 3
+ildouble: 4
+ldouble: 4
+
 # end of automatic generation
diff --git a/sysdeps/ieee754/dbl-64/e_jn.c b/sysdeps/ieee754/dbl-64/e_jn.c
index 236878b..bbe0426 100644
--- a/sysdeps/ieee754/dbl-64/e_jn.c
+++ b/sysdeps/ieee754/dbl-64/e_jn.c
@@ -37,6 +37,7 @@ 
  */
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 
@@ -248,7 +249,7 @@  __ieee754_yn (int n, double x)
 {
   int32_t i, hx, ix, lx;
   int32_t sign;
-  double a, b, temp;
+  double a, b, temp, ret;
 
   EXTRACT_WORDS (hx, lx, x);
   ix = 0x7fffffff & hx;
@@ -268,57 +269,67 @@  __ieee754_yn (int n, double x)
     }
   if (n == 0)
     return (__ieee754_y0 (x));
-  if (n == 1)
-    return (sign * __ieee754_y1 (x));
-  if (__glibc_unlikely (ix == 0x7ff00000))
-    return zero;
-  if (ix >= 0x52D00000)      /* x > 2**302 */
-    { /* (x >> n**2)
-		 *	    Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-		 *	    Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-		 *	    Let s=sin(x), c=cos(x),
-		 *		xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
-		 *
-		 *		   n	sin(xn)*sqt2	cos(xn)*sqt2
-		 *		----------------------------------
-		 *		   0	 s-c		 c+s
-		 *		   1	-s-c		-c+s
-		 *		   2	-s+c		-c-s
-		 *		   3	 s+c		 c-s
-		 */
-      double c;
-      double s;
-      __sincos (x, &s, &c);
-      switch (n & 3)
-	{
-	case 0: temp = s - c; break;
-	case 1: temp = -s - c; break;
-	case 2: temp = -s + c; break;
-	case 3: temp = s + c; break;
-	}
-      b = invsqrtpi * temp / __ieee754_sqrt (x);
-    }
-  else
-    {
-      u_int32_t high;
-      a = __ieee754_y0 (x);
-      b = __ieee754_y1 (x);
-      /* quit if b is -inf */
-      GET_HIGH_WORD (high, b);
-      for (i = 1; i < n && high != 0xfff00000; i++)
-	{
-	  temp = b;
-	  b = ((double) (i + i) / x) * b - a;
-	  GET_HIGH_WORD (high, b);
-	  a = temp;
-	}
-      /* If B is +-Inf, set up errno accordingly.  */
-      if (!__finite (b))
-	__set_errno (ERANGE);
-    }
-  if (sign > 0)
-    return b;
-  else
-    return -b;
+  {
+    SET_RESTORE_ROUND (FE_TONEAREST);
+    if (n == 1)
+      {
+	ret = sign * __ieee754_y1 (x);
+	goto out;
+      }
+    if (__glibc_unlikely (ix == 0x7ff00000))
+      return zero;
+    if (ix >= 0x52D00000)      /* x > 2**302 */
+      { /* (x >> n**2)
+	 *	    Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *	    Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *	    Let s=sin(x), c=cos(x),
+	 *		xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
+	 *
+	 *		   n	sin(xn)*sqt2	cos(xn)*sqt2
+	 *		----------------------------------
+	 *		   0	 s-c		 c+s
+	 *		   1	-s-c		-c+s
+	 *		   2	-s+c		-c-s
+	 *		   3	 s+c		 c-s
+	 */
+	double c;
+	double s;
+	__sincos (x, &s, &c);
+	switch (n & 3)
+	  {
+	  case 0: temp = s - c; break;
+	  case 1: temp = -s - c; break;
+	  case 2: temp = -s + c; break;
+	  case 3: temp = s + c; break;
+	  }
+	b = invsqrtpi * temp / __ieee754_sqrt (x);
+      }
+    else
+      {
+	u_int32_t high;
+	a = __ieee754_y0 (x);
+	b = __ieee754_y1 (x);
+	/* quit if b is -inf */
+	GET_HIGH_WORD (high, b);
+	for (i = 1; i < n && high != 0xfff00000; i++)
+	  {
+	    temp = b;
+	    b = ((double) (i + i) / x) * b - a;
+	    GET_HIGH_WORD (high, b);
+	    a = temp;
+	  }
+	/* If B is +-Inf, set up errno accordingly.  */
+	if (!__finite (b))
+	  __set_errno (ERANGE);
+      }
+    if (sign > 0)
+      ret = b;
+    else
+      ret = -b;
+  }
+ out:
+  if (__isinf (ret))
+    ret = __copysign (DBL_MAX, ret) * DBL_MAX;
+  return ret;
 }
 strong_alias (__ieee754_yn, __yn_finite)
diff --git a/sysdeps/ieee754/flt-32/e_jnf.c b/sysdeps/ieee754/flt-32/e_jnf.c
index 5984d94..86085cc 100644
--- a/sysdeps/ieee754/flt-32/e_jnf.c
+++ b/sysdeps/ieee754/flt-32/e_jnf.c
@@ -14,6 +14,7 @@ 
  */
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 
@@ -169,6 +170,8 @@  strong_alias (__ieee754_jnf, __jnf_finite)
 float
 __ieee754_ynf(int n, float x)
 {
+    float ret;
+    {
 	int32_t i,hx,ix;
 	u_int32_t ib;
 	int32_t sign;
@@ -187,7 +190,11 @@  __ieee754_ynf(int n, float x)
 		sign = 1 - ((n&1)<<1);
 	}
 	if(n==0) return(__ieee754_y0f(x));
-	if(n==1) return(sign*__ieee754_y1f(x));
+	SET_RESTORE_ROUNDF (FE_TONEAREST);
+	if(n==1) {
+	    ret = sign*__ieee754_y1f(x);
+	    goto out;
+	}
 	if(__builtin_expect(ix==0x7f800000, 0)) return zero;
 
 	a = __ieee754_y0f(x);
@@ -203,6 +210,11 @@  __ieee754_ynf(int n, float x)
 	/* If B is +-Inf, set up errno accordingly.  */
 	if (! __finitef (b))
 	  __set_errno (ERANGE);
-	if(sign>0) return b; else return -b;
+	if(sign>0) ret = b; else ret = -b;
+    }
+ out:
+    if (__isinff (ret))
+	ret = __copysignf (FLT_MAX, ret) * FLT_MAX;
+    return ret;
 }
 strong_alias (__ieee754_ynf, __ynf_finite)
diff --git a/sysdeps/ieee754/ldbl-128/e_jnl.c b/sysdeps/ieee754/ldbl-128/e_jnl.c
index c2a4923..4e32d38 100644
--- a/sysdeps/ieee754/ldbl-128/e_jnl.c
+++ b/sysdeps/ieee754/ldbl-128/e_jnl.c
@@ -57,6 +57,7 @@ 
  */
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 
@@ -300,7 +301,7 @@  __ieee754_ynl (int n, long double x)
   u_int32_t se;
   int32_t i, ix;
   int32_t sign;
-  long double a, b, temp;
+  long double a, b, temp, ret;
   ieee854_long_double_shape_type u;
 
   u.value = x;
@@ -328,70 +329,80 @@  __ieee754_ynl (int n, long double x)
     }
   if (n == 0)
     return (__ieee754_y0l (x));
-  if (n == 1)
-    return (sign * __ieee754_y1l (x));
-  if (ix >= 0x7fff0000)
-    return zero;
-  if (ix >= 0x412D0000)
-    {				/* x > 2**302 */
+  {
+    SET_RESTORE_ROUNDL (FE_TONEAREST);
+    if (n == 1)
+      {
+	ret = sign * __ieee754_y1l (x);
+	goto out;
+      }
+    if (ix >= 0x7fff0000)
+      return zero;
+    if (ix >= 0x412D0000)
+      {				/* x > 2**302 */
 
-      /* ??? See comment above on the possible futility of this.  */
+	/* ??? See comment above on the possible futility of this.  */
 
-      /* (x >> n**2)
-       *      Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-       *      Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-       *      Let s=sin(x), c=cos(x),
-       *          xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
-       *
-       *             n    sin(xn)*sqt2    cos(xn)*sqt2
-       *          ----------------------------------
-       *             0     s-c             c+s
-       *             1    -s-c            -c+s
-       *             2    -s+c            -c-s
-       *             3     s+c             c-s
-       */
-      long double s;
-      long double c;
-      __sincosl (x, &s, &c);
-      switch (n & 3)
-	{
-	case 0:
-	  temp = s - c;
-	  break;
-	case 1:
-	  temp = -s - c;
-	  break;
-	case 2:
-	  temp = -s + c;
-	  break;
-	case 3:
-	  temp = s + c;
-	  break;
-	}
-      b = invsqrtpi * temp / __ieee754_sqrtl (x);
-    }
-  else
-    {
-      a = __ieee754_y0l (x);
-      b = __ieee754_y1l (x);
-      /* quit if b is -inf */
-      u.value = b;
-      se = u.parts32.w0 & 0xffff0000;
-      for (i = 1; i < n && se != 0xffff0000; i++)
-	{
-	  temp = b;
-	  b = ((long double) (i + i) / x) * b - a;
-	  u.value = b;
-	  se = u.parts32.w0 & 0xffff0000;
-	  a = temp;
-	}
-    }
-  /* If B is +-Inf, set up errno accordingly.  */
-  if (! __finitel (b))
-    __set_errno (ERANGE);
-  if (sign > 0)
-    return b;
-  else
-    return -b;
+	/* (x >> n**2)
+	 *      Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *      Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *      Let s=sin(x), c=cos(x),
+	 *          xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
+	 *
+	 *             n    sin(xn)*sqt2    cos(xn)*sqt2
+	 *          ----------------------------------
+	 *             0     s-c             c+s
+	 *             1    -s-c            -c+s
+	 *             2    -s+c            -c-s
+	 *             3     s+c             c-s
+	 */
+	long double s;
+	long double c;
+	__sincosl (x, &s, &c);
+	switch (n & 3)
+	  {
+	  case 0:
+	    temp = s - c;
+	    break;
+	  case 1:
+	    temp = -s - c;
+	    break;
+	  case 2:
+	    temp = -s + c;
+	    break;
+	  case 3:
+	    temp = s + c;
+	    break;
+	  }
+	b = invsqrtpi * temp / __ieee754_sqrtl (x);
+      }
+    else
+      {
+	a = __ieee754_y0l (x);
+	b = __ieee754_y1l (x);
+	/* quit if b is -inf */
+	u.value = b;
+	se = u.parts32.w0 & 0xffff0000;
+	for (i = 1; i < n && se != 0xffff0000; i++)
+	  {
+	    temp = b;
+	    b = ((long double) (i + i) / x) * b - a;
+	    u.value = b;
+	    se = u.parts32.w0 & 0xffff0000;
+	    a = temp;
+	  }
+      }
+    /* If B is +-Inf, set up errno accordingly.  */
+    if (! __finitel (b))
+      __set_errno (ERANGE);
+    if (sign > 0)
+      ret = b;
+    else
+      ret = -b;
+  }
+ out:
+  if (__isinfl (ret))
+    ret = __copysignl (LDBL_MAX, ret) * LDBL_MAX;
+  return ret;
 }
 strong_alias (__ieee754_ynl, __ynl_finite)
diff --git a/sysdeps/ieee754/ldbl-128ibm/e_jnl.c b/sysdeps/ieee754/ldbl-128ibm/e_jnl.c
index 6761a0d..589f1f8 100644
--- a/sysdeps/ieee754/ldbl-128ibm/e_jnl.c
+++ b/sysdeps/ieee754/ldbl-128ibm/e_jnl.c
@@ -57,6 +57,7 @@ 
  */
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 
@@ -300,7 +301,7 @@  __ieee754_ynl (int n, long double x)
   uint32_t se, lx;
   int32_t i, ix;
   int32_t sign;
-  long double a, b, temp;
+  long double a, b, temp, ret;
   double xhi;
 
   xhi = ldbl_high (x);
@@ -328,72 +329,82 @@  __ieee754_ynl (int n, long double x)
     }
   if (n == 0)
     return (__ieee754_y0l (x));
-  if (n == 1)
-    return (sign * __ieee754_y1l (x));
-  if (ix >= 0x7ff00000)
-    return zero;
-  if (ix >= 0x52D00000)
-    {				/* x > 2**302 */
+  {
+    SET_RESTORE_ROUNDL (FE_TONEAREST);
+    if (n == 1)
+      {
+	ret = sign * __ieee754_y1l (x);
+	goto out;
+      }
+    if (ix >= 0x7ff00000)
+      return zero;
+    if (ix >= 0x52D00000)
+      {				/* x > 2**302 */
 
-      /* ??? See comment above on the possible futility of this.  */
+	/* ??? See comment above on the possible futility of this.  */
 
-      /* (x >> n**2)
-       *      Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-       *      Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-       *      Let s=sin(x), c=cos(x),
-       *          xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
-       *
-       *             n    sin(xn)*sqt2    cos(xn)*sqt2
-       *          ----------------------------------
-       *             0     s-c             c+s
-       *             1    -s-c            -c+s
-       *             2    -s+c            -c-s
-       *             3     s+c             c-s
-       */
-      long double s;
-      long double c;
-      __sincosl (x, &s, &c);
-      switch (n & 3)
-	{
-	case 0:
-	  temp = s - c;
-	  break;
-	case 1:
-	  temp = -s - c;
-	  break;
-	case 2:
-	  temp = -s + c;
-	  break;
-	case 3:
-	  temp = s + c;
-	  break;
-	}
-      b = invsqrtpi * temp / __ieee754_sqrtl (x);
-    }
-  else
-    {
-      a = __ieee754_y0l (x);
-      b = __ieee754_y1l (x);
-      /* quit if b is -inf */
-      xhi = ldbl_high (b);
-      GET_HIGH_WORD (se, xhi);
-      se &= 0xfff00000;
-      for (i = 1; i < n && se != 0xfff00000; i++)
-	{
-	  temp = b;
-	  b = ((long double) (i + i) / x) * b - a;
-	  xhi = ldbl_high (b);
-	  GET_HIGH_WORD (se, xhi);
-	  se &= 0xfff00000;
-	  a = temp;
-	}
-    }
-  /* If B is +-Inf, set up errno accordingly.  */
-  if (! __finitel (b))
-    __set_errno (ERANGE);
-  if (sign > 0)
-    return b;
-  else
-    return -b;
+	/* (x >> n**2)
+	 *      Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *      Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *      Let s=sin(x), c=cos(x),
+	 *          xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
+	 *
+	 *             n    sin(xn)*sqt2    cos(xn)*sqt2
+	 *          ----------------------------------
+	 *             0     s-c             c+s
+	 *             1    -s-c            -c+s
+	 *             2    -s+c            -c-s
+	 *             3     s+c             c-s
+	 */
+	long double s;
+	long double c;
+	__sincosl (x, &s, &c);
+	switch (n & 3)
+	  {
+	  case 0:
+	    temp = s - c;
+	    break;
+	  case 1:
+	    temp = -s - c;
+	    break;
+	  case 2:
+	    temp = -s + c;
+	    break;
+	  case 3:
+	    temp = s + c;
+	    break;
+	  }
+	b = invsqrtpi * temp / __ieee754_sqrtl (x);
+      }
+    else
+      {
+	a = __ieee754_y0l (x);
+	b = __ieee754_y1l (x);
+	/* quit if b is -inf */
+	xhi = ldbl_high (b);
+	GET_HIGH_WORD (se, xhi);
+	se &= 0xfff00000;
+	for (i = 1; i < n && se != 0xfff00000; i++)
+	  {
+	    temp = b;
+	    b = ((long double) (i + i) / x) * b - a;
+	    xhi = ldbl_high (b);
+	    GET_HIGH_WORD (se, xhi);
+	    se &= 0xfff00000;
+	    a = temp;
+	  }
+      }
+    /* If B is +-Inf, set up errno accordingly.  */
+    if (! __finitel (b))
+      __set_errno (ERANGE);
+    if (sign > 0)
+      ret = b;
+    else
+      ret = -b;
+  }
+ out:
+  if (__isinfl (ret))
+    ret = __copysignl (LDBL_MAX, ret) * LDBL_MAX;
+  return ret;
 }
 strong_alias (__ieee754_ynl, __ynl_finite)
diff --git a/sysdeps/ieee754/ldbl-96/e_jnl.c b/sysdeps/ieee754/ldbl-96/e_jnl.c
index 11d097c..95ff242 100644
--- a/sysdeps/ieee754/ldbl-96/e_jnl.c
+++ b/sysdeps/ieee754/ldbl-96/e_jnl.c
@@ -57,6 +57,7 @@ 
  */
 
 #include <errno.h>
+#include <float.h>
 #include <math.h>
 #include <math_private.h>
 
@@ -293,7 +294,7 @@  __ieee754_ynl (int n, long double x)
   u_int32_t se, i0, i1;
   int32_t i, ix;
   int32_t sign;
-  long double a, b, temp;
+  long double a, b, temp, ret;
 
 
   GET_LDOUBLE_WORDS (se, i0, i1, x);
@@ -314,69 +315,79 @@  __ieee754_ynl (int n, long double x)
     }
   if (n == 0)
     return (__ieee754_y0l (x));
-  if (n == 1)
-    return (sign * __ieee754_y1l (x));
-  if (__glibc_unlikely (ix == 0x7fff))
-    return zero;
-  if (ix >= 0x412D)
-    {				/* x > 2**302 */
+  {
+    SET_RESTORE_ROUNDL (FE_TONEAREST);
+    if (n == 1)
+      {
+	ret = sign * __ieee754_y1l (x);
+	goto out;
+      }
+    if (__glibc_unlikely (ix == 0x7fff))
+      return zero;
+    if (ix >= 0x412D)
+      {				/* x > 2**302 */
 
-      /* ??? See comment above on the possible futility of this.  */
+	/* ??? See comment above on the possible futility of this.  */
 
-      /* (x >> n**2)
-       *      Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-       *      Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
-       *      Let s=sin(x), c=cos(x),
-       *          xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
-       *
-       *             n    sin(xn)*sqt2    cos(xn)*sqt2
-       *          ----------------------------------
-       *             0     s-c             c+s
-       *             1    -s-c            -c+s
-       *             2    -s+c            -c-s
-       *             3     s+c             c-s
-       */
-      long double s;
-      long double c;
-      __sincosl (x, &s, &c);
-      switch (n & 3)
-	{
-	case 0:
-	  temp = s - c;
-	  break;
-	case 1:
-	  temp = -s - c;
-	  break;
-	case 2:
-	  temp = -s + c;
-	  break;
-	case 3:
-	  temp = s + c;
-	  break;
-	}
-      b = invsqrtpi * temp / __ieee754_sqrtl (x);
-    }
-  else
-    {
-      a = __ieee754_y0l (x);
-      b = __ieee754_y1l (x);
-      /* quit if b is -inf */
-      GET_LDOUBLE_WORDS (se, i0, i1, b);
-      /* Use 0xffffffff since GET_LDOUBLE_WORDS sign-extends SE.  */
-      for (i = 1; i < n && se != 0xffffffff; i++)
-	{
-	  temp = b;
-	  b = ((long double) (i + i) / x) * b - a;
-	  GET_LDOUBLE_WORDS (se, i0, i1, b);
-	  a = temp;
-	}
-    }
-  /* If B is +-Inf, set up errno accordingly.  */
-  if (! __finitel (b))
-    __set_errno (ERANGE);
-  if (sign > 0)
-    return b;
-  else
-    return -b;
+	/* (x >> n**2)
+	 *      Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *      Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
+	 *      Let s=sin(x), c=cos(x),
+	 *          xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
+	 *
+	 *             n    sin(xn)*sqt2    cos(xn)*sqt2
+	 *          ----------------------------------
+	 *             0     s-c             c+s
+	 *             1    -s-c            -c+s
+	 *             2    -s+c            -c-s
+	 *             3     s+c             c-s
+	 */
+	long double s;
+	long double c;
+	__sincosl (x, &s, &c);
+	switch (n & 3)
+	  {
+	  case 0:
+	    temp = s - c;
+	    break;
+	  case 1:
+	    temp = -s - c;
+	    break;
+	  case 2:
+	    temp = -s + c;
+	    break;
+	  case 3:
+	    temp = s + c;
+	    break;
+	  }
+	b = invsqrtpi * temp / __ieee754_sqrtl (x);
+      }
+    else
+      {
+	a = __ieee754_y0l (x);
+	b = __ieee754_y1l (x);
+	/* quit if b is -inf */
+	GET_LDOUBLE_WORDS (se, i0, i1, b);
+	/* Use 0xffffffff since GET_LDOUBLE_WORDS sign-extends SE.  */
+	for (i = 1; i < n && se != 0xffffffff; i++)
+	  {
+	    temp = b;
+	    b = ((long double) (i + i) / x) * b - a;
+	    GET_LDOUBLE_WORDS (se, i0, i1, b);
+	    a = temp;
+	  }
+      }
+    /* If B is +-Inf, set up errno accordingly.  */
+    if (! __finitel (b))
+      __set_errno (ERANGE);
+    if (sign > 0)
+      ret = b;
+    else
+      ret = -b;
+  }
+ out:
+  if (__isinfl (ret))
+    ret = __copysignl (LDBL_MAX, ret) * LDBL_MAX;
+  return ret;
 }
 strong_alias (__ieee754_ynl, __ynl_finite)
diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps
index 5469887..7904d3e 100644
--- a/sysdeps/x86_64/fpu/libm-test-ulps
+++ b/sysdeps/x86_64/fpu/libm-test-ulps
@@ -2025,4 +2025,28 @@  ifloat: 3
 ildouble: 4
 ldouble: 4
 
+Function: "yn_downward":
+double: 3
+float: 4
+idouble: 3
+ifloat: 4
+ildouble: 5
+ldouble: 5
+
+Function: "yn_towardzero":
+double: 3
+float: 3
+idouble: 3
+ifloat: 3
+ildouble: 5
+ldouble: 5
+
+Function: "yn_upward":
+double: 4
+float: 5
+idouble: 4
+ifloat: 5
+ildouble: 4
+ldouble: 4
+
 # end of automatic generation