This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Fix ldbl-128ibm acoshl inaccuracy (bug 18019) [committed]
- From: Joseph Myers <joseph at codesourcery dot com>
- To: <libc-alpha at sourceware dot org>
- Date: Wed, 25 Feb 2015 00:02:24 +0000
- Subject: Fix ldbl-128ibm acoshl inaccuracy (bug 18019) [committed]
- Authentication-results: sourceware.org; auth=none
The ldbl-128ibm implementation of acoshl uses a cut-off of 0x1p28 to
determine when to use log(x) + log(2) as a formula. That cut-off is
too small for this format, resulting in large errors. This patch
changes it to a more appropriate cut-off of 0x1p56, adding tests
around the cut-offs for various floating-point formats.
Tested for powerpc. Also tested for x86_64 and x86 and updated ulps.
Committed.
(auto-libm-test-out diffs omitted below.)
2015-02-24 Joseph Myers <joseph@codesourcery.com>
[BZ #18019]
* sysdeps/ieee754/ldbl-128ibm/e_acoshl.c (__ieee754_acoshl): Use
2**56 not 2**28 as threshold for log (2x) formula.
* math/auto-libm-test-in: Add more tests of acosh.
* math/auto-libm-test-out: Regenerated.
* sysdeps/i386/fpu/libm-test-ulps: Update.
* sysdeps/x86_64/fpu/libm-test-ulps: Likewise.
diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in
index 62e1439..a4bd972 100644
--- a/math/auto-libm-test-in
+++ b/math/auto-libm-test-in
@@ -43,6 +43,37 @@ acosh 1.625
acosh 7
acosh 100
acosh 1e5
+acosh 0x1p8
+acosh 0x1p9
+acosh 0x1p10
+acosh 0x1p11
+acosh 0x1p12
+acosh 0x1p13
+acosh 0x1p24
+acosh 0x1p25
+acosh 0x1p26
+acosh 0x1p27
+acosh 0x1p28
+acosh 0x1p29
+acosh 0x1p30
+acosh 0x1p31
+acosh 0x1p32
+acosh 0x1p33
+acosh 0x1p48
+acosh 0x1p49
+acosh 0x1p50
+acosh 0x1p51
+acosh 0x1p52
+acosh 0x1p53
+acosh 0x1p54
+acosh 0x1p55
+acosh 0x1p56
+acosh 0x1p57
+acosh 0x1p58
+acosh 0x1p59
+acosh 0x1p100
+acosh 0x1p500
+acosh 0x1p5000
acosh max no-test-inline
asin 0
diff --git a/sysdeps/i386/fpu/libm-test-ulps b/sysdeps/i386/fpu/libm-test-ulps
index 5e79307..0977557 100644
--- a/sysdeps/i386/fpu/libm-test-ulps
+++ b/sysdeps/i386/fpu/libm-test-ulps
@@ -21,6 +21,10 @@ Function: "acos_upward":
ildouble: 1
ldouble: 1
+Function: "acosh":
+ildouble: 1
+ldouble: 1
+
Function: "acosh_downward":
ildouble: 2
ldouble: 2
diff --git a/sysdeps/ieee754/ldbl-128ibm/e_acoshl.c b/sysdeps/ieee754/ldbl-128ibm/e_acoshl.c
index b0b33f7..cab1da9 100644
--- a/sysdeps/ieee754/ldbl-128ibm/e_acoshl.c
+++ b/sysdeps/ieee754/ldbl-128ibm/e_acoshl.c
@@ -44,14 +44,14 @@ __ieee754_acoshl(long double x)
EXTRACT_WORDS64 (lx, xlo);
if(hx<0x3ff0000000000000LL) { /* x < 1 */
return (x-x)/(x-x);
- } else if(hx >=0x41b0000000000000LL) { /* x > 2**28 */
+ } else if(hx >=0x4370000000000000LL) { /* x >= 2**56 */
if(hx >=0x7ff0000000000000LL) { /* x is inf of NaN */
return x+x;
} else
return __ieee754_logl(x)+ln2; /* acosh(huge)=log(2x) */
} else if (((hx-0x3ff0000000000000LL)|(lx&0x7fffffffffffffffLL))==0) {
return 0.0; /* acosh(1) = 0 */
- } else if (hx > 0x4000000000000000LL) { /* 2**28 > x > 2 */
+ } else if (hx > 0x4000000000000000LL) { /* 2**56 > x > 2 */
t=x*x;
return __ieee754_logl(2.0*x-one/(x+__ieee754_sqrtl(t-one)));
} else { /* 1<x<2 */
diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps
index 36e1b76..cd9e44f 100644
--- a/sysdeps/x86_64/fpu/libm-test-ulps
+++ b/sysdeps/x86_64/fpu/libm-test-ulps
@@ -29,20 +29,33 @@ ldouble: 1
Function: "acosh":
double: 1
+float: 1
idouble: 1
+ifloat: 1
+ildouble: 1
+ldouble: 1
Function: "acosh_downward":
+double: 1
float: 1
-ildouble: 1
+idouble: 1
+ifloat: 1
+ildouble: 2
ldouble: 2
Function: "acosh_towardzero":
+double: 1
float: 1
-ildouble: 1
+idouble: 1
+ifloat: 1
+ildouble: 2
ldouble: 2
Function: "acosh_upward":
double: 1
+float: 1
+idouble: 1
+ifloat: 1
ildouble: 1
ldouble: 1
--
Joseph S. Myers
joseph@codesourcery.com