This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Fix atanhl missing underflows (bug 16352) [committed]
- From: Joseph Myers <joseph at codesourcery dot com>
- To: <libc-alpha at sourceware dot org>
- Date: Fri, 15 May 2015 22:08:46 +0000
- Subject: Fix atanhl missing underflows (bug 16352) [committed]
- Authentication-results: sourceware.org; auth=none
Similar to various other bugs in this area, some atanh implementations
do not raise the underflow exception for subnormal arguments, when the
result is tiny and inexact. This patch forces the exception in a
similar way to previous fixes. (No change in this regard is needed
for the i386 implementation; special handling to force underflows in
these cases will only be needed there when the spurious underflows,
bug 18049, get fixed.)
Tested for x86_64, x86, powerpc and mips64. Committed.
(auto-libm-test-out diffs omitted below.)
2015-05-15 Joseph Myers <joseph@codesourcery.com>
[BZ #16352]
* sysdeps/i386/fpu/e_atanh.S (dbl_min): New object.
(__ieee754_atanh): Force underflow exception for results with
small absolute value.
* sysdeps/i386/fpu/e_atanhf.S (flt_min): New object.
(__ieee754_atanhf): Force underflow exception for results with
small absolute value.
* sysdeps/ieee754/dbl-64/e_atanh.c: Include <float.h>.
(__ieee754_atanh): Force underflow exception for results with
small absolute value.
* sysdeps/ieee754/flt-32/e_atanhf.c: Include <float.h>.
(__ieee754_atanhf): Force underflow exception for results with
small absolute value.
* sysdeps/ieee754/ldbl-128/e_atanhl.c: Include <float.h>.
(__ieee754_atanhl): Force underflow exception for results with
small absolute value.
* sysdeps/ieee754/ldbl-128ibm/e_atanhl.c: Include <float.h>.
(__ieee754_atanhl): Force underflow exception for results with
small absolute value.
* sysdeps/ieee754/ldbl-96/e_atanhl.c: Include <float.h>.
(__ieee754_atanhl): Force underflow exception for results with
small absolute value.
* math/auto-libm-test-in: Do not allow missing underflow
exceptions from atanh.
* math/auto-libm-test-out: Regenerated.
diff --git a/math/auto-libm-test-in b/math/auto-libm-test-in
index e483569..6fe0ed7 100644
--- a/math/auto-libm-test-in
+++ b/math/auto-libm-test-in
@@ -382,13 +382,12 @@ atanh 0x3.91d9f3c80c72d7acp-4
atanh -0x2.6c52c26567198p-4
atanh 0x3.a274ecp-4
atanh -0x3.f0f519a687b64p-8
-# Bug 16352: underflow exception may be missing.
-atanh 0x1p-500 missing-underflow
-atanh 0x1p-5000 missing-underflow
-atanh min missing-underflow
-atanh -min missing-underflow
-atanh min_subnorm missing-underflow
-atanh -min_subnorm missing-underflow
+atanh 0x1p-500
+atanh 0x1p-5000
+atanh min
+atanh -min
+atanh min_subnorm
+atanh -min_subnorm
# cabs (x,y) == cabs (y,x).
cabs 0.75 12.390625
diff --git a/sysdeps/i386/fpu/e_atanh.S b/sysdeps/i386/fpu/e_atanh.S
index 16e149b..90d19bc 100644
--- a/sysdeps/i386/fpu/e_atanh.S
+++ b/sysdeps/i386/fpu/e_atanh.S
@@ -35,6 +35,13 @@ limit: .double 0.29
ln2_2: .tfloat 0.3465735902799726547086160
ASM_SIZE_DIRECTIVE(ln2_2)
+ .section .rodata.cst8,"aM",@progbits,8
+
+ .p2align 3
+ .type dbl_min,@object
+dbl_min: .byte 0, 0, 0, 0, 0, 0, 0x10, 0
+ ASM_SIZE_DIRECTIVE(dbl_min)
+
#ifdef PIC
#define MO(op) op##@GOTOFF(%edx)
#else
@@ -81,7 +88,18 @@ ENTRY(__ieee754_atanh)
sahf
jae 4f
fyl2xp1 // 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|))
- jecxz 3f
+ fcoml MO(dbl_min)
+ fnstsw
+ sahf
+ jae 8f
+ subl $8, %esp
+ cfi_adjust_cfa_offset (8)
+ fld %st(0)
+ fmul %st(0)
+ fstpl (%esp)
+ addl $8, %esp
+ cfi_adjust_cfa_offset (-8)
+8: jecxz 3f
fchs // 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x))
3: ret
diff --git a/sysdeps/i386/fpu/e_atanhf.S b/sysdeps/i386/fpu/e_atanhf.S
index 0021d9c..1c8969e 100644
--- a/sysdeps/i386/fpu/e_atanhf.S
+++ b/sysdeps/i386/fpu/e_atanhf.S
@@ -36,6 +36,13 @@ limit: .double 0.29
ln2_2: .tfloat 0.3465735902799726547086160
ASM_SIZE_DIRECTIVE(ln2_2)
+ .section .rodata.cst4,"aM",@progbits,4
+
+ .p2align 2
+ .type flt_min,@object
+flt_min: .byte 0, 0, 0x80, 0
+ ASM_SIZE_DIRECTIVE(flt_min)
+
#ifdef PIC
#define MO(op) op##@GOTOFF(%edx)
#else
@@ -77,7 +84,18 @@ ENTRY(__ieee754_atanhf)
sahf
jae 4f
fyl2xp1 // 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|))
- jecxz 3f
+ fcoms MO(flt_min)
+ fnstsw
+ sahf
+ jae 6f
+ subl $4, %esp
+ cfi_adjust_cfa_offset (4)
+ fld %st(0)
+ fmul %st(0)
+ fstps (%esp)
+ addl $4, %esp
+ cfi_adjust_cfa_offset (-4)
+6: jecxz 3f
fchs // 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x))
3: ret
diff --git a/sysdeps/ieee754/dbl-64/e_atanh.c b/sysdeps/ieee754/dbl-64/e_atanh.c
index b95399d..6b00b80 100644
--- a/sysdeps/ieee754/dbl-64/e_atanh.c
+++ b/sysdeps/ieee754/dbl-64/e_atanh.c
@@ -35,6 +35,7 @@
*/
+#include <float.h>
#include <inttypes.h>
#include <math.h>
#include <math_private.h>
@@ -51,6 +52,11 @@ __ieee754_atanh (double x)
if (__glibc_unlikely (xa < 0x1.0p-28))
{
math_force_eval (huge + x);
+ if (fabs (x) < DBL_MIN)
+ {
+ double force_underflow = x * x;
+ math_force_eval (force_underflow);
+ }
return x;
}
diff --git a/sysdeps/ieee754/flt-32/e_atanhf.c b/sysdeps/ieee754/flt-32/e_atanhf.c
index a6d8bd1..bc74960 100644
--- a/sysdeps/ieee754/flt-32/e_atanhf.c
+++ b/sysdeps/ieee754/flt-32/e_atanhf.c
@@ -35,6 +35,7 @@
*/
+#include <float.h>
#include <inttypes.h>
#include <math.h>
#include <math_private.h>
@@ -51,6 +52,11 @@ __ieee754_atanhf (float x)
if (__glibc_unlikely (xa < 0x1.0p-28f))
{
math_force_eval (huge + x);
+ if (fabsf (x) < FLT_MIN)
+ {
+ float force_underflow = x * x;
+ math_force_eval (force_underflow);
+ }
return x;
}
diff --git a/sysdeps/ieee754/ldbl-128/e_atanhl.c b/sysdeps/ieee754/ldbl-128/e_atanhl.c
index c5cceb5..a5a7ee0 100644
--- a/sysdeps/ieee754/ldbl-128/e_atanhl.c
+++ b/sysdeps/ieee754/ldbl-128/e_atanhl.c
@@ -32,6 +32,7 @@
*
*/
+#include <float.h>
#include <math.h>
#include <math_private.h>
@@ -57,7 +58,15 @@ __ieee754_atanhl(long double x)
else
return (x-x)/(x-x);
}
- if(ix<0x3fc60000 && (huge+x)>zero) return x; /* x < 2^-57 */
+ if(ix<0x3fc60000 && (huge+x)>zero) /* x < 2^-57 */
+ {
+ if (fabsl (x) < LDBL_MIN)
+ {
+ long double force_underflow = x * x;
+ math_force_eval (force_underflow);
+ }
+ return x;
+ }
if(ix<0x3ffe0000) { /* x < 0.5 */
t = u.value+u.value;
diff --git a/sysdeps/ieee754/ldbl-128ibm/e_atanhl.c b/sysdeps/ieee754/ldbl-128ibm/e_atanhl.c
index 5a98999..bcd1fce 100644
--- a/sysdeps/ieee754/ldbl-128ibm/e_atanhl.c
+++ b/sysdeps/ieee754/ldbl-128ibm/e_atanhl.c
@@ -28,6 +28,7 @@
*
*/
+#include <float.h>
#include <math.h>
#include <math_private.h>
@@ -54,7 +55,15 @@ __ieee754_atanhl(long double x)
if (t == one)
return x/zero;
}
- if(ix<0x3c70000000000000LL&&(huge+x)>zero) return x; /* x<2**-56 */
+ if(ix<0x3c70000000000000LL&&(huge+x)>zero) /* x<2**-56 */
+ {
+ if (fabsl (x) < LDBL_MIN)
+ {
+ long double force_underflow = x * x;
+ math_force_eval (force_underflow);
+ }
+ return x;
+ }
x = fabsl (x);
if(ix<0x3fe0000000000000LL) { /* x < 0.5 */
t = x+x;
diff --git a/sysdeps/ieee754/ldbl-96/e_atanhl.c b/sysdeps/ieee754/ldbl-96/e_atanhl.c
index 305d50e..9a957c9 100644
--- a/sysdeps/ieee754/ldbl-96/e_atanhl.c
+++ b/sysdeps/ieee754/ldbl-96/e_atanhl.c
@@ -32,6 +32,7 @@
*
*/
+#include <float.h>
#include <math.h>
#include <math_private.h>
@@ -54,6 +55,11 @@ __ieee754_atanhl(long double x)
return x/zero;
if(ix<0x3fdf) {
math_force_eval(huge+x);
+ if (fabsl (x) < LDBL_MIN)
+ {
+ long double force_underflow = x * x;
+ math_force_eval (force_underflow);
+ }
return x; /* x<2**-32 */
}
SET_LDOUBLE_EXP(x,ix);
--
Joseph S. Myers
joseph@codesourcery.com