This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Fix ldbl-128ibm fmodl handling of subnormal results (bug 19595) [committed]
- From: Joseph Myers <joseph at codesourcery dot com>
- To: <libc-alpha at sourceware dot org>
- Date: Thu, 18 Feb 2016 22:42:47 +0000
- Subject: Fix ldbl-128ibm fmodl handling of subnormal results (bug 19595) [committed]
- Authentication-results: sourceware.org; auth=none
The ldbl-128ibm implementation of fmodl has completely bogus logic for
subnormal results (in this context, that means results for which the
result is in the subnormal range for double, not results with absolute
value below LDBL_MIN), based on code used for ldbl-128 that is correct
in that case but incorrect in the ldbl-128ibm use. This patch fixes
it to convert the mantissa into the correct form expected by
ldbl_insert_mantissa, removing the other cases of the code that were
incorrect and in one case unreachable for ldbl-128ibm. A correct
exponent value is then passed to ldbl_insert_mantissa to reflect the
shifted result.
Tested for powerpc. Committed.
2016-02-18 Joseph Myers <joseph@codesourcery.com>
[BZ #19595]
* sysdeps/ieee754/ldbl-128ibm/e_fmodl.c (__ieee754_fmodl): Use
common logic for all cases of shifting subnormal results. Do not
insert sign bit in shifted mantissa. Always pass -1023 as biased
exponent to ldbl_insert_mantissa in subnormal case.
diff --git a/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c b/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c
index 205097d..d756e3e 100644
--- a/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c
+++ b/sysdeps/ieee754/ldbl-128ibm/e_fmodl.c
@@ -130,15 +130,11 @@ __ieee754_fmodl (long double x, long double y)
x = ldbl_insert_mantissa((sx>>63), iy, hx, lx);
} else { /* subnormal output */
n = -1022 - iy;
- if(n<=48) {
- lx = (lx>>n)|((u_int64_t)hx<<(64-n));
- hx >>= n;
- } else if (n<=63) {
- lx = (hx<<(64-n))|(lx>>n); hx = sx;
- } else {
- lx = hx>>(n-64); hx = sx;
- }
- x = ldbl_insert_mantissa((sx>>63), iy, hx, lx);
+ /* We know 1 <= N <= 52, and that there are no nonzero
+ bits in places below 2^-1074. */
+ lx = (lx >> n) | ((u_int64_t) hx << (64 - n));
+ hx >>= n;
+ x = ldbl_insert_mantissa((sx>>63), -1023, hx, lx);
x *= one; /* create necessary signal */
}
return x; /* exact output */
--
Joseph S. Myers
joseph@codesourcery.com