[PATCH] BZ#2466 More errors in long double (ldbl-128ibm) rounding
Steven Munroe
munroesj@us.ibm.com
Thu Mar 16 20:47:00 GMT 2006
The implementations of the (IBM double double format) long double
(./ldbld-128ibm/) for llrintl and llroundl, round incorrectly
return incorrect values when the integer portion is contained in
the high double, but the fractional postion is contained in the
low double. There are also problems where the input value
approaches __LONG_LONG_MAX__.
Also the IEEE 1003.1 spec for llrintl states "These functions
shall round their argument to the nearest integer value, rounding
according to the current rounding direction." So I added tests
to libm-test.inc to cover llrint in the four rounding modes
(FE_TONEAREST, FE_TOWARDZERO, FE_UPWARD, FE_DOWNWARD) in addition
to the default rounding.
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: ppc-llroundl-20060315.txt
URL: <http://sourceware.org/pipermail/libc-alpha/attachments/20060316/c483e1c5/attachment.txt>
More information about the Libc-alpha
mailing list