See test-case in patch referenced soon. All hardware (including i387) I have access to get these numbers "right" (less than 0.5ulp error) on multiplication.
Patch and testcase at <URL:http://gcc.gnu.org/ml/gcc-patches/2004-03/msg00003.html>.
Subject: Bug 14354 CVSROOT: /cvs/gcc Module name: gcc Changes by: hp@gcc.gnu.org 2004-03-05 00:46:06 Modified files: gcc/testsuite : ChangeLog Log message: Forgot to mark PR for <URL:http://gcc.gnu.org/ml/gcc-cvs/2004-03/msg00238.html>. PR other/14354 * gcc.c-torture/execute/ieee/mul-subnormal-single-1.c: New test. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.3570&r2=1.3571
Subject: Bug 14354 CVSROOT: /cvs/gcc Module name: gcc Changes by: hp@gcc.gnu.org 2004-03-05 00:49:22 Modified files: gcc : ChangeLog Log message: Forgot to mark PR for <URL:http://gcc.gnu.org/ml/gcc-cvs/2004-03/msg00237.html> PR other/14354 * config/fp-bit.c (_fpdiv_parts): Do not round when pack_d would round the same. When rounding, clear bits that would cause a second rounding in pack_d. (_fpmul_parts): Ditto. Remove #if 0:d code. Patches: http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.3051&r2=2.3052
Testcase and fix committed.