This is the mail archive of the libc-hacker@sourceware.org mailing list for the glibc project.

Note that libc-hacker is a closed list. You may look at the archives of this list, but subscription and posting are not open.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

test-misc is failing for powerpc32/64

From: Steven Munroe <munroesj at us dot ibm dot com>
To: Jakub Jelinek <jakub at redhat dot com>, GNU libc hacker <libc-hacker at sources dot redhat dot com>
Date: Tue, 22 May 2007 13:51:18 -0500
Subject: test-misc is failing for powerpc32/64

Jakub your change:

2007-05-03  Jakub Jelinek  <jakub@redhat.com>

    * soft-fp/op-common.h (FP_TRUNC): Replace raising of FP_EX_INEXACT
    with setting the sticky bit.
    * math/test-misc.c (main): Add more truncation tests.

Is causing failures on powerpc. The denomal rounding test you added
assume that __LDBL_DENORM_MIN__ is less than __DBL_DENORM_MIN__ by at
least 4 bits which is not true for IBM128 long double.

#define __DBL_DENORM_MIN__   4.9406564584124654e-324
#define __LDBL_DENORM_MIN__ 4.94065645841246544176568792868221e-324L

This is implicit in the statement:

    volatile long double ld5 = nextafter (0.0, 1.0) / 16.0L;

This will always return 0.0L for IBM128.

So we could just disable this part of the test for IBM128 with for example

   #if LDBL_MANT_DIG != 106

or replace each usage of  "nextafter (0.0, 1.0)" with LDBL_MIN

#define __LDBL_MIN__ 2.00416836000897277799610805135016e-292L

In IBM128 LDBL_MIN is the where the low order double can be subnormal
but as the high and low double can not overlap, the high order double
must be >= 0x1.0p-969

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]