Re: Test results for m4-1.4.9b

Gary V. Vaughan Mon, 04 Jun 2007 08:35:53 -0700

Hi Bruno,

On 4 Jun 2007, at 12:42, Bruno Haible wrote:

Gary V. Vaughan wrote:
Run till exit from #0 my_ldexp (x=1.0121050218045312617883299999606451e-292, d=969) at test- printf- frexpl.c:56
This value x has the mantissa 0x2.051EB851EB8520000000000000. Which
means, the hardware truncates the precision when converting a number from
normalized to denormalized range.

Can you try this patch, please? (Not yet committed in CVS.)


*** tests/test-printf-frexpl.c  29 Apr 2007 09:15:13 -0000      1.4
--- tests/test-printf-frexpl.c  4 Jun 2007 11:41:57 -0000
***************
*** 39,49 ****
    while (0)
/* On MIPS IRIX machines, LDBL_MIN_EXP is -1021, but the smallest reliable ! exponent for 'long double' is -964. For exponents below that, the ! precision may be truncated to the precision used for 'double'. */
  #ifdef __sgi
  # define MIN_NORMAL_EXP (LDBL_MIN_EXP + 57)
  # define MIN_SUBNORMAL_EXP MIN_NORMAL_EXP
  #else
  # define MIN_NORMAL_EXP LDBL_MIN_EXP
  # define MIN_SUBNORMAL_EXP (LDBL_MIN_EXP - 100)
--- 39,54 ----
    while (0)
/* On MIPS IRIX machines, LDBL_MIN_EXP is -1021, but the smallest reliable ! exponent for 'long double' is -964. Similarly, on PowerPC machines, ! LDBL_MIN_EXP is -1021, but the smallest reliable exponent for 'long double' ! is -968. For exponents below that, the precision may be truncated to the
!    precision used for 'double'.  */
  #ifdef __sgi
  # define MIN_NORMAL_EXP (LDBL_MIN_EXP + 57)
  # define MIN_SUBNORMAL_EXP MIN_NORMAL_EXP
+ #elif defined __ppc || defined __ppc__ || defined __powerpc || defined __powerpc__
+ # define MIN_NORMAL_EXP (LDBL_MIN_EXP + 53)
+ # define MIN_SUBNORMAL_EXP MIN_NORMAL_EXP
  #else
  # define MIN_NORMAL_EXP LDBL_MIN_EXP
  # define MIN_SUBNORMAL_EXP (LDBL_MIN_EXP - 100)


That fixes test-frexpl.c on my ppc machine.

*** tests/test-frexpl.c 29 Apr 2007 09:15:13 -0000      1.6
--- tests/test-frexpl.c 4 Jun 2007 11:41:57 -0000
***************
*** 40,49 ****
    while (0)
/* On MIPS IRIX machines, LDBL_MIN_EXP is -1021, but the smallest reliable ! exponent for 'long double' is -964. For exponents below that, the ! precision may be truncated to the precision used for 'double'. */
  #ifdef __sgi
  # define MIN_NORMAL_EXP (LDBL_MIN_EXP + 57)
  #else
  # define MIN_NORMAL_EXP LDBL_MIN_EXP
  #endif
--- 40,53 ----
    while (0)
/* On MIPS IRIX machines, LDBL_MIN_EXP is -1021, but the smallest reliable ! exponent for 'long double' is -964. Similarly, on PowerPC machines, ! LDBL_MIN_EXP is -1021, but the smallest reliable exponent for 'long double' ! is -968. For exponents below that, the precision may be truncated to the
!    precision used for 'double'.  */
  #ifdef __sgi
  # define MIN_NORMAL_EXP (LDBL_MIN_EXP + 57)
+ #elif defined __ppc || defined __ppc__ || defined __powerpc || defined __powerpc__
+ # define MIN_NORMAL_EXP (LDBL_MIN_EXP + 53)
  #else
  # define MIN_NORMAL_EXP LDBL_MIN_EXP
  #endif

This doesn't have chance to come into play, as the loop aborts in an assertion

failure on the first iteration:

(gdb) run
Starting program: /Users/gary/m4-1.4.9b/tests/test-frexpl

Breakpoint 1, main () at test-frexpl.c:160
160       for (i = 1, x = 1.01L; i <= LDBL_MAX_EXP; i++, x *= 2.0L)
(gdb) n
162           int exp = -9999;
(gdb)
163           long double mantissa = frexpl (x, &exp);
(gdb)
164           ASSERT (exp == i);
(gdb)
165           ASSERT (mantissa == 0.505L);
(gdb) p mantissa
$1 = 0.50499999999999989341858963598497212
(gdb) p exp
$2 = 1
(gdb) p i
$3 = 1
(gdb) p x
$4 = 1.0100000000000000088817841970012523
(gdb) n
test-frexpl.c:165: assertion failed

Program received signal SIGABRT, Aborted.
0x9004822c in kill ()

Cheers,
        Gary
--
  ())_.              Email me: [EMAIL PROTECTED]
  ( '/           Read my blog: http://blog.azazil.net
  / )=         ...and my book: http://sources.redhat.com/autobook
`(_~)_ Join my AGLOCO Network: http://www.agloco.com/r/BBBS7912

PGP.sig
Description: This is a digitally signed message part

Re: Test results for m4-1.4.9b

Reply via email to