Re: [Patch, Fortran] Another floating point speedup

[Tobias Burnus <burnus at net-b dot de>]

Hi Dominique,

Daniel Kraft wrote:
> Dominique Dhumieres wrote:
>> I have two failures with the NIST
>> tests: FM110 and 900. The failures for FM110 can be reduced to
>> character(15) :: str="+ .339567+2"
>> character(15) :: str1="+.339567+2"
>> real :: x
>> read(str1,*) x
>> print *, x
>> read(str,*) x
>> print *, x
>> end
>>
>> where STR gives:
>>
>> At line 6 of file read_tst.f90
>> Fortran runtime error: Bad real number in item 1 of list input
> I don't think this has something to do with it, though, as this error
> message is not generated by my code; and gfortran 4.3 gives this
> error, too, on my system (unless I messed something up).  This looks
> like something to fix with the following up list_read patches, maybe.
It fails here also with other compiles. NAG f95 prints:
  Invalid input for real editing
  Program terminated by I/O error on internal file
and ifort prints:
  forrtl: severe (59): list-directed I/O syntax error, unit -5, file
Internal List-Directed Read
whereas sunf95 shows
  Error 1084:  unexpected character in real value
while g95 accepts it but the result is not as expected: "0." for STR.

(Admittedly I have not checked the standard.)

Thus I think either the NIST test case is wrong or (more likely) while
reducing the failure there was an algorithmic changed which made test
invalid.

Tobias