G77: bitwise .AND. ?

G77: bitwise .AND. ?

[Hassan Aurag]

 Hi,


 this is a small question. On IBM with xlf, if I do

	a = b .and. ( c .xor. TRUE ) 

 where TRUE = X'01010101', and a, b, c logical*4's equivalenced to 4
logical*1's, I'd get the correct info in a, eg:

 a = X'01000100' for example.

 But here, I'd only get the boolean .AND. result so for example
X'00000001'.

 Any options or ways of making g77 behave in the same way?


 Thanks in advance for any help

Re: G77: bitwise .AND. ?

[Toon Moene]

Hassan Aurag wrote:

>  this is a small question. On IBM with xlf, if I do
> 
>         a = b .and. ( c .xor. TRUE )
> 
>  where TRUE = X'01010101', and a, b, c logical*4's equivalenced to 4
> logical*1's, I'd get the correct info in a, eg:
> 
>  a = X'01000100' for example.
> 
>  But here, I'd only get the boolean .AND. result so for example
> X'00000001'.
> 
>  Any options or ways of making g77 behave in the same way?

Please supply a complete, compilable example and I'll try to answer this
question.

Thanks,

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)

Re: G77: bitwise .AND. ?

[Hassan Aurag]

[-- Attachment #1: Type: text/plain, Size: 1077 bytes --]

Here you go:

if you compile the attached file on Linux using 'g77 -fugly-logint' and
run it, you get:

T F F F

On AIX (RISC) with 'xlf -qintlog', you get:

T F T F

Thanks in advance

On Tue, 2002-04-23 at 17:23, Toon Moene wrote:
> Hassan Aurag wrote:
> 
> >  this is a small question. On IBM with xlf, if I do
> > 
> >         a = b .and. ( c .xor. TRUE )
> > 
> >  where TRUE = X'01010101', and a, b, c logical*4's equivalenced to 4
> > logical*1's, I'd get the correct info in a, eg:
> > 
> >  a = X'01000100' for example.
> > 
> >  But here, I'd only get the boolean .AND. result so for example
> > X'00000001'.
> > 
> >  Any options or ways of making g77 behave in the same way?
> 
> Please supply a complete, compilable example and I'll try to answer this
> question.
> 
> Thanks,
> 
> -- 
> Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
> Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
> Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
> Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)


[-- Attachment #2: ands.f --]
[-- Type: text/x-fortran, Size: 389 bytes --]

        program parallel
        logical*4 l4, ll4
        logical*1 l1(4), ll1(4)
        logical*4 true /X'01010101'/
        integer i
        equivalence(l1, l4), (ll1, ll4)

        ll1(1) = .true.
        ll1(3) = .true.

        l4 = ll4 .and. true

        do i =1, 4
                print *, l1(i)
        enddo
        end
       

        

Re: G77: bitwise .AND. ?

[Toon Moene]

Hassan Aurag wrote:

> if you compile the attached file on Linux using 'g77 -fugly-logint' and
> run it, you get:
> 
> T F F F
> 
> On AIX (RISC) with 'xlf -qintlog', you get:
> 
> T F T F

I think the problem is that g77 considers

      l4 = ll4 .and. true

a logical expression - which will have the value 0 or 1 - because all
the elements are logicals (so in spite of the -fugly-logint).

I don't know an easy way out of this, unfortunately.

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)

Re: G77: bitwise .AND. ?

[Hassan Aurag]

Ok the problem below is fixed by doing

and(ll4, true) instead of ll4.and.true


In my opinion, this is a bug in G77.



On Wed, 2002-04-24 at 17:31, Toon Moene wrote:
> Hassan Aurag wrote:
> 
> > if you compile the attached file on Linux using 'g77 -fugly-logint' and
> > run it, you get:
> > 
> > T F F F
> > 
> > On AIX (RISC) with 'xlf -qintlog', you get:
> > 
> > T F T F
> 
> I think the problem is that g77 considers
> 
>       l4 = ll4 .and. true
> 
> a logical expression - which will have the value 0 or 1 - because all
> the elements are logicals (so in spite of the -fugly-logint).
> 
> I don't know an easy way out of this, unfortunately.
> 
> -- 
> Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
> Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
> Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
> Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)

Re: G77: bitwise .AND. ?

[Toon Moene]

Hassan Aurag wrote:

> Ok the problem below is fixed by doing
> 
> and(ll4, true) instead of ll4.and.true
> 
> In my opinion, this is a bug in G77.

Yes, it could be argued that if we want to support -fugly-logint that
the code you supplied should work as-is.

Could you file a bug report using GNATS ?

Thanks in advance,

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)

RE: G77: bitwise .AND. ?

[Hassan Aurag]

 I'll do that. It is also the same with OR and someone told me even
XOR??!!!!


-----Original Message-----
From: Toon Moene [mailto:toon@moene.indiv.nluug.nl]
Sent: Friday, April 26, 2002 4:02 PM
To: Hassan Aurag
Cc: Hassan Aurag; gcc-help@gcc.gnu.org; Remi Simard
Subject: Re: G77: bitwise .AND. ?


Hassan Aurag wrote:

> Ok the problem below is fixed by doing
> 
> and(ll4, true) instead of ll4.and.true
> 
> In my opinion, this is a bug in G77.

Yes, it could be argued that if we want to support -fugly-logint that
the code you supplied should work as-is.

Could you file a bug report using GNATS ?

Thanks in advance,

-- 
Toon Moene - mailto:toon@moene.indiv.nluug.nl - phoneto: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands
Maintainer, GNU Fortran 77: http://gcc.gnu.org/onlinedocs/g77_news.html
Join GNU Fortran 95: http://g95.sourceforge.net/ (under construction)