fortran, IVF continues to disappoint (relatively)

awgreyno..
#1 / 25

IVF continues to disappoint (relatively)

I just downloaded the latest Intel Visual fortran 9.1 evaluation.
Right off the bat it choked on some standard Fortran:

exports.f90(849) : Error: Unbalanced parentheses
write(lu,'(3(1x,'//c(:3)//'.7),:,",")')
((plane(:,n,j)+gelem(:,i,n,j),i=0,MentPoints(j)),j=l1,l2)
------------------------------------------------^
exports.f90(849) : Error: Unbalanced parentheses
write(lu,'(3(1x,'//c(:3)//'.7),:,",")')
((plane(:,n,j)+gelem(:,i,n,j),i=0,MentPoints(j)),j=l1,l2)
-------------------------------------------------------------------------------------------------^
exports.f90(849) : Error: Unbalanced parentheses
write(lu,'(3(1x,'//c(:3)//'.7),:,",")')
((plane(:,n,j)+gelem(:,i,n,j),i=0,MentPoints(j)),j=l1,l2)
----------------------------------------------------------------------------------------------------------^

where "c" is a character variable. Of the many other commercial and
free Fortran-95 compilers I use, none have a problem with this code.

I was a big fan of DVF/CVF, but ever since it became IVF, I have not
seen any significant improvement in the quality of the compiler. Yes,
it does produce the fastest code on the latest Intel processors, but
really not that much faster than the free G95 (don't pay any attentiion
to the Polyhedron benchmarks since they are using a very out-of-date
version of G95).

Al Greynolds

P.S. To Steve Lionel: I will formally submit this bug when I get a
chance.

Fri, 31 Oct 2008 00:19:06 GMT

Dick Hendrickso
#2 / 25

IVF continues to disappoint (relatively)

From the way the error carets are pointing, it looks like
a source form error. Are you sure you've got the free/fixed
source form choice correctly set? If so, some of the lines
look awful long, are you sure you're not running past column
71/132? It's hard to tell in this post how your lines join.
At least, my newsreader sometimes wraps lines and sometimes
doesn't. Makes it hard for me to see where they really end.

Dick Hendrickson

Quote:

> I just downloaded the latest Intel Visual Fortran 9.1 evaluation.
> Right off the bat it choked on some standard Fortran:

> exports.f90(849) : Error: Unbalanced parentheses
> write(lu,'(3(1x,'//c(:3)//'.7),:,",")')
> ((plane(:,n,j)+gelem(:,i,n,j),i=0,MentPoints(j)),j=l1,l2)
> ------------------------------------------------^
> exports.f90(849) : Error: Unbalanced parentheses
> write(lu,'(3(1x,'//c(:3)//'.7),:,",")')
> ((plane(:,n,j)+gelem(:,i,n,j),i=0,MentPoints(j)),j=l1,l2)
> -------------------------------------------------------------------------------------------------^
> exports.f90(849) : Error: Unbalanced parentheses
> write(lu,'(3(1x,'//c(:3)//'.7),:,",")')
> ((plane(:,n,j)+gelem(:,i,n,j),i=0,MentPoints(j)),j=l1,l2)
> ----------------------------------------------------------------------------------------------------------^

> where "c" is a character variable. Of the many other commercial and
> free Fortran-95 compilers I use, none have a problem with this code.

> I was a big fan of DVF/CVF, but ever since it became IVF, I have not
> seen any significant improvement in the quality of the compiler. Yes,
> it does produce the fastest code on the latest Intel processors, but
> really not that much faster than the free G95 (don't pay any attentiion
> to the Polyhedron benchmarks since they are using a very out-of-date
> version of G95).

> Al Greynolds

> P.S. To Steve Lionel: I will formally submit this bug when I get a
> chance.

Fri, 31 Oct 2008 01:10:40 GMT

awgreyno..
#3 / 25

IVF continues to disappoint (relatively)

The file has a ".f90" extension which I assume implies free-form
source. Right? IVF doesn't {*filter*}on the hundreds of other lines in the
file that go past column 72, start statements before column 7, and also
use the trailing "&" for continuation.

BTW, I get the error in other files only when a character expression
involving concatenation is used as a format specifier. So I'm pretty
sure its a "basic" problem with their syntax parser.

Fri, 31 Oct 2008 01:51:37 GMT

Dan Nagl
#4 / 25

IVF continues to disappoint (relatively)

Hello,

<snip>

Quote:

> So I'm pretty
> sure its a "basic" problem with their syntax parser.

I'd be very surprised if any commercial compiler
has a "basic" problem with its parser.

Have you tried something like cut -c1-132 | diff ?

--
Cheers!

Dan Nagle
Purple Sage Computing Solutions, Inc.

Fri, 31 Oct 2008 02:19:30 GMT

awgreyno..
#5 / 25

IVF continues to disappoint (relatively)

Dan,

Thanks for introducing me to the cut command. The subsequent diff
turned up nothing, so all my lines are 132 characters or less.

Again, I want to stress that the exact same code compiles and runs
without a hitch using several other Fortran-95 compilers (lf95, g95,
xlf) on a couple different platforms (Win32, MacOSX-PPC).

Al

Fri, 31 Oct 2008 03:13:27 GMT

Steve Lione
#6 / 25

IVF continues to disappoint (relatively)

Quote:

> I was a big fan of DVF/CVF, but ever since it became IVF, I have not
> seen any significant improvement in the quality of the compiler. Yes,
> it does produce the fastest code on the latest Intel processors, but
> really not that much faster than the free G95 (don't pay any attentiion
> to the Polyhedron benchmarks since they are using a very out-of-date
> version of G95).

Well, I could come up with hundreds of bugs that CVF had that we've
fixed and many other asked-for features that we've implemented. I'll be
waiting for the bug report and the actual source.

Steve

Fri, 31 Oct 2008 09:41:04 GMT

Gary L. Scot
#7 / 25

IVF continues to disappoint (relatively)

Quote:

>>I was a big fan of DVF/CVF, but ever since it became IVF, I have not
>>seen any significant improvement in the quality of the compiler. Yes,
>>it does produce the fastest code on the latest Intel processors, but
>>really not that much faster than the free G95 (don't pay any attentiion
>>to the Polyhedron benchmarks since they are using a very out-of-date
>>version of G95).

> Well, I could come up with hundreds of bugs that CVF had that we've
> fixed and many other asked-for features that we've implemented. I'll be
> waiting for the bug report and the actual source.

But, had the CVF line continued, I'm certain that most if not all of
those compiler bugs would have been fixed as well in that line. The
fact that IVF is overall a diminished product from CVF remains and no
amount of compiler perfecting or modest execution performance
improvements will change that.

Quote:

> Steve

--

Gary Scott

Fortran Library: http://www.fortranlib.com

Support the Original G95 Project: http://www.g95.org
-OR-
Support the GNU GFortran Project: http://gcc.gnu.org/fortran/index.html

Why are there two? God only knows.

If you want to do the impossible, don't hire an expert because he knows
it can't be done.

-- Henry Ford

Fri, 31 Oct 2008 10:26:06 GMT

awgreyno..
#8 / 25

IVF continues to disappoint (relatively)

Steve,

I registered for support at the time of the download, but I haven't
been able to login several hours later. Does it take a while before an
"account" is activated? I'll file the bug report with a small test
case when I'm able to login.

Al

Fri, 31 Oct 2008 10:28:43 GMT

awgreyno..
#9 / 25

IVF continues to disappoint (relatively)

Here's the complete source for the smallest routine in my application
that generates the error. I encourage anyone to compile it on their
own installation of IVF and any other Fortran-95 compiler.

subroutine RealArrayStats(name,list,m,array,n,mask,stats)
implicit real(8) (a-h,o-z)
dimension array(m,n),stats(m,-1:2)
intent(in) name,list,m,array,n,mask; intent(out) stats
character(*) name,list; logical mask(n)
!
external ListParse
character(*),parameter :: FMT='g13.7'
dimension a(m),mloc(m,-1:1)
logical msk(m,n)
dimension l(100)
s=count(mask); if (s==0) return
msk=spread(mask,1,m); mloc(:,0)=0
mloc(:,-1)=minloc(array,2,msk); forall (i=1:m)
stats(i,-1)=array(i,mloc(i,-1))
a=sum(array,2,msk)/s; stats(:,0)=a
mloc(:,1)=maxloc(array,2,msk); forall (i=1:m)
stats(i,1)=array(i,mloc(i,1))
stats(:,2)=sqrt(sum((array-spread(a,2,n))**2,2,msk)/s)
if (list==' ') return
k=ListParse(list,',',l); j=len_trim(list)/m
if (k==1.and.m>1) then
j=1; l=(/(i,i=0,99)/)
endif
write(*,'(a,5(3x,a7,4x))') repeat('
',j),'Minimum','Average','Maximum',' RMSvar','Max-Min'
do i=1,m
if (real(stats(i,-1))==real(stats(i,1))) then
write(*,'(a,1x,'//FMT//')') list(l(i)+1:l(i)+j),stats(i,-1)
else
write(*,'(a,5(1x,'//FMT//'))')
list(l(i)+1:l(i)+j),stats(i,:),stats(i,1)-stats(i,-1)
write(*,'(a,3(i10.0,4x))') repeat(' ',j),mloc(i,:)
endif
enddo !i
return
end

Fri, 31 Oct 2008 12:12:49 GMT

Tim Princ
#10 / 25

IVF continues to disappoint (relatively)

Quote:

> Here's the complete source for the smallest routine in my application
> that generates the error. I encourage anyone to compile it on their
> own installation of IVF and any other Fortran-95 compiler.

> subroutine RealArrayStats(name,list,m,array,n,mask,stats)
> implicit real(8) (a-h,o-z)
> dimension array(m,n),stats(m,-1:2)
> intent(in) name,list,m,array,n,mask; intent(out) stats
> character(*) name,list; logical mask(n)
> !
> external ListParse
> character(*),parameter :: FMT='g13.7'
> dimension a(m),mloc(m,-1:1)
> logical msk(m,n)
> dimension l(100)
> s=count(mask); if (s==0) return
> msk=spread(mask,1,m); mloc(:,0)=0
> mloc(:,-1)=minloc(array,2,msk); forall (i=1:m)
> stats(i,-1)=array(i,mloc(i,-1))
> a=sum(array,2,msk)/s; stats(:,0)=a
> mloc(:,1)=maxloc(array,2,msk); forall (i=1:m)
> stats(i,1)=array(i,mloc(i,1))
> stats(:,2)=sqrt(sum((array-spread(a,2,n))**2,2,msk)/s)
> if (list==' ') return
> k=ListParse(list,',',l); j=len_trim(list)/m
> if (k==1.and.m>1) then
> j=1; l=(/(i,i=0,99)/)
> endif
> write(*,'(a,5(3x,a7,4x))') repeat('
> ',j),'Minimum','Average','Maximum',' RMSvar','Max-Min'
> do i=1,m
> if (real(stats(i,-1))==real(stats(i,1))) then
> write(*,'(a,1x,'//FMT//')') list(l(i)+1:l(i)+j),stats(i,-1)
> else
> write(*,'(a,5(1x,'//FMT//'))')
> list(l(i)+1:l(i)+j),stats(i,:),stats(i,1)-stats(i,-1)
> write(*,'(a,3(i10.0,4x))') repeat(' ',j),mloc(i,:)
> endif
> enddo !i
> return
> end

$ gfortran -O -Wall al.f90
In file al.f90:20

if (list==' ') return
1
Error: Unexpected Simple IF statement in FORALL block at (1)
In file al.f90:22

if (k==1.and.m>1) then
1
Error: Unexpected block IF statement in FORALL block at (1)
In file al.f90:24

endif
1
Error: Expecting END FORALL statement at (1)
In file al.f90:25

write(*,'(a,5(3x,a7,4x))') repeat('
1
Error: Unterminated character constant beginning at (1)
In file al.f90:26

',j),'Minimum','Average','Maximum',' RMSvar','Max-Min'
1
Error: Unclassifiable statement at (1)
In file al.f90:27

do i=1,m
1
Error: Unexpected DO statement in FORALL block at (1)
In file al.f90:28

if (real(stats(i,-1))==real(stats(i,1))) then
1
Error: Unexpected block IF statement in FORALL block at (1)
In file al.f90:29

write(*,'(a,1x,'//FMT//')') list(l(i)+1:l(i)+j),stats

Error: Unexpected WRITE statement in FORALL block at (1)
In file al.f90:30

else
1
Error: Unexpected ELSE statement in FORALL block at (1)
In file al.f90:31

write(*,'(a,5(1x,'//FMT//'))')
1
Error: Unexpected WRITE statement in FORALL block at (1)
In file al.f90:32

list(l(i)+1:l(i)+j),stats(i,:),stats(i,1)-stats(i,-1)
1
Error: Unclassifiable statement at (1)
In file al.f90:33

write(*,'(a,3(i10.0,4x))') repeat(' ',j),mloc(i,:)
1
Error: Unexpected WRITE statement in FORALL block at (1)
In file al.f90:34

endif
1
Error: Expecting END FORALL statement at (1)
In file al.f90:35

enddo !i
1
Error: Expecting END FORALL statement at (1)
In file al.f90:36

return
1
Error: Unexpected RETURN statement in FORALL block at (1)
In file al.f90:37

end
1
Error: END FORALL statement expected at (1)
Error: Unexpected end of file in 'al.f90'

Fri, 31 Oct 2008 12:37:10 GMT

awgreyno..
#11 / 25

IVF continues to disappoint (relatively)

The Internet Explorer/Google Groups client obviously wrapped some of
the source code lines. Its not too hard to figure out which ones (2
foralls and 1 write), so unwrap them and try again.

Quote:

> > Here's the complete source for the smallest routine in my application
> > that generates the error. I encourage anyone to compile it on their
> > own installation of IVF and any other Fortran-95 compiler.

> > subroutine RealArrayStats(name,list,m,array,n,mask,stats)
> > implicit real(8) (a-h,o-z)
> > dimension array(m,n),stats(m,-1:2)
> > intent(in) name,list,m,array,n,mask; intent(out) stats
> > character(*) name,list; logical mask(n)
> > !
> > external ListParse
> > character(*),parameter :: FMT='g13.7'
> > dimension a(m),mloc(m,-1:1)
> > logical msk(m,n)
> > dimension l(100)
> > s=count(mask); if (s==0) return
> > msk=spread(mask,1,m); mloc(:,0)=0
> > mloc(:,-1)=minloc(array,2,msk); forall (i=1:m)
> > stats(i,-1)=array(i,mloc(i,-1))
> > a=sum(array,2,msk)/s; stats(:,0)=a
> > mloc(:,1)=maxloc(array,2,msk); forall (i=1:m)
> > stats(i,1)=array(i,mloc(i,1))
> > stats(:,2)=sqrt(sum((array-spread(a,2,n))**2,2,msk)/s)
> > if (list==' ') return
> > k=ListParse(list,',',l); j=len_trim(list)/m
> > if (k==1.and.m>1) then
> > j=1; l=(/(i,i=0,99)/)
> > endif
> > write(*,'(a,5(3x,a7,4x))') repeat('
> > ',j),'Minimum','Average','Maximum',' RMSvar','Max-Min'
> > do i=1,m
> > if (real(stats(i,-1))==real(stats(i,1))) then
> > write(*,'(a,1x,'//FMT//')') list(l(i)+1:l(i)+j),stats(i,-1)
> > else
> > write(*,'(a,5(1x,'//FMT//'))')
> > list(l(i)+1:l(i)+j),stats(i,:),stats(i,1)-stats(i,-1)
> > write(*,'(a,3(i10.0,4x))') repeat(' ',j),mloc(i,:)
> > endif
> > enddo !i
> > return
> > end

> $ gfortran -O -Wall al.f90
> In file al.f90:20

> if (list==' ') return
> 1
> Error: Unexpected Simple IF statement in FORALL block at (1)
> In file al.f90:22

> if (k==1.and.m>1) then
> 1
> Error: Unexpected block IF statement in FORALL block at (1)
> In file al.f90:24

> endif
> 1
> Error: Expecting END FORALL statement at (1)
> In file al.f90:25

> write(*,'(a,5(3x,a7,4x))') repeat('
> 1
> Error: Unterminated character constant beginning at (1)
> In file al.f90:26

> ',j),'Minimum','Average','Maximum',' RMSvar','Max-Min'
> 1
> Error: Unclassifiable statement at (1)
> In file al.f90:27

> do i=1,m
> 1
> Error: Unexpected DO statement in FORALL block at (1)
> In file al.f90:28

> if (real(stats(i,-1))==real(stats(i,1))) then
> 1
> Error: Unexpected block IF statement in FORALL block at (1)
> In file al.f90:29

> write(*,'(a,1x,'//FMT//')') list(l(i)+1:l(i)+j),stats

> Error: Unexpected WRITE statement in FORALL block at (1)
> In file al.f90:30

> else
> 1
> Error: Unexpected ELSE statement in FORALL block at (1)
> In file al.f90:31

> write(*,'(a,5(1x,'//FMT//'))')
> 1
> Error: Unexpected WRITE statement in FORALL block at (1)
> In file al.f90:32

> list(l(i)+1:l(i)+j),stats(i,:),stats(i,1)-stats(i,-1)
> 1
> Error: Unclassifiable statement at (1)
> In file al.f90:33

> write(*,'(a,3(i10.0,4x))') repeat(' ',j),mloc(i,:)
> 1
> Error: Unexpected WRITE statement in FORALL block at (1)
> In file al.f90:34

> endif
> 1
> Error: Expecting END FORALL statement at (1)
> In file al.f90:35

> enddo !i
> 1
> Error: Expecting END FORALL statement at (1)
> In file al.f90:36

> return
> 1
> Error: Unexpected RETURN statement in FORALL block at (1)
> In file al.f90:37

> end
> 1
> Error: END FORALL statement expected at (1)
> Error: Unexpected end of file in 'al.f90'

Fri, 31 Oct 2008 12:43:58 GMT

Tim Princ
#12 / 25

IVF continues to disappoint (relatively)

Quote:

> The Internet Explorer/Google Groups client obviously wrapped some of
> the source code lines. Its not too hard to figure out which ones (2
> foralls and 1 write), so unwrap them and try again.

Don't use broken methods of posting, and don't top post if you're trying
to prove a point. How do we guess which errors you want us to undo, and
which you think the compiler should undo?

Fri, 31 Oct 2008 12:48:39 GMT

Dan Nagl
#13 / 25

IVF continues to disappoint (relatively)

Hello,

When you compile your code, do you use the option
to generate a warning on all non-standard usage?
That's helpful when you want the code to work
with many compilers.

Quote:

> The Internet Explorer/Google Groups client obviously wrapped some of
> the source code lines. Its not too hard to figure out which ones (2
> foralls and 1 write), so unwrap them and try again.

--
Cheers!

Dan Nagle
Purple Sage Computing Solutions, Inc.

Fri, 31 Oct 2008 19:11:55 GMT

awgreyno..
#14 / 25

IVF continues to disappoint (relatively)

Dan,

I set Fortran-95 conformance checking on 4 different compilers
(including IVF) and got no warning messages from any. Only IVF aborts
on multiple "Unbalanced parentheses" errors. Here's the test case
again (without I hope any line wrapping problems):

subroutine RealArrayStats(name,list,m,array,n,mask,stats)
implicit real(8) (a-h,o-z)
dimension array(m,n),stats(m,-1:2)
intent(in) name,list,m,array,n,mask; intent(out) stats
character(*) name,list; logical mask(n)
!
external ListParse
character(*),parameter :: FMT='g13.7'
dimension a(m),mloc(m,-1:1)
logical msk(m,n)
dimension l(100)
s=count(mask); if (s==0) return
msk=spread(mask,1,m); mloc(:,0)=0
mloc(:,-1)=minloc(array,2,msk)
forall (i=1:m) stats(i,-1)=array(i,mloc(i,-1))
a=sum(array,2,msk)/s; stats(:,0)=a
mloc(:,1)=maxloc(array,2,msk)
forall (i=1:m) stats(i,1)=array(i,mloc(i,1))
stats(:,2)=sqrt(sum((array-spread(a,2,n))**2,2,msk)/s)
if (list==' ') return
k=ListParse(list,',',l); j=len_trim(list)/m
if (k==1.and.m>1) then
j=1; l=(/(i,i=0,99)/)
endif
write(*,'(a,5(3x,a7,4x))') &
repeat(' ',j),'Minimum','Average','Maximum',' RMSvar','Max-Min'
do i=1,m
if (real(stats(i,-1))==real(stats(i,1))) then
write(*,'(a,1x,'//FMT//')') list(l(i)+1:l(i)+j),stats(i,-1)
else
write(*,'(a,5(1x,'//FMT//'))') &
list(l(i)+1:l(i)+j),stats(i,:),stats(i,1)-stats(i,-1)
write(*,'(a,3(i10.0,4x))') repeat(' ',j),mloc(i,:)
endif
enddo !i
return
end

Fri, 31 Oct 2008 20:44:13 GMT

Herman D. Knobl
#15 / 25

IVF continues to disappoint (relatively)

Lahey FTN95 (5.70f) and G95 (5 May version)and CVF (v6.6C)
find nothing wrong with this program's syntax.

Skip Knoble

Fri, 31 Oct 2008 20:52:10 GMT