Re: Error compiling with ifort 9.1

pi2002 · ‎06-03-2006

Dear IFC developers,

I'm trying to compile a code with ifc 9.1.032 (em64t). I encountered some errors, which are very similar to one old post in this board here:

http://softwareforums.intel.com/ids/board/message?board.id=11&message.id=2290

I didn't get any error message while compiling with ifc 9.0.

Any suggestions?

#ifort ./para_p.f -o ./para_p.o

fortcom: Error: ./para_p.f, line 136: Unbalanced parentheses

&'" with ",I3," procs): I am the superparent. ")')

----------------------------------------------------------------^

fortcom: Error: ./para_p.f, line 136: Syntax error, found ''' when expecting one of: ( ...

&'" with ",I3," procs): I am the superparent. ")')

----------------^

fortcom: Error: ./para_p.f, line 136: Syntax error, found ''' when expecting one of: ( , ; ...

&'" with ",I3," procs): I am the superparent. ")')

---------------------------------------------------------------^

fortcom: Error: ./para_p.f, line 140: Syntax error, found ''' when expecting one of: ( ...

&'" with ",I3," procs): Output --> ",A5,".")')

----------------^

fortcom: Error: ./para_p.f, line 141: Syntax error, found IDENTIFIER 'UT_' when expecting one of: .EQV. .NEQV. .XOR. .OR. .AND. .LT. .LE. = .EQ. == .NE. /= .GT. > ...

&ME,nmr_mygroup,nproc,'out_'//zahlenchars(nmr_mygroup)

---------------------------------------^

compilation aborted for ./para_p.f (code 1)

Here is the source code of para_p.f:

[...]

C This is line 130.

C ==--------------------------------------------------------------==

do i=1,nmr_threads

if (ME.eq.parents(i)) then

if (nmr_superparent) then

write (6,'("PROCESSOR ",I3," (GROUP ",I1,'//

&'" with ",I3," procs): I am the superparent. ")')

&ME,nmr_mygroup,nproc

else

write (6,'("PROCESSOR ",I3," (GROUP ",I1,'//

&'" with ",I3," procs): Output --> ",A5,".")')

&ME,nmr_mygroup,nproc,'out_'//zahlenchars(nmr_mygroup)

close(6)

open(unit=6,file='out_'//zahlenchars(nmr_mygroup))

endif

endif

call my_sync(nmr_supergroup)

enddo

C ==--------------------------------------------------------------==

C This is line 149.

[...]

cheers,

Rex

p.s. I'm sorry the html format above is too awkward, please see the error message and source code in the log file below.

Message Edited by Pi2002 on 06-03-200603:22 AM

TimP · ‎06-03-2006

It's difficult to do anything with this, if you don't supply a working example. Did you try ifort 9.1.033 yet?

pi2002 · ‎06-03-2006

tim18 wrote:

It's difficult to do anything with this, if you don't supply a working example. Did you try ifort 9.1.033 yet?

Well, I just tried the evaluation version of ifc, and 9.1.032 is the latest version I could find in your website yet.

P.S. Please find the code I tested below.

pi2002 · ‎06-08-2006

In addition, I tested ifort 9.1.032 on a P4-32bit platform, the same error also appears. What's your suggestion?

TimP · ‎06-08-2006

Still, we can't guess what would be in your include files. I think there was a bug confirmed in ifort on use of // string concatenation in context such as this, where it generates a string temporary.
Until that is fixed, it should be possible to work around it by using // without going so far beyond the f77 standard limitations. You would have to declare a string to receive the full concatenated string, and use that string in the write. I can't check whether 9,1,033 fixed the bug for your case unless you make the case complete.

Steven_L_Intel1 · ‎06-08-2006

Please send a complete example along with a detailed description of the problem to Intel Premier Support - See here for more information.

pi2002 · ‎06-08-2006

I'm sorry, tim. Please find the tested code file and all include files packed below.

ps. I just could not logon on the Intel Premier Support website:
https://premier.intel.com/
The brower redirected me to:
http://or1grebe.cps.intel.com/scripts-util/404.asp?Redirect=Y
Maybe that's the problem of my proxy, I'll try again later.

Message Edited by Pi2002 on 06-08-200607:58 AM

pi2002 · ‎06-08-2006

Just test it with:

ifort -c nmr_para_p.f

TimP · ‎06-08-2006

The current em64t linux (9.1.033) and 32-bit Windows ifort compilers appear to have no trouble compiling this e.g.
ifort -xW -c nmr_para_p.f

A few odd details (besides 5 unused declared variables):
Your Cray pointer usage seems to be agreeable to several compilers.
ifort doesn't mention the possibly un-initialized myfirst.
ifort em64t linux appears to be more confident about data alignments than ifort 32-bit Windows, so uses several instances of parallel SSE2.

The problem you mention with access to Premier is usually occasioned by cookie usage. You may have to delete all cookies from intel.com, possibly prior to each login. I have been prevented at times from logging into premier while logged into the forum, and then must close my browser (which then automatically deletes cookies) and start it again. The premier site administrators haven't documented exact expected cookie treatment settings, to my knowledge. About a month ago, a number of internal URL paths were changed, such that they can be broken by using paths stored in cookies. The site became easier for me to use on linux, with more than one browser, than on Windows.

pi2002 · ‎06-10-2006

Thanks,Tim. The ifort 9.1.033 (em64t) version really works.

Well, it runs a bit slower than the 9.0.031 version in my case, I have used PGO and -O3 level optimizations. Beside of the new CPUs support, what's the main optimization update between ifort 9.1 version and the 9.0 version?

pi2002 · ‎06-10-2006

I'm running my code on a Linux 64-bit cluster. Is this reduction of performance due to the modification of ifort with Cluster OpenMP feature mentioned here?
http://www.intel.com/cd/software/products/asmo-na/eng/compilers/flin/282048.htm#new

Well, I'm really not sure about this and could not do such a test. That's why I posted a topic previously on this board here:
http://softwareforums.intel.com/ids/board/message?board.id=11&message.id=3895

Thanks again!

Rex

Steven_L_Intel1 · ‎06-12-2006

If you are not using Cluster OpenMP, it has no effect on your program. The Release Notes discuss some of the changes between 9.0 and 9.1.