[Wien] unaligned access

Shunli Shang S.Shang at tnw.tudelft.nl
Fri Dec 12 10:01:20 CET 2003


Dear Torsten,

Thanks for your suggestion. I solved my problem by changing the compiler
option -O3 to -O2, i.e. I turned off the aggressive optimization option.
In efc compiler, I found only -Zp[n] (Specify alignment constraint for
structures (n=1,2,4,8,16)) has the relationship with align, I tried n=4, 8
(default) and 16, but no influences on my results, the unaligned faults
still appear when I use the -O3 option.

Best regards,
Shunli

----- Original Message ----- 
From: "Torsten Andersen" <thor at physik.uni-kl.de>
To: <wien at zeus.theochem.tuwien.ac.at>
Sent: Thursday, December 11, 2003 6:04 PM
Subject: Re: [Wien] unaligned access


Dear Dr. Shang,

I would guess that the SGI compiler has an option to align data on
certain boundaries, e.g., 8-byte words, or, since you are using a 64-bit
processor, an option that compiles to 64-bit binaries.

However, isn't it just warnings? Then turn off the "-g" flag (debug?)...

Best regards,
Torsten Andersen.

Shunli Shang wrote:
> Dear Prof. P.Blaha and Wien users,
>
> I installed the wien2k code (successfully?) in a SGI supercomputer. There
> are no error messages during the compiling process, and results for TiC
case
> is correct. But, the unaligned access messages appeared in lapw1 when I
ran
> the TiC case. Could you please tell me how to fix the unaligned faults?
>
> Details of SGI computer:
> Computer:             SGI Altix 3700 system (Intel Itanium-2 CPU)
> Operating system:  RedHat Linux
> Compiler:              efc: the Intel Fortran compiler
> Libraries:               MKL: Numerical software from INTEL
>
> Compile settings for wien2k:
> O   Compiler options:    -FR -mp -w -O3 -g
> L   Linker Flags:            -openmp -Vaxlib
> P   Preprocessor flags     '-DParallel'
> R   R_LIB (LAPACK+BLAS):     -lmkl_lapack -lmkl_itp -lmkl_vml_itp
>
> When I run the TiC case, the following messages appear:
> lapw1(26213): unaligned access to 0x2000000000e8b048,
ip=0x400000000000d780
> lapw1(26213): unaligned access to 0x2000000000ebf048,
ip=0x400000000000d781
> lapw1(26213): unaligned access to 0x2000000000e8b018,
ip=0x400000000000d7a0
> lapw1(26213): unaligned access to 0x2000000000ebf018,
ip=0x400000000000d7a1
>
> Using the gdb command, I found the unaligned faults are located in lines
261
> and 262 of atpar.f  (SRC_lapw1)
> 0x400000000000d780 is in atpar (atpar.f:261).
> 0x400000000000d7a0 is in atpar (atpar.f:261).
> 0x400000000000d781 is in atpar (atpar.f:262).
> 0x400000000000d7a1 is in atpar (atpar.f:262).
>
> ------in atpar.f-------
> 256     !
> 257                 CALL RINT13(CFEIN1,CFEIN2,A,B,AE,BE,CROSS,JATOM)
> 258                 TRY = -CROSS
> 259                 IF (TRY .LT. -0.05D+0) WRITE (6,6070) L, TRY, OVLP
> 260                 DO 130 M = 1, JRI
> 261                    AE(M) = AE(M) + TRY*A(M)
> 262                    BE(M) = BE(M) + TRY*B(M)
> 263      130        CONTINUE
> 264                 PE(J,JATOM) = UVE + TRY*P(J,JATOM)
> 265                 DPE(J,JATOM) = DUVE + TRY*DP(J,JATOM)
> --------------------
>
> Thanks for your help!
>
> Best regards,
> Shunli
>
> ********************************************************
> Dr. Shunli Shang
> Laboratory of Materials Science
> Delft University of Technology
> Rotterdamseweg 137, 2628AL, Delft
> The Netherlands
> Tel:  +31-15-278 8397
> Fax: +31-15-278 6730
>
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>

-- 
Dr. Torsten Andersen                     TA-web: http://deep.at/myspace/
AG Hübner, Department of Physics, Kaiserslautern University, and
Condensed Matter Theory Group, Department of Physics, Uppsala University
Web: http://www.fysik4.fysik.uu.se/         http://www.physik.uni-kl.de/


_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien







More information about the Wien mailing list