[mpich-discuss] Assembler messages at build ?

[Lou Picciano]

Huiwei, Thanks for your ideas. Responses follow:

----- Original Message -----
From: "Huiwei Lu" <huiweilu at mcs.anl.gov>
To: discuss at mpich.org
Sent: Saturday, November 23, 2013 10:04:23 PM
Subject: Re: [mpich-discuss] Assembler messages at build ?

Because I am not familiar with solaris, I will have some wild guesses here.

1. Does it has to be ‘-m32’? I searched a little bit and find SunOS 5.11 actually support 64 bit if the processor is 64 bit. Have you tried to configure and compile without ‘-m32’? What’s the output?

Without CFLAGS or FFLAGS set, configure sets up a 32-bit build, with same Assembler messages as below.

As an aside: When attempting an explicit configure for 64-bit:

$ ./configure --prefix=/usr/local/mpi32 \
FC=gfortran \
F77=gfortran \
CFLAGS="-m64" FFLAGS="-m64" \
--with-pm=hydra \
| tee c.txt

configure: error: The selected Fortran 90 compiler gfortran does not work with the selected Fortran 77 compiler gfortran.  Use the environment variables FC and F77 respectively to select compatible Fortran compilers.  The check here tests to see if a main program compiled with the Fortran 90 compiler can link with a subroutine compiled with the Fortran 77 compiler.
(same result without FC and F77 being set)

2. Can you give us the output of ‘isainfo -v’?

64-bit amd64 applications
        vmx avx xsave pclmulqdq aes sse4.2 sse4.1 ssse3 popcnt tscp cx16 
        sse3 sse2 sse fxsr mmx cmov amd_sysc cx8 tsc fpu 
32-bit i386 applications
        vmx avx xsave pclmulqdq aes sse4.2 sse4.1 ssse3 popcnt tscp ahf 
        cx16 sse3 sse2 sse fxsr mmx cmov sep cx8 tsc fpu

3. Have you tried the newest version of gcc? Because SunOS 5.11 is quite old now, maybe it is a bug in an old version of gcc which has been fixed.

We're using 4.7.3, with some significant tweaks for the Illumos kernel. (We may be slipping away, at this point, from an 'exact match' test on SunOS 5.11!)

Lou Picciano

— Huiwei Lu
http://www.mcs.anl.gov/~huiweilu/

On Nov 23, 2013, at 9:02 PM, Junchao Zhang <jczhang at mcs.anl.gov> wrote:

> Hi, Lou,
>   It seems you met a compiler bug which shows up at the assemble stage.
>   Let's look at what it is. Try make V=1 src/mpi/coll/helper_fns.lo
>   Copy the command which gcc used to compile helper_fns.c , Then add extra flags " -v -save-temps" to the command. You should find the error happens at which line of the helper_fns.s file. Let's see what you will find.
>   Thanks.
> 
> --Junchao Zhang
> 
> 
> On Sat, Nov 23, 2013 at 11:13 AM, Lou Picciano <loupicciano at comcast.net> wrote:
> mpich Wizards,
> 
> Trying our first build of mpich-3.0.4, and seeing Assembler messages at build time:
> (there are clearly a lot of config options to become familiar with; but, as a first pass):
> 
> Configure is:
> ./configure --prefix=/usr/local/mpi32 \
> CFLAGFS="-m32" FFLAGS="-m32" \
> | tee c.txt
> 
> $ make
> ...
>   CC       src/mpi/attr/attrutil.lo
>   CC       src/mpi/attr/dup_fn.lo
>   CC       src/mpi/coll/allred_group.lo
>   CC       src/mpi/coll/barrier_group.lo
>   CC       src/mpi/coll/helper_fns.lo
> /var/tmp//ccf4aaqC.s: Assembler messages:
> /var/tmp//ccf4aaqC.s:851: Error: number of operands mismatch for `mov'
> /var/tmp//ccf4aaqC.s:859: Error: number of operands mismatch for `mov'
> make[2]: *** [src/mpi/coll/helper_fns.lo] Error 1
> make[2]: Leaving directory `<build_path>/mpich-3.0.4'
> make[1]: *** [all-recursive] Error 1
> 
> gcc 4.7.3
> Platform is i86 OpenIndiana (Illumos, Solaris-ish)
> 
> Can anyone point me in the right direction?
> 
> Lou
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
> 
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss

_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss