[mpich-discuss] Build in Solaris 10 x86 HELP...

renato at dokfile.com renato at dokfile.com
Tue Nov 8 16:21:17 CST 2011


Hi Dave, NOW WORK!!! THANKS, THANKS, THANKS!!! you're the guy! Thanks again!

-----Original Message-----
From: Dave Goodell [mailto:goodell at mcs.anl.gov]
Sent: Tuesday, November 8, 2011 05:15 PM
To: mpich-discuss at mcs.anl.gov
Subject: Re: [mpich-discuss] Build in Solaris 10 x86 HELP...

There should be no assembly language programming, just a small tweak to a C preprocessor directive. You just need to take the line that currently says:----8<----#if defined(HAVE_GCC_AND_PENTIUM_ASM)----8<----and change it to say:----8<----#if 0----8<----instead. This should be line 12 of "src/mpid/ch3/channels/nemesis/nemesis/include/mpid_nem_memdefs.h".Then re-running make should work.-DaveOn Nov 8, 2011, at 4:12 PM CST, renato at dokfile.com wrote:> Hi Dave,> > The second option is not for me, I cant upgrade it.> > > A heavier weight option would be to upgrade your GCC and assembler packages so that the fancier assembly code is understood correctly. I personally wouldn't bother with this, but it should work.> > I am not an advanced Guy on Assembly. I will try.> > > In the short run, you should be able to get unstuck by editing 12 of that file to read "#if 0" instead. If the relevant section of the configure.in file was written correctly then there would be other easy ways to get around this, but unfortunately it appears to be buggy.> > Can you giveme more detail, that show how I do this... > > Thanks> -----Original Message-----> From: Dave Goodell [mailto:goodell at mcs.anl.gov]> Sent: Tuesday, November 8, 2011 02:08 PM> To: mpich-discuss at mcs.anl.gov> Subject: Re: [mpich-discuss] Build in Solaris 10 x86 HELP...> > for the archives: http://trac.mcs.anl.gov/projects/mpich2/ticket/1545 On Nov 8, 2011, at 1:03 PM CST, Dave Goodell wrote: > It looks like the assembler on your machine doesn't like some of the enhanced memcpy code that we've written into nemesis. This appears to be a Solaris on x86 machine, and the configure script decides that x86 assembly is supported by the assembler by using a small test program. Unfortunately, the assembler does not support a more complicated hunk of inline assembly listed here: > > http://trac.mcs.anl.gov/projects/mpich2/browser/mpich2/tags/release/mpich2-1.4.1p1/src/mpid/ch3/channels/nemesis/nemesis/include/mpid_nem_memdefs.h > > In the short run, you should be able to get unstuck by editing 12 of that file to read "#if 0" instead. If the relevant section of the configure.in file was written correctly then there would be other easy ways to get around this, but unfortunately it appears to be buggy. > > A heavier weight option would be to upgrade your GCC and assembler packages so that the fancier assembly code is understood correctly. I personally wouldn't bother with this, but it should work. > > In the long run we should probably make the configure test more robust, and also add a "--disable-asm-memcpy" option or similar so that no code editing is required. I'll add a ticket for this in trac. > > -Dave > > On Nov 8, 2011, at 11:57 AM CST, renato at dokfile.com wrote: > >> Hi All, >> >> I am tring build MPICH2 in Solaris 10, and I dont have success. >> >> Fist I have problems in configuration, now I put it to work, but now I have problems in build. >> >> Whell, this is what I did -> >> >> ./configure CC=gcc CXX=g++ --disable-f77 --disable-fc --prefix=/MPICH/mpich2-install/ 2>&1 | tee c.txt >> >> The result is OK, see -> >> >> .. >> Configuration completed. >> -bash-3.00# >> >> Then, I did make -> >> >> make 2>&1 | tee m.txt >> >> Assembler: dataloop.c >> "/var/tmp//ccuee48e.s", line 493 : Syntax error >> Near line: "mov $PREFETCHBLOCK/16, %eax" >> "/var/tmp//ccuee48e.s", line 501 : Syntax error >> Near line: "mov $PREFETCHBLOCK/8, %eax" >> make[5]: ** [dataloop.o] Erro 1 >> make[5]: Saindo do diretório `/MPICH/Source/mpich2-1.4.1p1/src/mpid/common/datatype/dataloop' >> make[4]: ** [all-redirect] Erro 2 >> make[4]: Saindo do diretório `/MPICH/Source/mpich2-1.4.1p1/src/mpid/common/datatype' >> make[3]: ** [all-redirect] Erro 2 >> make[3]: Saindo do diretório `/MPICH/Source/mpich2-1.4.1p1/src/mpid/common' >> make[2]: ** [all-redirect] Erro 2 >> make[2]: Saindo do diretório `/MPICH/Source/mpich2-1.4.1p1/src/mpid' >> make[1]: ** [all-redirect] Erro 2 >> make[1]: Saindo do diretório `/MPICH/Source/mpich2-1.4.1p1/src' >> make: ** [all-redirect] Erro 2 >> -bash-3.00# >> >> And error. >> >> What is wrong..? >> >> Thank you... >> >> >> _______________________________________________ >> mpich-discuss mailing list mpich-discuss at mcs.anl.gov >> To manage subscription options or unsubscribe: >> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss > > _______________________________________________ > mpich-discuss mailing list mpich-discuss at mcs.anl.gov > To manage subscription options or unsubscribe: > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss _______________________________________________ mpich-discuss mailing list mpich-discuss at mcs.anl.gov To manage subscription options or unsubscribe: https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss> _______________________________________________> mpich-discuss mailing list mpich-discuss at mcs.anl.gov> To manage subscription options or unsubscribe:> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss_______________________________________________mpich-discuss mailing list mpich-discuss at mcs.anl.govTo manage subscription options or unsubscribe:https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20111108/d0fd42eb/attachment.htm>


More information about the mpich-discuss mailing list