You are here

Error during installation of rosetta2.3 in Blue gene

5 posts / 0 new
Last post
Error during installation of rosetta2.3 in Blue gene


I am trying to install rosetta2.3 in IBM Blue gene
Compiler: xlc++
Mpi : mpixlC

The error got during compilation:
mpixlC -c -I. -Isrc -Iexternal/include -Isrc/platform/linux/ -O3 -qhot=nosimd -DNDEBUG -DUSEMPI -qarch=440 -qtune=440 -qmaxmem=-1 -qlanglvl=c99__func__ -MT.xlc++_ppc.map_sequence.d -MF.xlc++_ppc.map_sequence.d -o .xlc++_ppc.map_sequence.o
"src/ObjexxFCL/FArray1DB.hh", line 477.29: 1540-0274 (S) The name lookup for "equal_dimensions" did not find a declaration.
"src/ObjexxFCL/FArray4DB.hh", line 565.1: 1540-1298 (I) "template bool equal_dimensions(const FArray4DB &, const FArray4DB &)" needs to be declared in the containing scope to be found by name lookup.
"src/ObjexxFCL/FArray3DB.hh", line 530.1: 1540-1298 (I) "template bool equal_dimensions(const FArray3DB &, const FArray3DB &)" needs to be declared in the containing scope to be found by name lookup.
"src/ObjexxFCL/FArray2DB.hh", line 683.1: 1540-1298 (I) "template bool equal_dimensions(const FArray2DB &, const FArray2DB &)" needs to be declared in the containing scope to be found by name lookup.
"src/ObjexxFCL/FArray1DB.hh", line 585.1: 1540-1298 (I) "template bool equal_dimensions(const FArray1DB &, const FArray1DB &)" needs to be declared in the containing scope to be found by name lookup.
make: *** [.xlc++_ppc.map_sequence.o] Error 1

I tried using extern "C" in src/ObjexxFCL/FArray1DB.hh, but, could not get rid of the problem. I shall be thankful if helped in this matter.

Post Situation: 
Mon, 2011-01-31 09:49

I assume you do mean Rosetta2, not Rosetta3. Let me know which you meant and I'll delete the other topic.

Nobody's ever run Rosetta2 on a bluegene to my knowledge. The Bluegene compiler is pretty finicky, you'll have to figure out what it wants changed...

We have had issues in the past with getting Rosetta3 on bluegenes; some were fixed and some were not. (In fact, I have access to a bluegene I can't use, because the compiler won't compile Rosetta3 and we can't get the sysops to upgrade the compiler...)

Mon, 2011-01-31 11:15


I want to install Rosetta 2.3. The Readme files for installation clearly mentions that the software can be installed on BlueGene with the compilers compatible. I am unable to debug a small error in the code which is commonly encountered if compiled with xlc compiler. Since, installation on BlueGene recomends mpixlC for compilation parallely, I wanted to sort out the error.


Mon, 2011-01-31 22:22

Dear All

I had tried compiling rosetta2.3 on IBM BlueGene using mpixlC compiler. I had compiled the program files at frontend mode of BlueGene to generate object files. I got the following warning:
mpixlC -c -I. -Isrc -Iexternal/include -Isrc/platform/linux/ -O3 -qhot=nosimd -DNDEBUG -DUSEMPI -qarch=440 -qtune=440 -qmaxmem=-1 -qlanglvl=c99__func__ -MT.xlc++_ppc.void.d -MF.xlc++_ppc.void.d -o .xlc++_ppc.void.o
1500-036: (I) The NOSTRICT option (default at OPT(3)) has the potential to alter the semantics of a program. Please refer to documentation on the STRICT/NOSTRICT option for more information.

This warning was common to mostly all the object files created. I then tried to combine al the object files to get the final executable file: rosetta.xlc++_ppc at the backend mode using make xlc++_ppc, but, without the -q64 option in makefile i.e I compiled it using mpirun32.

Though I was able to strip rosetta.xlc++_ppc from the object files generated, I was not able to run the program successfully. The program haults in the middle after generating few core files (error).

core file:
ended with software signal.......0x0000000b (SIGSEGV - segmentation violation)
generated by interrupt...........0x0000000d (data TLB error interrupt)
while executing instruction at...0x00c0b820
dereferencing memory at..........0x48284444

General Purpose Registers:
r00=0x00c0b068 r01=0x3ffb4000 r02=0xdeadbeef r03=0x00000004
r04=0x48284440 r05=0x01d7ee54 r06=0x3ffb4120 r07=0x02720000
r08=0x02720000 r09=0x01e00000 r10=0x01d80000 r11=0x0b4e7ffc
r12=0x00000004 r13=0xdeadbeef r14=0x00000004 r15=0x00000007
r16=0x0270ac20 r17=0x00000005 r18=0x00000004 r19=0x00000240
r20=0x00000000 r21=0x02709e60 r22=0x01d7f670 r23=0x3ffb4580
r24=0x3ffb4430 r25=0x0000000a r26=0x3ffb4490 r27=0x3ffb44c0
r28=0x00000000 r29=0x024904d0 r30=0x01d7f670 r31=0x3ffb40f0

The program haults at :

Filling missing atoms
optimizing missing atoms
[T/F OPT]Default FALSE value for [-disallow_hbond_o4star]
[T/F OPT]Default FALSE value for [-disallow_hbond_ho2star]

I am looking forward for any suggestion regarding this.

Thanking you.......

Sun, 2011-02-06 22:17

I've asked around and nobody that I can find has run 2.3 on a BlueGene and/or encountered errors like this. Maybe another user can help...

Mon, 2011-02-07 13:10