Installing ctffind-4.1.14 on IBM Power9

Submitted by d.molinari on Thu, 2021-04-08 12:50

Forums

ctffind & ctftilt

Dear support,

I am trying to install ctffind software on a IBM cluster with Power9 architecture. During compilation I get the following error

g++ -DPACKAGE_NAME=\"ctffind\" -DPACKAGE_TARNAME=\"ctffind\" -DPACKAGE_VERSION=\"4.1.14\" -DPACKAGE_STRING=\"ctffind\ 4.1.14\" -DPACKAGE_BUGREPORT=\"\" -DPACKAGE_URL=\"\" -DPACKAGE=\"ctffind\" -DVERSION=\"4.1.14\" -DSTDC_HEADERS=1 -DHAVE_SYS_TYPES_H=1 -DHAVE_SYS_STAT_H=1 -DHAVE_STDLIB_H=1 -DHAVE_STRING_H=1 -DHAVE_MEMORY_H=1 -DHAVE_STRINGS_H=1 -DHAVE_INTTYPES_H=1 -DHAVE_STDINT_H=1 -DHAVE_UNISTD_H=1 -DHAVE_DLFCN_H=1 -DLT_OBJDIR=\".libs/\" -Dwx_is_available=1 -I. -I..  -I/m100_scratch/userinternal/dmolina1/CCTF/wxwidget/wxWidgets/install/lib/wx/include/gtk2-unicode-3.1 -I/m100_scratch/userinternal/dmolina1/CCTF/wxwidget/wxWidgets/install/include/wx-3.1 -D_FILE_OFFSET_BITS=64 -DwxDEBUG_LEVEL=0 -DWXUSINGDLL -D__WXGTK__ -DwxUSE_GUI=0 -DNDEBUG -funroll-loops -O3 -Wall -pipe -fexpensive-optimizations -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -I/m100_scratch/userinternal/dmolina1/CCTF/wxwidget/wxWidgets/install/lib/wx/include/gtk2-unicode-3.1 -I/m100_scratch/userinternal/dmolina1/CCTF/wxwidget/wxWidgets/install/include/wx-3.1 -D_FILE_OFFSET_BITS=64 -DwxDEBUG_LEVEL=0 -DWXUSINGDLL -D__WXGTK__ -DwxUSE_GUI=0 -DNDEBUG -funroll-loops -O3 -Wall -pipe -fexpensive-optimizations -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -fopenmp -MT src/core/ctffind-matrix.o -MD -MP -MF src/core/.deps/ctffind-matrix.Tpo -c -o src/core/ctffind-matrix.o `test -f 'src/core/matrix.cpp' || echo '../'`src/core/matrix.cpp

 

../src/core/matrix.cpp:16:33: error: impossible constraint in ‘asm’
     FLOATSINCOS(x, sin_x, cos_x);  \
                                 ^
../src/core/matrix.cpp:208:5: note: in expansion of macro ‘MAKE_ROTATION_f’
     MAKE_ROTATION_f(x, y, z);
     ^~~~~~~~~~~~~~~
../src/core/matrix.cpp:17:33: error: impossible constraint in ‘asm’
     FLOATSINCOS(y, sin_y, cos_y);  \
                                 ^
../src/core/matrix.cpp:208:5: note: in expansion of macro ‘MAKE_ROTATION_f’
     MAKE_ROTATION_f(x, y, z);
     ^~~~~~~~~~~~~~~
../src/core/matrix.cpp:18:33: error: impossible constraint in ‘asm’
     FLOATSINCOS(z, sin_z, cos_z);  \
                                 ^
../src/core/matrix.cpp:208:5: note: in expansion of macro ‘MAKE_ROTATION_f’
     MAKE_ROTATION_f(x, y, z);
     ^~~~~~~~~~~~~~~

make: *** [Makefile:814: src/core/ctffind-matrix.o] Error 1

I configured like this

../configure --enable-static --enable-openmp FC=gfortran F77=gfortran

I am afraid that the problem is related to the function __asm__ that is present only in matrix.cpp  file. Do you have a suggestion to solve this issue? Does the software support IBM Power9 systems?

Many thanks for your availability.

Best regards,

Diego

 

 

Alexis Tue, 2021-04-13 17:48

Hello Diego,

Apologies for this problem. 

Could you please change line 7 of src/core/matrix.cpp from:

#define _AL_SINCOS(x, s, c)  __asm__ ("fsincos" : "=t" (c), "=u" (s) : "0" (x))

to 

#if defined(__i386__) || defined(__x86_64__)
#define _AL_SINCOS(x, s, c) __asm__ ("fsincos" : "=t" (c), "=u" (s) : "0" (x))
#else
#define _AL_SINCOS(x, s, c) s = sinf(x); c = cosf(x);
#endif

and try again?

Thanks,
Alexis