[petsc-dev] [petsc-maint #129715] __float128 complex

[Jose E. Roman]

El 23/08/2012, a las 18:11, Barry Smith escribió:

> 
>   Jose,
> 
>     This is great, thanks so much!
> 
>     Does Satish need to roll a new http://ftp.mcs.anl.gov/pub/petsc/externalpackages/f2cblaslapack-3.1.1.q.tar.gz  tar ball?   I'm guessing yes?
> 

No. I used the one that is currently brought by --download-f2cblaslapack
Anyway, it would be a good opportunity to upgrade to Lapack 3.4 (the toclapack.sh script should work for this version as well).

Jose

> 
> 
>     Barry
> 
> 
> 
> 
> On Aug 23, 2012, at 11:02 AM, "Jose E. Roman" <jroman at dsic.upv.es> wrote:
> 
>> 
>> El 22/08/2012, a las 18:52, Jose E. Roman escribió:
>> 
>>> f2cblaslapack is more or less prepared for this case already (it is not much tested though, we will make some tests for the w* subroutines tomorrow).
>>> 
>>> What is not ready is PETSc itself. Some definitions are missing. For instance, PetscScalar is undefined, so one should add the following to petscmath.h:106:
>>> 
>>> #elif defined(PETSC_USE_REAL___FLOAT128)
>>> typedef __complex128 PetscScalar;
>>> 
>>> #define PetscRealPart(a)      crealq(a)
>>> #define PetscImaginaryPart(a) cimagq(a)
>>> #define PetscAbsScalar(a)     cabsq(a)
>>> #define PetscConj(a)          conjq(a)
>>> #define PetscSqrtScalar(a)    csqrtq(a)
>>> #define PetscPowScalar(a,b)   cpowq(a,b)
>>> #define PetscExpScalar(a)     cexpq(a)
>>> #define PetscLogScalar(a)     clogq(a)
>>> #define PetscSinScalar(a)     csinq(a)
>>> #define PetscCosScalar(a)     ccosq(a)
>>> 
>>> Similarly, MPIU_SCALAR should be defined. Maybe other as well.
>>> 
>>> Jose
>> 
>> I have pushed a change to petsc-dev that allows to configure --with-precision=__float128 --with-scalar-type=complex
>> 
>> I have tested this with src/ksp/ksp/examples/tutorials/ex12.c with 1 and 2 processes. It seems to work.
>> 
>> Please review the change, since I had to modify pinit.c
>> 
>> A couple of comments:
>> 
>> 1) The f2cblaslapack interface to quad precision subroutines (q* and w*) is defined only in the case of "uscore" mangling. Maybe a test in configure should assure this.
>> 
>> 2) I could not make it work --with-clanguage=c++ - I get compilation errors like this. Any idea?
>> 
>> /usr/include/c++/4.6/complex: In function ‘_Tp std::__complex_abs(const std::complex<_Tp>&) [with _Tp = __float128]’:
>> /usr/include/c++/4.6/complex:592:68:   instantiated from ‘_Tp std::abs(const std::complex<_Tp>&) [with _Tp = __float128]’
>> /home/jroman/soft/petsc-dev/src/sys/utils/sortd.c:116:26:   instantiated from here
>> /usr/include/c++/4.6/complex:571:50: error: call of overloaded ‘abs(__float128&)’ is ambiguous
>> /usr/include/c++/4.6/complex:571:50: note: candidates are:
>> /usr/include/c++/4.6/cmath:82:3: note: double std::abs(double)
>> /usr/include/c++/4.6/cmath:86:3: note: float std::abs(float)
>> /usr/include/c++/4.6/cmath:90:3: note: long double std::abs(long double)
>> /usr/include/c++/4.6/complex:571:50: error: call of overloaded ‘abs(__float128&)’ is ambiguous
>> /usr/include/c++/4.6/complex:571:50: note: candidates are:
>> /usr/include/c++/4.6/cmath:82:3: note: double std::abs(double)
>> /usr/include/c++/4.6/cmath:86:3: note: float std::abs(float)
>> /usr/include/c++/4.6/cmath:90:3: note: long double std::abs(long double)
>> /usr/include/c++/4.6/complex:576:46: error: call of overloaded ‘sqrt(__float128)’ is ambiguous
>> /usr/include/c++/4.6/complex:576:46: note: candidates are:
>> /usr/include/x86_64-linux-gnu/bits/mathcalls.h:158:1: note: double sqrt(double)
>> /usr/include/c++/4.6/cmath:416:3: note: float std::sqrt(float)
>> /usr/include/c++/4.6/cmath:420:3: note: long double std::sqrt(long double)
>> 
>> 
>> 
>