[petsc-users] Problem in compiling PETSc+ML

[Ali Reza Khaz'ali]

Thank you very much for the assistance. I have compiled ML (and HYPRE, 
too) separately; and then linked them to PETSc with 
--with-ml-include,--with-ml-lib and 
--with-hypre-include,--with-hypre-lib. The modules are accessible from 
PETSc interface now, and working as expected.

Ali


On 8/17/2018 2:32 AM, Satish Balay wrote:
> Here is one way to get --download-ml working. [its a bit hakey]
>
> 1. update ml.py to run configure with additional options: "--disable-libcheck FLIBS=Ws2_32.lib"
>
> diff --git a/config/BuildSystem/config/packages/ml.py b/config/BuildSystem/config/packages/ml.py
> index 2987cf2122..e921db0ef5 100644
> --- a/config/BuildSystem/config/packages/ml.py
> +++ b/config/BuildSystem/config/packages/ml.py
> @@ -35,7 +35,7 @@ class Configure(config.package.GNUPackage):
>       args.append('--disable-ml-aztecoo')
>       args.append('--disable-ml-examples')
>       args.append('--disable-tests')
> -    args.append('--enable-libcheck')
> +    args.append('--disable-libcheck')
>   
>       self.framework.pushLanguage('C')
>       args.append('--with-cflags="'+self.removeWarningFlags(self.framework.getCompilerFlags())+' -DMPICH_SKIP_MPICXX -DOMPI_SKIP_MPICXX '+ self.headers.toStringNoDupes(self.mpi.include)+'"')
> @@ -45,6 +45,7 @@ class Configure(config.package.GNUPackage):
>       if hasattr(self.compilers, 'FC'):
>         self.framework.pushLanguage('FC')
>         args.append('--with-fflags="'+self.framework.getCompilerFlags()+' '+ self.headers.toStringNoDupes(self.mpi.include)+'"')
> +      args.append('FLIBS=Ws2_32.lib')
>         self.framework.popLanguage()
>       else:
>         args.append('F77=""')
>
> 2. run configure with option CPPFLAGS=-DICL [this works arround some include file issues inside ML]
>
> Here ML gets successfully built  - but its built as libml.a - not libml.lib
>
> 3. copy PETSC_ARCH/lib/libml.a to PETSC_ARCH/lib/libml.lib
>
> 4. rerun previous configure command [don't forget CPPFLAGS=-DICL option]
>
> Satish
>
> On Thu, 16 Aug 2018, Satish Balay wrote:
>
>> If you can built it from Trilinos - you should be able to use it from PETSc.
>> [by specifing ml to petsc with --with-ml-include,--with-ml-lib options]
>>
>> This is unverified though..
>>
>> Satish
>>
>>
>> On Thu, 16 Aug 2018, Ali Reza Khaz'ali wrote:
>>
>>> Thanks a lot. In fact I had built ML from Trilinos before, and I wondered if I
>>> could use it in PETSc.
>>>
>>> Best wishes,
>>>
>>> Ali
>>>
>>>
>>> On 8/16/2018 10:12 PM, Satish Balay wrote:
>>>> Sorry - most external packages won't work with --download-pacakge
>>>> option on windows [with MS/Intel compilers].
>>>>
>>>> If you need ML on windows - you might have to install it separately
>>>> [by checking its windows installation instructions - if they exist]
>>>>
>>>> And then specify it to petsc configure with --with-ml-include
>>>> --with-ml-lib options.
>>>>
>>>> Satish
>>>>
>>>> On Thu, 16 Aug 2018, Ali Reza Khaz'ali wrote:
>>>>
>>>>> Hi,
>>>>>
>>>>> I have been using PETSc on a MS windows 7 x64 system with Intel MKL, MPI
>>>>> and
>>>>> compilers without problem. For an specific problem, I had to use ML.
>>>>> However,
>>>>> I could not compile ML with the before-mentioned toolsets.  The the log
>>>>> files
>>>>> (PETSc and ML) are attached. Any help is much appreciated.
>>>>>
>>>>>
>>>>> Thanks,
>>>>>
>>>>> Ali
>>>>>
>>>>>
>>>

-- 
Ali Reza Khaz’ali
Assistant Professor of Petroleum Engineering,
Department of Chemical Engineering
Isfahan University of Technology
Isfahan, Iran