Re: OpenMP + ACLiC

 Hi Akira,

What are all the steps you take to compile the standalone case?

Cheers,
Philippe.

On 9/15/10 1:25 PM, OKUMURA, Akira wrote:
> Hello Philippe,
>
> Thank you again. But I get the same error again. Does this work on your machine? I tried gcc 4.2.1 on Snow Leopard and gcc 4.4.0 on Scientific Linux 5.
>
> $ root
> root [0] TString cmd(gSystem->GetMakeSharedLib())
> root [1] cmd.ReplaceAll("g++", "g++ -fopenmp")
> (class TString)"cd $BuildDir ; g++ -fopenmp -c $Opt -m64 -pipe -Wshadow -W -Wall -Woverloaded-virtual -fsigned-char -fno-common -D_REENTRANT -pthread $IncludePath $SourceFiles ; MACOSX_DEPLOYMENT_TARGET=10.6 g++ -fopenmp $ObjectFiles -dynamiclib -single_module -undefined dynamic_lookup -lwcs $DepLibs -o $SharedLib"
> root [2] gSystem->SetMakeSharedLib(cmd)
> root [3] .L pi.C++
> Info in <TUnixSystem::ACLiC>: creating shared library /Users/oxon/./pi_C.so
> In file included from /Users/oxon/pi_C_ACLiC_dict.h:34,
> from /Users/oxon/pi_C_ACLiC_dict.cxx:16:
> /Users/oxon/./pi.C: In function ‘void pi()’:
> /Users/oxon/./pi.C:14: error: expected ‘#pragma omp’ clause before ‘public’
> i686-apple-darwin10-g++-4.2.1: /Users/oxon/pi_C_ACLiC_dict.o: No such file or directory
> Error in <ACLiC>: Compilation failed!
>
> Regards,
>
> --
> OKUMURA, Akira oxon@{astro.isas.jaxa.jp,stanford.edu}
> Institute of Space and Astronautical Science (ISAS/JAXA)
> Now at KIPAC/SLAC/Stanford
> Varian Physics #306
> 382 Via Pueblo Mall, MC 406 Stanford, CA94305
> TEL 650-736-0971/FAX 650-724-5065
> Skype : okumura.akira
>
> On 2010/09/15, at 10:30, Philippe Canal wrote:
>
>> Hi Akira,
>>
>> You also need to add *-fopenmp to the compiler invocations.
>> For that do:
>> TString cmd( gSystem->GetMakeSharedLibs );
>> cmd.ReplaceAll("g++","g++ -fopenmp");
>> gSystem->SetMakeSharedLibs(cmd);
>>
>> Cheers,
>> Philippe.
>> *
>> On 9/15/10 11:25 AM, OKUMURA, Akira wrote:
>>> Hello Philippe,
>>>
>>> Thanks for the advise. But I still get the same error.
>>>
>>> #include <omp.h>
>>> #include <time.h>
>>> #include <iostream>
>>>
>>> void pi()
>>> {
>>> const Int_t kN = 100000000;
>>> long double top = 0.L;
>>> long double bottom = 0.L;
>>> long double kStep = 1.L/kN;
>>> long double sum = 0.L;
>>>
>>> #ifndef __CINT__
>>> #pragma omp parallel private(top, bottom)
>>> #endif
>>> {
>>> #ifndef __CINT__
>>> #pragma omp for reduction(+:sum)
>>> #endif
>>> for(Int_t i = 0; i < kN; i++) {
>>> top = 4.L/(1.L + (kStep*i)*(kStep*i));
>>> bottom = 4.L/(1.L + (kStep*(i+1))*(kStep*(i + 1)));
>>> sum += (top + bottom)*kStep/2.L;
>>> } // i
>>> }
>>> std::cout.precision(16);
>>> std::cout << "pi = " << sum << "\n";
>>> }
>>>
>>> root [4] .L pi.C++
>>> Info in <TUnixSystem::ACLiC>: creating shared library /Users/oxon/./pi_C.so
>>> In file included from /Users/oxon/pi_C_ACLiC_dict.h:34,
>>> from /Users/oxon/pi_C_ACLiC_dict.cxx:16:
>>> /Users/oxon/./pi.C: In function ‘void pi()’:
>>> /Users/oxon/./pi.C:14: error: expected ‘#pragma omp’ clause before ‘public’
>>> i686-apple-darwin10-g++-4.2.1: /Users/oxon/pi_C_ACLiC_dict.o: No such file or directory
>>> Error in <ACLiC>: Compilation failed!
>>>
>>> Regards,
>>>
>>> --
>>> OKUMURA, Akira oxon@{astro.isas.jaxa.jp,stanford.edu}
>>> Institute of Space and Astronautical Science (ISAS/JAXA)
>>> Now at KIPAC/SLAC/Stanford
>>> Varian Physics #306
>>> 382 Via Pueblo Mall, MC 406 Stanford, CA94305
>>> TEL 650-736-0971/FAX 650-724-5065
>>> Skype : okumura.akira
>>>
>>> On 2010/09/15, at 8:53, Philippe Canal wrote:
>>>
>>>> Hi,
>>>>
>>>> You need to at least hide the #pragma from CINT:
>>>>
>>>> #ifndef __CINT__
>>>> #pragma omp parallel private(top, bottom)
>>>> #endif
>>>> {
>>>> #ifndef __CINT__
>>>> #pragma omp for reduction(+:sum)
>>>> #endif
>>>> for(Int_t i = 0; i < kN; i++) {
>>>>
>>>> Cheers,
>>>> Philippe.
>>>>
>>>> On 9/15/10 10:47 AM, OKUMURA, Akira wrote:
>>>>> Hello ROOTers,
>>>>>
>>>>> Does ACLiC support OpenMP?
>>>>>
>>>>> I started learning OpenMP to improve the speed of my analysis. The following is an example to calculate pi with OpenMP.
>>>>>
>>>>> === pi.C ===
>>>>> #include <omp.h>
>>>>> #include <time.h>
>>>>> #include <iostream>
>>>>>
>>>>> void pi()
>>>>> {
>>>>> const Int_t kN = 100000000;
>>>>> long double top = 0.L;
>>>>> long double bottom = 0.L;
>>>>> long double kStep = 1.L/kN;
>>>>> long double sum = 0.L;
>>>>>
>>>>> #pragma omp parallel private(top, bottom)
>>>>> {
>>>>> #pragma omp for reduction(+:sum)
>>>>> for(Int_t i = 0; i < kN; i++) {
>>>>> top = 4.L/(1.L + (kStep*i)*(kStep*i));
>>>>> bottom = 4.L/(1.L + (kStep*(i+1))*(kStep*(i + 1)));
>>>>> sum += (top + bottom)*kStep/2.L;
>>>>> } // i
>>>>> }
>>>>> std::cout.precision(16);
>>>>> std::cout << "pi = " << sum << "\n";
>>>>> }
>>>>> ============
>>>>>
>>>>> But I got the following error.
>>>>> ============
>>>>> root [0] gSystem->SetIncludePath("-fopenmp")
>>>>> root [1] gSystem->AddLinkedLibs("-lgomp")
>>>>> root [2] .L pi.C+O
>>>>> Info in <TUnixSystem::ACLiC>: creating shared library /Users/oxon/./pi_C.so
>>>>> In file included from /Users/oxon/pi_C_ACLiC_dict.h:34,
>>>>> from /Users/oxon/pi_C_ACLiC_dict.cxx:16:
>>>>> /Users/oxon/./pi.C: In function ‘void pi()’:
>>>>> /Users/oxon/./pi.C:13: error: expected ‘#pragma omp’ clause before ‘public’
>>>>> i686-apple-darwin10-g++-4.2.1: /Users/oxon/pi_C_ACLiC_dict.o: No such file or directory
>>>>> Error in <ACLiC>: Compilation failed!
>>>>> ============
>>>>>
>>>>> It seems that some of OpenMP words are not handled properly in generation of a dictionary. I suppose ACLiC does not support OpenMP yet.
>>>>>
>>>>> If ACLiC supports OpenMP, how can I run the script in ROOT? This example runs as I expected in standalone C++ program.
>>>>>
>>>>> I am using ROOT trunk and GCC (i686-apple-darwin10-g++-4.2.1 (GCC) 4.2.1 (Apple Inc. build 5664)) on Snow Leopard.
>>>>>
>>>>> Regards,
>>>>>
>>>>> --
>>>>> OKUMURA, Akira oxon@{astro.isas.jaxa.jp,stanford.edu}
>>>>> Institute of Space and Astronautical Science (ISAS/JAXA)
>>>>> Now at KIPAC/SLAC/Stanford
>>>>> Varian Physics #306
>>>>> 382 Via Pueblo Mall, MC 406 Stanford, CA94305
>>>>> TEL 650-736-0971/FAX 650-724-5065
>>>>> Skype : okumura.akira
>>>>>
Received on Wed Sep 15 2010 - 21:50:19 CEST