Problem for make

[Liang ZHENG]

I followed your instructions to build it on my workstastion (OpenSuse
11.2, two GPUs Tesla 1060 + GTX 480),
When I tried :
make all install

it shows:

cc1plus: warnings being treated as errors
../../../../libgmac/src/api/cudadrv/compat.cpp: In function 鈉
udaError_t __getCUDAError(CUresult)?
../../../../libgmac/src/api/cudadrv/compat.cpp:96: error: enumeration
value 釩UDA_ERROR_ECC_UNCORRECTABLE?not handled in switch
make[4]: *** [compat.lo] Error 1
make[4]: Leaving directory `/home/lzheng/2010/gmac/adsm/build/src/api/
cudadrv'
make[3]: *** [all-recursive] Error 1
make[3]: Leaving directory `/home/lzheng/2010/gmac/adsm/build/src/api'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/home/lzheng/2010/gmac/adsm/build/src'
make[1]: *** [all] Error 2
make[1]: Leaving directory `/home/lzheng/2010/gmac/adsm/build/src'
make: *** [all-recursive] Error 1

I google "cc1plus: warnings being treated as errors" , it's said I
should delete -Werror in Makefile.
I did it but it is still wrong ,
What can I do? I beg your help!

Thanks

[Liang ZHENG]

Forgot to say, I am using GCC 4.4 so in order to run cuda wich gcc 4.4 I modified something in cuda sdk

does it matter to GMAC?

2010/6/27 Liang ZHENG <liang...@gmail.com>

--
Liang ZHENG http://liangzh.cug.googlepages.com

Lab of Computational Geodynamics, Chinese Academy of Sciences

[Liang ZHENG]

I solved the problem by adding 

        case CUDA_ERROR_ECC_UNCORRECTABLE:

in compat.cpp 

I don't know if it's the right way, however it's passed.. (-:

[Liang ZHENG]

another question:

when I tried to gmcMalloc as your example like:

        gmacMalloc(&a,VECTOR_SIZE*sizeof(float));

then 

nvcc test.cu -o test

it showed;

test.cu(9): error: argument of type "float **" is incompatible with parameter of type "void **"

[Isaac Gelado]

Dear Liang,
sorry for the late reply. The error you were facing is related to new
error codes in CUDA 3.0 which were not supported by GMAC 0.0.8. We
have just released a the version 0.10 of GMAC which now supports CUDA
3.0 and 3.1. However, this new version is not working properly on
MacOS depending on the version of XCode you use.

Thank you for highlighting the error in the example in the wiki. In
that case, casting the first parameter is compulsory, so it becomes:

gmacMalloc((void **)&a, VECTOR_SIZE * sizeof(float));

Hope this helps you. Please, let us know any further problem you find
using GMAC.

Best,
Isaac

2010/6/27 Liang ZHENG <liang...@gmail.com>:

--
Isaac Gelado

'As gold which he cannot spend will make no man rich
so knowledge which he cannot apply will make no man wise'