When will cuda support gcc 4.4

Accelerated Computing CUDA CUDA Programming and Performance
#1

I’ve seen threads here of people with work-arounds for gcc 4.4 using CUDA 3.0. Will CUDA officially support gcc 4.4 in a planned version, and if so when?

I’ve run into a bug in gcc 4.3 related to OpenMP so I really need 4.4.x, and would prefer official support.

By the way, 4.3.0 is just about unmaintained (as much as is possible with free software).

Thanks,
Matt

#2

I’m bumping this post, with a similar question, but regarding gcc 4.5.

I’ve been successfully able to use gcc 4.4 with CUDA, thanks to the work-arounds found somewhere on this forum.

I would like to use gcc 4.5, and probably the workarounds fo gcc 4.4 would work, however, binaries generated by gcc 4.5 may only be debugged using gdb 7,0 or later, but cuda-gdb is based on gdb 6.6 and therefore it may be unusable :(.

Are there any plans on upgrading cuda-gdb?

Thanks

Witold Andrzejewski

#3

Let me add a few lines here. I managed to build the SDK on 64 bit Fedora 12 with the
following patches, extracted from other messages in this forum:

  • in common/common.mk, line 92:
    NVCCFLAGS := --compiler-options -fno-inline

  • in common/common.mk, line 275:
    LIB += -lcuda {OPENGLLIB} (PARAMGLLIB) (RENDERCHECKGLLIB) (CUDPPLIB) ${LIB} -L/usr/lib64/nvidia

  • in /usr/local/cuda/bin/nvcc.profile, line 7:
    INCLUDES += “-I$(TOP)/include” “-I$(TOP)/include/cudart” $(SPACE) “-D__builtin_stdarg_start=__builtin_va_start”

I’m not claiming originality, but I thought a summary would be helpful.

Bram