cuda-gdb 3.0:how to step into the kernel help me~~

hey,guys,

I can’t step into my kernel code when I use cuda-gdb 3.0. And my develop enviroment is: ubuntu 9.10, cuda toolkit3.0, gtx 480, gcc 4.3.

In my makefile, I use"include ./…/…/common/common.mk", and I modified the common.mk at the line which defined NVCCFLAGS:

NVCCFLAGS:= -g -G

[/codeï¼½

 also commented the 

[code]

//GENCODE_SM10 := -gencode=arch=compute_10,code=\"sm_10,compute_10\"

when I went to the statement “cudaMalloc”, I get the error " Segment fault".

If I only modified the complie command to add “-g -G” after $NVCCFLAGS just like below:

$(OBJDIR)/%.cu.o : $(SRCDIR)%.cu $(CU_DEPS)

$(VERBOSE)$(NVCC) $(GENCODE_SM10) $(GENCODE_ARCH) $(GENCODE_SM20) $(NVCCFLAGS) <b><u>-g -G </u></b>$(SMVERSIONFLAGS) -o $@ -c $<

and my .cu file and .cpp file:

CUFILES_sm_20 :=	mytest.cu \

			 

 # CUDA dependency files

 CU_DEPS	 :=  mytest_kernel.cu \

 # C/C++ source files (compiled with gcc / c++)

  CCFILES	 :=  init.cpp \

I compiled it using " make dbg=1".

I set the breakpoint in the kernel: “b mytest_kernel.cu: 15”

run…

and Cant get into the kernel.

Any suggestion?