I downloaded CUDA Toolkit and SDK 2.1 for 32-bit Fedora 9 from CUDA zoom.
It fails at emu mode.
However, the following version of toolkit and SDK work on my 32-bit Fedora 9 at emu mode.
CUDA toolkit 2.1 beta for 32-bit Fedora 9 from this link:
http://forums.nvidia.com/index.php?showtopic=82290
CUDA SDK 2.1 for 32-bit REL 5.x from CUDA zoom.
Fails how?
Sorry, I forget to post the error message.
I typed the command “make emu=1” to build the SDK.
Then I run the matrixMul example in bin/linux/emurelease folder.
I got the following error:
cutilCheckMsg() CUTIL CUDA error: Kernel execution failed in file <matrixMul.cu>, line 140 : feature is not yet implemented.
Other examples got the same error, complaining “feature is not yet implemented”.
I tried to recompile the example codes with “make emu=1”, but still got the same error.
i am also getting the same error on opensuse 11
i don’t have the NVIDIA GPU but i want to run on my computer.
i did make emu=1
some of the examples are running though
SobelFiilter
oceanFFT
clock
boxFilter
simpleMultiGpu
others report an error in the middle
./alignedTypes
Allocating memory…
Generating host input data array…
Uploading input data to GPU memory…
Testing misaligned types…
uint8…
cutilCheckMsg() CUTIL CUDA error: testKernel() execution failed
in file <alignedTypes.cu>, line 223 : feature is not yet implemented.
./particles
grid: 64 x 64 x 64 = 262144 cells
Segmentation fault
were you able to solve it?
I am not able to make 2.1’s emu mode work on my machine.
I am using earlier version of toolkit and SDK now.