[...]CUDA cuPrintf » Jeremy Kemp[...]…

As far as I’m aware, there hasn’t been an update since its initial release.

Do you know if there are updates out there for newer architectures?

Thanks for getting back to me. Unfortunately, we’re don’t have Fermi architecture. We’re using a Tesla C1060 with a CPU running Ubuntu 9.10, and Cuda 3.1. This was the file that we’re trying to test. Ideally, we want it to print out “42″.
———————————————–
#include
#include “cuPrintf.cu”

__global__ void printTest() {

int tid = 42;
cuPrintf(“%d\n”, tid);
}

int main() {
cudaPrintfInit();
printTest<<>>();
cudaPrintfDisplay(stdout, true);
fflush(stdout);
cudaPrintfEnd();
return 0;
}

———————————————–

Incidentally, we tried adding the cuPrintf files to our Makefile (standard NVIDIA make) which gave us redefinition errors, so our makefile looks like this:

———————————————–
# Add source files here
EXECUTABLE := printTest
# Cuda source files (compiled with cudacc)
CUFILES := printTest.cu
# C/C++ source files (compiled with gcc / c++)
CCFILES :=

include ../../common/common.mk

———————————————–

Do you see anything obvious?

That is strange that the code doesn’t work. What GPU are you using? Fermi GPU’s now support the standard printf statement, without the need for cuPrintf.

So I know this thread is a little old, but it is no less relevant.

I’m trying to get print statements to work with cuda 3.1 (on Ubuntu linux 9.10) and having very little success. I tried copying your code into a toy project, and it compiled fine, but it didn’t print anything. I’ve done a little digging into the source code, but I can’t find anything obvious. Any idea what might be going on?