Hi there,
I’m pretty new to CUDA-Programming. I’ve got a large Matrix (width: 64 ints, height: ~119000) that I want to copy to device.
I stuck at getting cudaMalloc2D working. I tried to get any information or example-code from the net but no success yet.
An excerpt of my code looks like this
#include <stdio.h>
#include <math.h>
using namespace std;
#include "read_file.h" //read_file is for parsing the csv-file!
#include <cutil_inline.h>
//kernel includes
#include <my_project_kernel.cu>
int main( int argc, char** argv) {
//Get data from CSV-File
long linescount = 0;
int matrix = **getData(argv[1],linescount); //read data an put it in matrix...
unsigned int* pitch;
//Allocate memory on device
cudaMalloc2d((void**) &matrix, pitch, sizeof(int)*64, linescount);
There might be another error in usage of pitch but before I got rid of the following message, theres no chance of getting the point in using pitch.
my_project.cu(24): error: identifier "cudaMalloc2d" is undefined
1 error detected in the compilation of "/tmp/tmpxft_000043d2_00000000-4_my_project.cpp1.ii".
g++: obj/release/my_project.cu.o: No such file or directory
make: *** [../../bin/linux/release/uni02] Error 1
Some more possibly needed information:
Linux (Ubuntu) 64bit
my_project was derived from template from NVIDIA_GPU_Computing_SDK (using marginally altered Makefile)
CUDA version 3.0beta
All the other cuda-functions (like cudamalloc) work
What do I have to do (incluldes?) to get CudaMalloc2D working?
Is there any working cudaMalloc2D-example around where I can have a look?
What other alternatives do I have?
Greets