I tried all the steps you mentioned but still I get error as :
run:
Error while loading native library with base name “JCudaRuntime”
Operating system name: Windows Vista
Architecture : x86
Architecture bit size: 32
Exception in thread “main” java.lang.UnsatisfiedLinkError: Could not load native library
at jcuda.LibUtils.loadLibrary(LibUtils.java:79)
at jcuda.runtime.JCuda.assertInit(JCuda.java:225)
at jcuda.runtime.JCuda.cudaMalloc(JCuda.java:1775)
at cuda.JCUDARuntimTest.main(JCUDARuntimTest.java:15)
Java Result: 1
BUILD SUCCESSFUL (total time: 0 seconds)
I have copied the libraries to all the paths ( even replicated it to system32 and project directory ) But still I get this error . When I run the sample projects of NVIDIA sdk I can run it . Its says 1 Cuda device found .
Please help me . I have to paraellize my Random forest code in JCUDA .
The same question as in this post: Which version of the CUDA driver and toolkit do you have installed, and are you sure you have the matching JCuda version?
Windows Vista 32 bit
I Installed the Developer Driver (263.06), and the Cuda Toolkit 3.2 (build 16)
Am using the latest version of JCuda binaries from jcuda website (0.3.2)
I have the dlls in the program directory (have tried explicitly specifying the path too)
I’m able to run the code sample examples on the nVidia site (Vector Addition, etc, etc) fine (they tell me everything passes).
Is there anything else I can try?
Note I did have this working before on version 0.3.1 (build 9), but I got a new graphics card and had to reinstall drivers, etc, so figured I may as well upgrade to version 0.3.2. Not sure why it isn’t working now (I completely uninstalled the old toolkit etc, before installing the new versions).
Maybe there’s some old DLL left abandoned somewhere - maybe if you copied the CUDA DLL from the Toolkit directory to any other (for example the Windows) directory… If you are sure that there is no old DLL (maybe by searching for “cudart*.dll” on the PC), you may want to copy the Version 32_16 of the DLL into your program directory as well (just for testing!).
It’s a pity that the JVM does not give more specific information when an UnsatisfiedLinkError occurs -_-
I copied the cudart dll (& the other ones) to the project directory and it seems to work fine now.
Btw, when I ran a search for cudart, I got:
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v3.2\bin\cudart32_32_16 (the correct one)
C:\Program Files\NVIDIA Corporation\PhysX\Common\cudart32_30_9 (this should be ok?)
[QUOTE=James]I copied the cudart dll (& the other ones) to the project directory and it seems to work fine now.
[/quote]
OK, it’s good to know that it works in general, although copying this DLL should not be necessary - and may lead to the same problem again after the next update, when it finds this …_32_16 version although it needs a version like …_33_XX …
C:\Program Files\NVIDIA Corporation\PhysX\Common\cudart32_30_9 (this should be ok?)
If this path is located in the „PATH“ einvironment variable, it **might ** be the case that - for whatever reason - it attempts to use this „wrong“ DLL, but I’m not quite sure about that…