Open smith6jt-cop opened 11 months ago
Maybe this isn't a good idea. I try to run the test data and get an error:
Exception in thread "Thread-6" jcuda.CudaException: CUDA_ERROR_INVALID_PTX at KernelLauncher.checkResult(KernelLauncher.java:647) at KernelLauncher.initModule(KernelLauncher.java:433) at KernelLauncher.load(KernelLauncher.java:196) at Kernelsetup.cudaFilesetup(Auto_LF_Deconvolution.java:1756) at Psfcompute_Thread.run(Auto_LF_Deconvolution.java:1855)
Here is my info:
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2023 NVIDIA Corporation Built on Tue_Jun_13_19:42:34_Pacific_Daylight_Time_2023 Cuda compilation tools, release 12.2, V12.2.91 Build cuda_12.2.r12.2/compiler.32965470_0
PC with two Quadro K620, 2GB each
Apache Maven 3.9.6 (bc0240f3c744dd6b6ec2920b3cd08dcc295161ae) Maven home: C:\Program Files\apache-maven-3.9.6 Java version: 17.0.9, vendor: Oracle Corporation, runtime: C:\Program Files\Java\jdk-17 Default locale: en_US, platform encoding: Cp1252 OS name: "windows 10", version: "10.0", arch: "amd64", family: "windows"
Maybe this isn't a good idea. I try to run the test data and get an error:
Exception in thread "Thread-6" jcuda.CudaException: CUDA_ERROR_INVALID_PTX at KernelLauncher.checkResult(KernelLauncher.java:647) at KernelLauncher.initModule(KernelLauncher.java:433) at KernelLauncher.load(KernelLauncher.java:196) at Kernelsetup.cudaFilesetup(Auto_LF_Deconvolution.java:1756) at Psfcompute_Thread.run(Auto_LF_Deconvolution.java:1855)
Here is my info:
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2023 NVIDIA Corporation Built on Tue_Jun_13_19:42:34_Pacific_Daylight_Time_2023 Cuda compilation tools, release 12.2, V12.2.91 Build cuda_12.2.r12.2/compiler.32965470_0
PC with two Quadro K620, 2GB each
Apache Maven 3.9.6 (bc0240f3c744dd6b6ec2920b3cd08dcc295161ae) Maven home: C:\Program Files\apache-maven-3.9.6 Java version: 17.0.9, vendor: Oracle Corporation, runtime: C:\Program Files\Java\jdk-17 Default locale: en_US, platform encoding: Cp1252 OS name: "windows 10", version: "10.0", arch: "amd64", family: "windows"
I believe there might be two main issues: 1) The CUDA version may be too high, for example, you're using CUDA 12, while the version supported by jcuda might not have been updated to this latest version. The pom.xml file specifies CUDA 10.0, so adjustments will be needed during compilation. 2) The Java version might be too high; for instance, you're using Java 17, while ImageJ only supports Java 8.
The most convenient process to try is as follows:
Perhaps this is something I am doing incorrectly running build_cuda.but. Using Windows 10, I run the file and get an error that the files should be in src/java/kernel, but the files are actually in src/main/resources. My solution is to copy the files and create new directories, then the build xxx.cu to xxx.ptx proceeds as expected.