CUDA - Revision history

Boiarino at 18:52, 15 September 2026

2026-09-15T18:52:15Z

← Older revision		Revision as of 14:52, 15 September 2026
Line 32:		Line 32:

	[root@clonfarm11 ~]# dkms install -m nvidia -v 615.71.09		[root@clonfarm11 ~]# dkms install -m nvidia -v 615.71.09

			CUDA should be installed in ''/usr/local/cuda-13.4'' directory. Users should include following in .cshrc:

			setenv CUDA /usr/local/cuda-13.4
			setenv PATH ${PATH}:${CUDA}/bin
			setenv LD_LIBRARY_PATH "$LD_LIBRARY_PATH":$CUDA/lib64


	Reboot machine.		Reboot machine.

Boiarino at 18:49, 15 September 2026

2026-09-15T18:49:28Z

← Older revision		Revision as of 14:49, 15 September 2026
Line 33:		Line 33:
	[root@clonfarm11 ~]# dkms install -m nvidia -v 615.71.09		[root@clonfarm11 ~]# dkms install -m nvidia -v 615.71.09

	[root@clonfarm11 ~]# nvidia-smi		Reboot machine.

			Check installation:

			[root@clonfarm11 ~]# nvidia-smi
	Tue Sep 15 14:26:10 2026		Tue Sep 15 14:26:10 2026
	+-----------------------------------------------------------------------------------------+		+-----------------------------------------------------------------------------------------+

Boiarino at 18:44, 15 September 2026

2026-09-15T18:44:08Z

← Older revision		Revision as of 14:44, 15 September 2026
Line 9:		Line 9:
	'''Install software:'''		'''Install software:'''

	Disable jlab/rhel repo with cuda first, and get cuda repo from nvidia website:		Disable jlab/rhel repo with cuda first, and get cuda repo from nvidia website https://developer.nvidia.com/cuda-downloads. Click Linux -> x86_64 -> RHEL -> 9 -> rpm(network). Follow instructions, for example:

			dnf config-manager --add-repo https://developer.download.nvidia.com/compute/cuda/repos/rhel9/x86_64/cuda-rhel9.repo
			dnf clean all
			dnf -y install cuda-toolkit-13-4

			Check that repo is enabled (if not, see notices in [[yum]]):

	[root@clonfarm11 ~]# yum repolist all		[root@clonfarm11 ~]# yum repolist all

Boiarino at 18:36, 15 September 2026

2026-09-15T18:36:50Z

← Older revision		Revision as of 14:36, 15 September 2026
Line 9:		Line 9:
	'''Install software:'''		'''Install software:'''

	Disable jlab/rhel repo with cuda first, and get cuda repo from nvidia website ~~! Than, install~~:		Disable jlab/rhel repo with cuda first, and get cuda repo from nvidia website:

	[root@clonfarm11 ~]# yum repolist all		[root@clonfarm11 ~]# yum repolist all
Line 15:		Line 15:
	cuda-rhel9-x86_64 cuda-rhel9-x86_64 enabled		cuda-rhel9-x86_64 cuda-rhel9-x86_64 enabled
	.....		.....

			Install:

	dnf install cuda		dnf install cuda

Boiarino at 18:36, 15 September 2026

2026-09-15T18:36:18Z

← Older revision		Revision as of 14:36, 15 September 2026
Line 52:		Line 52:


	~~'''~~Compile following test program by ''nvcc cudatest.cu -o cudatest'' ~~and run it'''~~		Compile following test program by command ''nvcc cudatest.cu -o cudatest'':

	/* cudatest.cu */		/* cudatest.cu */
Line 103:		Line 103:
	return 0;		return 0;
	}		}


			Run it:

			clonfarm11:boiarino> ./cudatest
			[1] Program started on CPU...
			[2] Found 2 CUDA-capable GPU(s).
			[3] Launching GPU kernel...
			Hello World from the GPU!
			[4] Program finished successfully on CPU.

Boiarino at 18:34, 15 September 2026

2026-09-15T18:34:45Z

← Older revision		Revision as of 14:34, 15 September 2026
Line 52:		Line 52:


	'''Compile and run ~~test program~~'''		'''Compile following test program by ''nvcc cudatest.cu -o cudatest'' and run it'''

	/* cudatest.cu */		/* cudatest.cu */

Boiarino at 18:33, 15 September 2026

2026-09-15T18:33:37Z

← Older revision		Revision as of 14:33, 15 September 2026
Line 50:		Line 50:
	\| No running processes found \|		\| No running processes found \|
	+-----------------------------------------------------------------------------------------+		+-----------------------------------------------------------------------------------------+


			'''Compile and run test program'''

			/* cudatest.cu */

			#include <iostream>
			#include <cstdio>

			// A tiny GPU kernel that actually prints from the GPU
			__global__ void helloFromGPU() {
			// Only let the very first thread print to avoid flooding the console
			if (threadIdx.x == 0 && blockIdx.x == 0) {
			printf("Hello World from the GPU!\n");
			}
			}

			int main() {
			std::cout << "[1] Program started on CPU..." << std::endl;

			// Check if a CUDA-capable GPU is even visible
			int deviceCount = 0;
			cudaError_t err = cudaGetDeviceCount(&deviceCount);

			if (err != cudaSuccess) {
			std::cout << "❌ CUDA Error: " << cudaGetErrorString(err) << std::endl;
			std::cout << "Reason: Your system might lack an NVIDIA GPU, or drivers are missing." << std::endl;
			return 1;
			}

			std::cout << "[2] Found " << deviceCount << " CUDA-capable GPU(s)." << std::endl;

			// Launch 1 block with 1 thread
			std::cout << "[3] Launching GPU kernel..." << std::endl;
			helloFromGPU<<<1, 1>>>();

			// Check if the kernel launch itself failed
			err = cudaGetLastError();
			if (err != cudaSuccess) {
			std::cout << "❌ Kernel Launch Error: " << cudaGetErrorString(err) << std::endl;
			return 1;
			}

			// Force CPU to wait and flush the GPU's printf buffer
			err = cudaDeviceSynchronize();
			if (err != cudaSuccess) {
			std::cout << "❌ Synchronization Error: " << cudaGetErrorString(err) << std::endl;
			return 1;
			}

			std::cout << "[4] Program finished successfully on CPU." << std::endl;
			return 0;
			}

Boiarino at 18:30, 15 September 2026

2026-09-15T18:30:29Z

← Older revision		Revision as of 14:30, 15 September 2026
Line 9:		Line 9:
	'''Install software:'''		'''Install software:'''

	Disable jlab/rhel repo with cuda first, and get cuda repo from nvidia website !		Disable jlab/rhel repo with cuda first, and get cuda repo from nvidia website ! Than, install:

	[root@clonfarm11 ~]# yum repolist all		[root@clonfarm11 ~]# yum repolist all

Boiarino at 18:30, 15 September 2026

2026-09-15T18:30:11Z

← Older revision		Revision as of 14:30, 15 September 2026
Line 1:		Line 1:
	Some notes for RHEL9.8 on clonfarm11:		Some notes for RHEL9.8 on clonfarm11:

	Find installed cards:		'''Find installed cards:'''

	[root@clonfarm11 ~]# lspci \| grep -i nvidia		[root@clonfarm11 ~]# lspci \| grep -i nvidia
Line 7:		Line 7:
	31:00.0 3D controller: NVIDIA Corporation TU104GL [Tesla T4] (rev a1)		31:00.0 3D controller: NVIDIA Corporation TU104GL [Tesla T4] (rev a1)

	Install software:		'''Install software:'''

			Disable jlab/rhel repo with cuda first, and get cuda repo from nvidia website !

	[root@clonfarm11 ~]# yum repolist all		[root@clonfarm11 ~]# yum repolist all

Boiarino at 18:27, 15 September 2026

2026-09-15T18:27:50Z

← Older revision		Revision as of 14:27, 15 September 2026
Line 14:		Line 14:
	.....		.....

			dnf install cuda
	dnf install nvidia-driver		dnf install nvidia-driver
	dnf install nvidia-driver-cuda		dnf install nvidia-driver-cuda