gpu_info.c File Reference

Prints information on available hardware and software on the cluster relating to GPUs. More...

#include "io.h"
#include "error.h"
#include "gpu.h"
#include "mpi.h"
#include <mpi-ext.h>

Include dependency graph for gpu_info.c:

Functions
void	print_device_count_info (input_gpu gpu_var_init)
	Query number of GPUs and print related information.
void	print_software_info (cudaDeviceProp device_prop)
	Print CUDA driver and runtime version information.
void	print_global_memory_info (cudaDeviceProp device_prop, input_gpu gpu_var_init)
	Print Global memory information including bandwidth paramters and supported features.
void	print_shared_memory_info (cudaDeviceProp device_prop)
	Print information on shared memory.
void	print_cache_info (cudaDeviceProp device_prop, input_gpu gpu_var_init)
	Print (L2) cache information.
void	print_constant_memory_info (cudaDeviceProp device_prop)
	Print information on constant memory, in particular amount available, alignment, texture and layered texture memory paramters.
void	print_memory_info (cudaDeviceProp device_prop, input_gpu gpu_var_init)
	Prints all memory related info, here on global, shared memory, cache and constant memory, see resp. functions for more information.
void	print_compute_info (cudaDeviceProp device_prop, input_gpu gpu_var_init)
	Prints information on compute parameters.
void	print_supported_features (cudaDeviceProp device_prop)
	Checks for a number of other supported features and prints information on them.
void	print_performance_metrics ()
	Prints peak performance metric estimates. This allows to check, whether we are fully utilizing the capabilities of the hardware.
void	print_hardware_info (cudaDeviceProp device_prop, input_gpu gpu_var_init)
	Prints all information on hardware, meaning memory, compute and features (no driver and runtime info). This assumes CUDART >= 4000.

Variables
const char *	sComputeMode []

Detailed Description

Prints information on available hardware and software on the cluster relating to GPUs.

Function Documentation

print_cache_info()

void print_cache_info	(	cudaDeviceProp	device_prop,
		input_gpu	gpu_var_init )

Print (L2) cache information.

Parameters

cudaDeviceProp	A CUDA class containing information on the device.
input_gpu	A struct containing parameters on the current GPU.

print_compute_info()

void print_compute_info	(	cudaDeviceProp	device_prop,
		input_gpu	gpu_var_init )

Prints information on compute parameters.

Parameters

cudaDeviceProp	A CUDA class containing information on the device.
input_gpu	A struct containing parameters on the current GPU.

print_constant_memory_info()

void print_constant_memory_info

(

cudaDeviceProp

device_prop

)

Print information on constant memory, in particular amount available, alignment, texture and layered texture memory paramters.

Parameters

cudaDeviceProp

A CUDA class containing information on the device.

print_device_count_info()

void print_device_count_info

(

input_gpu

gpu_var_init

)

Query number of GPUs and print related information.

Parameters

input_gpu

A struct containing parameters on the current GPU.

print_global_memory_info()

void print_global_memory_info	(	cudaDeviceProp	device_prop,
		input_gpu	gpu_var_init )

Print Global memory information including bandwidth paramters and supported features.

Parameters

cudaDeviceProp	A CUDA class containing information on the device.
input_gpu	A struct containing parameters on the current GPU.

print_hardware_info()

void print_hardware_info	(	cudaDeviceProp	device_prop,
		input_gpu	gpu_var_init )

Prints all information on hardware, meaning memory, compute and features (no driver and runtime info). This assumes CUDART >= 4000.

Parameters

input_gpu

A struct containing parameters on the current GPU.

print_memory_info()

void print_memory_info	(	cudaDeviceProp	device_prop,
		input_gpu	gpu_var_init )

Prints all memory related info, here on global, shared memory, cache and constant memory, see resp. functions for more information.

Parameters

cudaDeviceProp	A CUDA class containing information on the device.
input_gpu	A struct containing parameters on the current GPU.

print_performance_metrics()

void print_performance_metrics

(

)

Prints peak performance metric estimates. This allows to check, whether we are fully utilizing the capabilities of the hardware.

Code snippet based on: https://developer.nvidia.com/blog/how-query-device-properties-and-handle-errors-cuda-cc/ Access 2023-02-27

print_shared_memory_info()

void print_shared_memory_info

(

cudaDeviceProp

device_prop

)

Print information on shared memory.

Parameters

cudaDeviceProp

A CUDA class containing information on the device.

print_software_info()

void print_software_info

(

cudaDeviceProp

device_prop

)

Print CUDA driver and runtime version information.

Print CUDA software version information (runtime and driver version)

Parameters

input_gpu

A struct containing parameters on the current GPU.

print_supported_features()

void print_supported_features

(

cudaDeviceProp

device_prop

)

Checks for a number of other supported features and prints information on them.

Parameters

cudaDeviceProp

A CUDA class containing information on the device.

Variable Documentation

sComputeMode

const char* sComputeMode[]

Initial value:

= {
    "Default (multiple host threads can use ::cudaSetDevice() with device simultaneously)",
    "Exclusive (only one host thread in one process is able to use ::cudaSetDevice() with this device)",
    "Prohibited (no host thread can use ::cudaSetDevice() with this device)",
    "Exclusive Process (many threads in one process is able to use ::cudaSetDevice() with this device)",
    "Unknown",
    NULL
}