nvml: Update getDeviceFieldValue to check value of nvmlReturn and use correct valueType - #564

I'm reviewing this PR.

I tested these changes on the NVIDIA GH200 480GB device (Grace-Hopper architecture).

The utility papi_command_line indeed reports reasonable event counts:

~/papi/src$ papi_command_line nvml:::NVIDIA_GH200_480GB:device_0:gpu_inst_power

This utility lets you add events from the command line interface to see if they work.

Successfully added: nvml:::NVIDIA_GH200_480GB:device_0:gpu_inst_power

nvml:::NVIDIA_GH200_480GB:device_0:gpu_inst_power : 	112654 mW

----------------------------------

However, the nvml component test HelloWorld returns the following error:

~/papi/src$ ./components/nvml/tests/HelloWorld
Call to cudaGetDeviceCount(&num_devices) on line 148 failed with error code 805.

@Treece-Burgess Does this error occur on your end?

I tested these changes on the NVIDIA GH200 480GB device (Grace-Hopper architecture).

The utility papi_command_line indeed reports reasonable event counts:

~/papi/src$ papi_command_line nvml:::NVIDIA_GH200_480GB:device_0:gpu_inst_power

This utility lets you add events from the command line interface to see if they work.

Successfully added: nvml:::NVIDIA_GH200_480GB:device_0:gpu_inst_power

nvml:::NVIDIA_GH200_480GB:device_0:gpu_inst_power : 	112654 mW

----------------------------------

However, the nvml component test HelloWorld returns the following error:

~/papi/src$ ./components/nvml/tests/HelloWorld Call to cudaGetDeviceCount(&num_devices) on line 148 failed with error code 805.

@Treece-Burgess Does this error occur on your end?

This error occurs even in the sample code /packages/cuda/13.0.0/extras/CUPTI/samples/concurrent_profiling/concurrent_profiling.cu provided by NVIDIA on Hopper1 at Oregon (1 * GH200):

[tburgess@hopper1 concurrent_profiling]$ ./concurrent_profiling 


Error: concurrent_profiling.cu:218: Function cudaGetDeviceCount(&numDevices) failed with error(805): MPS client failed to connect to the MPS control daemon or the MPS server.

[tburgess@hopper1 concurrent_profiling]$ 

If I go to Illyad on Oregon (1 * H100) with Cuda Toolkit 13.0.0, this error does not occur:

[tburgess@illyad tests]$ ./HelloWorld 
Running the nvml component test HelloWorld.cu
PAPI version being used for this test: 7.3.0
The nvml component is assigned to component index: 2

Starting string:
Hello World!

Proceeding to mangle the starting string.
The mangled string is: HdjikhjcZ

Proceeding to unmangle the mangled string.
The unmangled string is: Hello World!

After PAPI_stop, the event nvml:::NVIDIA_H100_PCIe:device_0:graphics_clock produced the value 		1755

This appears to be a system issue rather than a PAPI issue.