feature request: add support for `nvmlDeviceGetGraphicsRunningProcesses_v2` about nvml-wrapper HOT 5 CLOSED

Hey! I have seen this and I will give it some thought, but I won't have time to reply for another couple of days 🙂.

from nvml-wrapper.

Thank you for your reply! Please feel free to tell me it when you're free.

from nvml-wrapper.

@Cldfire How is it going? May I submit one?

from nvml-wrapper.

Sorry for taking so long to reply! I got busy with life stuff.

So, I've put some thought into this and I have two main concerns here:

• In order for me to accept a PR for this we would need to end up with nvmlDeviceGetGraphicsRunningProcesses_v2 being generated and present as a symbol in nvml-wrapper-sys/src/bindings.rs. I don't want to mess around with opening and loading a second library definition to support legacy function calls. Doing this will require making sure bindgen generates bindings for the functions in this ifdef, and making sure that's accomplished without modifying nvml.h in the source tree:

  nvml-wrapper/nvml-wrapper-sys/nvml.h

  Lines 9106 to 9129 in 1e943a4

  	#ifdef NVML_NO_UNVERSIONED_FUNC_DEFS
  	nvmlReturn_t DECLDIR nvmlInit(void);
  	nvmlReturn_t DECLDIR nvmlDeviceGetCount(unsigned int *deviceCount);
  	nvmlReturn_t DECLDIR nvmlDeviceGetHandleByIndex(unsigned int index, nvmlDevice_t *device);
  	nvmlReturn_t DECLDIR nvmlDeviceGetHandleByPciBusId(const char *pciBusId, nvmlDevice_t *device);
  	nvmlReturn_t DECLDIR nvmlDeviceGetPciInfo(nvmlDevice_t device, nvmlPciInfo_t *pci);
  	nvmlReturn_t DECLDIR nvmlDeviceGetPciInfo_v2(nvmlDevice_t device, nvmlPciInfo_t *pci);
  	nvmlReturn_t DECLDIR nvmlDeviceGetNvLinkRemotePciInfo(nvmlDevice_t device, unsigned int link, nvmlPciInfo_t *pci);
  	nvmlReturn_t DECLDIR nvmlDeviceGetGridLicensableFeatures(nvmlDevice_t device, nvmlGridLicensableFeatures_t *pGridLicensableFeatures);
  	nvmlReturn_t DECLDIR nvmlDeviceGetGridLicensableFeatures_v2(nvmlDevice_t device, nvmlGridLicensableFeatures_t *pGridLicensableFeatures);
  	nvmlReturn_t DECLDIR nvmlDeviceGetGridLicensableFeatures_v3(nvmlDevice_t device, nvmlGridLicensableFeatures_t *pGridLicensableFeatures);
  	nvmlReturn_t DECLDIR nvmlDeviceRemoveGpu(nvmlPciInfo_t *pciInfo);
  	nvmlReturn_t DECLDIR nvmlEventSetWait(nvmlEventSet_t set, nvmlEventData_t * data, unsigned int timeoutms);
  	nvmlReturn_t DECLDIR nvmlDeviceGetAttributes(nvmlDevice_t device, nvmlDeviceAttributes_t *attributes);
  	nvmlReturn_t DECLDIR nvmlComputeInstanceGetInfo(nvmlComputeInstance_t computeInstance, nvmlComputeInstanceInfo_t *info);
  	nvmlReturn_t DECLDIR nvmlDeviceGetComputeRunningProcesses(nvmlDevice_t device, unsigned int *infoCount, nvmlProcessInfo_v1_t *infos);
  	nvmlReturn_t DECLDIR nvmlDeviceGetComputeRunningProcesses_v2(nvmlDevice_t device, unsigned int *infoCount, nvmlProcessInfo_v2_t *infos);
  	nvmlReturn_t DECLDIR nvmlDeviceGetGraphicsRunningProcesses(nvmlDevice_t device, unsigned int *infoCount, nvmlProcessInfo_v1_t *infos);
  	nvmlReturn_t DECLDIR nvmlDeviceGetGraphicsRunningProcesses_v2(nvmlDevice_t device, unsigned int *infoCount, nvmlProcessInfo_v2_t *infos);
  	nvmlReturn_t DECLDIR nvmlDeviceGetMPSComputeRunningProcesses(nvmlDevice_t device, unsigned int *infoCount, nvmlProcessInfo_v1_t *infos);
  	nvmlReturn_t DECLDIR nvmlDeviceGetMPSComputeRunningProcesses_v2(nvmlDevice_t device, unsigned int *infoCount, nvmlProcessInfo_v2_t *infos);
  	nvmlReturn_t DECLDIR nvmlDeviceGetGpuInstancePossiblePlacements(nvmlDevice_t device, unsigned int profileId, nvmlGpuInstancePlacement_t *placements, unsigned int *count);
  	nvmlReturn_t DECLDIR nvmlVgpuInstanceGetLicenseInfo(nvmlVgpuInstance_t vgpuInstance, nvmlVgpuLicenseInfo_t *licenseInfo);
  	#endif // #ifdef NVML_NO_UNVERSIONED_FUNC_DEFS

• I don't want to clog editor autocomplete with the names of legacy methods. This means avoiding a situation where the default autocomplete experience when using this library is seeing the following:

  device.running_graphics_<|>
  
      ------------------------------------------
      device.running_graphics_processes()
      device.running_graphics_processes_v1()
      device.running_graphics_processes_v2()
      ------------------------------------------
  

  This excludes a trait-based approach for bringing these into scope because rust-analyzer is smart enough to discover method calls that could be made available by importing a trait and suggest those for autocomplete. The only way to preserve a good default autocomplete experience here would be to have a crate feature such as "legacy-functions" that could be enabled to compile support for older function calls in the wrapper.

from nvml-wrapper.

@KisaragiEffective I put together a branch meeting both of the constraints I laid out above. It's in a PR here: #39

Please give that a look and let me know if it would satisfy your needs 🙂.

from nvml-wrapper.