Runtime resource availability examination ?

Runtime resource availability examination ?

The module I've been working on is designed with the intention to work in multiple instances simultaneously. 

Let's for a moment assume that Intel Media SDK hw-based implementation gives superb performance (and is preferred to run if available), whereas sw-based implementation performance falls behind the original/native implementation (C/C++/SSE/MMX...). 

For these reasons it would be beneficial to have some kind of run-time system examination mechanism which would report resource (GPU/surfaces...) availability, upon which the instantiation of different module flavors would be decided. 

Is there a good example of such 'runtime monitor' (source code example preferred) ?

3 posts / 0 nouveau(x)
Dernière contribution
Reportez-vous à notre Notice d'optimisation pour plus d'informations sur les choix et l'optimisation des performances dans les produits logiciels Intel.

Hi,

Intel Media SDK does not provide any way of querying the GPU or CPU for current load.

That said, for the SW execution case, developers may benefit from using the Media SDK Join feature which will help share the CPU resources appropriately.

Regards,
Petter 

>>...Intel Media SDK does not provide any way of querying the GPU or CPU for current load...

I very simple Resource Monitoring ( RM ) software subsystem could be implemented using Win32 or WMI ( Windows Management Instrumentation ) APIs for CPU. In case of GPU take a look at NVIDIA's SDKs since they provide lots of functions in their APIs. For example, in NVRESULT enumeration ( see NvApiError.h ) there is an error code NV_ALLOCATED and it means:

'...the request cannot be completed as the resource is already allocated...'.

Connectez-vous pour laisser un commentaire.