Test CPUs, GPUs and FPGAs with a sample code in the DevCloud

[Grassi__Claudio]

Hi, my goal is to try a simple code like a Matrix Multiplication to observe the different performances in the different devices in the DevCloud, but I have found some problems... I've tried the Matrix Multiplication that is present in here https://github.com/intel/BaseKit-code-samples and it's all okay if I do

qsub -I -l nodes=1:gpu:ppn=2 -d .

(or targetting fpga), but I get problem if I target cpu. However what I need to do is to specify a precise type of CPU, GPU or FPGA when I do the test, so I've seen the list of devices in https://devcloud.intel.com/edge/get_started/devcloud/ but I get error if I try to do

qsub -I -l nodes=1:intel-hd-505:ppn=2 -d .

(or inserting any other name of devices)

Can you tell me what do I have to do to reach my goal? Thank you very much in advance

[Andrey_Vladimirov]

I understand that you are a user of Intel DevCloud for oneAPI, but you are pointing to a Getting Started guide for Intel DevCloud for the Edge, which is a separate project with its own set of devices and practices. That would explain the discrepancy.

In your case, if you want to list the available compute nodes and use their properties to request a job, try this:

pbsnodes | grep "properties" | sort | uniq

This is what I got today:

     properties = fpga_compile,xeon,plat8153,skl,ram384gb,net1gbe
     properties = fpga_runtime,xeon,gold6128,skl,ram192gb,net1gbe,fpga,arria10
     properties = gen9,xeon,e-2176g,cfl,gpu,ram64gb,net1gbe,6cores,eus0024
     properties = jupyter,batch,xeon,gold6128,skl,ram192gb,net1gbe

Here, the first property is the type of node, the second is the brand of the CPU, the third (plat8153/gold6128/e-2176g) is the identifier of the CPU model. You can use one of these. 

For questions related to Intel DevCloud for oneAPI (your project), you'll get faster response if you post in this forum: https://software.intel.com/en-us/forums/intel-devcloud