Temporal Sharing Overhead Benchmark Test

[aktondak]

Problem solved by the commit

https://jira.xilinx.com/browse/VITIS-12999 : Temporal Sharing Overhead Benchmark Test

Bug / issue (if any) fixed, which PR introduced the bug, how it was discovered

https://jira.xilinx.com/browse/VITIS-12999

How problem was solved, alternative solutions (if any) and why they were rejected

The enhancement was made via a 2-phase test comparison between 2 threaded 2 hardware context run and a multi-threaded temporally shared hardware contexts to calculate time overhead. More details :
https://confluence.amd.com/display/~aktondak/spatial+sharing+overhead+via+xrt-smi+validate

Verbose: Enabling Verbosity
Test 1 [00c5:00:01.1]     : temporal-sharing-overhead
    Description           : Run Temporal Sharing Overhead Test
    Benchmarks            : C:\WINDOWS\System32\DriverStore\FileRepository\ipukmddrv.inf_amd64_3df722c6631619cb\benchmark_17f0_20.json
    Details               : Threshold is 100.0 ms
    Xclbin                : C:\WINDOWS\System32\DriverStore\FileRepository\ipukmddrv.inf_amd64_3df722c6631619cb\validate_17f0_20.xclbin
    Details               : Kernel name is 'DPU_PDI_0'
                            Spatially shared contexts latency: '6596.7' ms
                            Temporally shared contexts latency: '10898.0' ms
                            Overhead: '4301.3' ms
    Test Status           : [PASSED]

Risks (if any) associated the changes in the commit

None

What has been tested and how, request additional testing if necessary

Needs testing from DSV

Documentation impact (if any)

None

[gbuildx]

Can one of the admins verify this patch?

[AShivangi]

What's the variance when the test is run in validate back to back and when it is run by itself?

[aktondak]

What's the variance when the test is run in validate back to back and when it is run by itself?

The standard deviation with current parameters used for 10 subsequent runs is 13.9ms with a mean of 4300 (~0.3%)