how to create performance benchmarks....

[k_g_v_s_prasad]

Hello,
To check the performance improvements we need a benchmark. So I want to create a benchmark for my activity. How can I do this..? Where shall I do this..?
Regards,
Prasad.

[harmeet-duggal]

Quoting - k.g.v.s.prasad

Hello,
To check the performance improvements we need a benchmark. So I want to create a benchmark for my activity. How can I do this..? Where shall I do this..?
Regards,
Prasad.

any answers related this issue?

[Peter_W_Intel]

Quoting - k.g.v.s.prasad

Hello,
To check the performance improvements we need a benchmark. So I want to create a benchmark for my activity. How can I do this..? Where shall I do this..?
Regards,
Prasad.

Hi Prasad,

VTune Analyzer doesn't generate any benchmark,howeveryou cancreate manyactivitieswhich can use (record) variant parameters foryour application, inspect CPU_CLK_UNHALED.CORE samples/events to know time spending in yourprocess/modules.

The user can compare sampling results from same activity (multiple runs) or different activities. That is, open one sampling result and drag-drop another sampling result to existing "processes report" window. Thus, you can compare two sessions'results in one window. Note that samplingresults have their unique session IDs.

Thanks, Peter

[Vladimir_T_Intel]

You can also add a timer measurements to the code you are going to investigate: "envelope" the performance critical portion of your code with start/stop timer calls and then check it for improvements once you've changed the program. With the method proposed by Peter you can figure out what is the exact place/function/line you need to look at for performance improvement.

[bishgada]

Quoting - Peter Wang (Intel)

Hi Prasad,

VTune Analyzer doesn't generate any benchmark,howeveryou cancreate manyactivitieswhich can use (record) variant parameters foryour application, inspect CPU_CLK_UNHALED.CORE samples/events to know time spending in yourprocess/modules.

The user can compare sampling results from same activity (multiple runs) or different activities. That is, open one sampling result and drag-drop another sampling result to existing "processes report" window. Thus, you can compare two sessions'results in one window. Note that samplingresults have their unique session IDs.

Thanks, Peter

Hi,
I was wonderig how does the tool deal with the drag & drop when the two activity runs results with a different order or even a different number of rows?

Besides that it seems quite usefull method. (I used to open both results and just look at them in sequence... now everything becomes parallel :) )
Thanks,
Guy.

[Peter_W_Intel]

Quoting - bishgada

Hi,
I was wonderig how does the tool deal with the drag & drop when the two activity runs results with a different order or even a different number of rows?

Besides that it seems quite usefull method. (I used to open both results and just look at them in sequence... now everything becomes parallel :) )
Thanks,
Guy.

I have toemphasize when you compare two results.

1. Two sample sessions have to use same CPU events.
2. Some module(s) may be in firstresult (these modules may have 1-3 samples only),but maynot in second result (zero sample). It doesn't matter, because runnig environment was changed.

Remember activity IDs, so you can know which is first, which is second.

Regards, Peter