Energy from Likwid-perf tool

[Sonal Aggarwal]

Hello!

I am trying to measure the Energy and Power values for my application and I am using likwid-perfctr for it. This is the snippet from the output:

  Energy [J] STAT      |    6     |    3     |    0     |   0.375    |

|   Energy DRAM [J] STAT    |    6     |    3     |    0     |   0.375    |

|     Power [W] STAT       | 39.6916  | 19.8458  |    0     |  2.48072   |

|    Power DRAM [W] STAT    | 39.6916  | 19.8458  |    0     |  2.48072   |

  I believe that the energy/power for the CPU and memory are measured separately but they are always same. Why is that?

Thanks in advance.

[Thomas Röhl]

Hi.

in order to check your problem I need the version of LIKWID and the CPU architecture. If you are using an older version as 3.1.3, please update LIKWID to the newest release.

Have you tried likwid-powermeter? Does it also show the same values for the domains?

Greetings,
Thomas

[Sonal Aggarwal]

Thanks for reaching back Thomas!

I am using the current stable version of likwid which is likwid 3.1.2.

When I use powermeter, I get the following result:

Energy consumed: 275.445 Joules

Power consumed: 73.868 Watts

Domain: DRAM

Energy consumed: 0 Joules

Power consumed: 0 Watts

The DRAM group is always zero. Are these readings correct?

The output of lscpu on my machine is:

Architecture:          x86_64

CPU op-mode(s):        32-bit, 64-bit

Byte Order:            Little Endian

CPU(s):                32

On-line CPU(s) list:   0-31

Thread(s) per core:    2

Core(s) per socket:    8

Socket(s):             2

NUMA node(s):          2

Vendor ID:             GenuineIntel

CPU family:            6

Model:                 45

Stepping:              7

CPU MHz:               1200.000

BogoMIPS:              3992.33

Virtualization:        VT-x

L1d cache:             32K

L1i cache:             32K

L2 cache:              256K

L3 cache:              20480K

NUMA node0 CPU(s):     0-7,16-23

NUMA node1 CPU(s):     8-15,24-31

Thanks

Sonal Aggarwal

Computer Science

Stony Brook University

+1-(631)-504-7609

--

---
You received this message because you are subscribed to a topic in the Google Groups "likwid-developers" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/likwid-developers/FXPQX7Np6WI/unsubscribe.
To unsubscribe from this group and all its topics, send an email to likwid-develop...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Thomas Röhl]

Hi,

the current stable version of LIKWID is 3.1.3. Please update to it. You are using a SandyBridge E/EP/EN architecture and the 3.1.2 version has a faulty RAPL counter definition for the SandyBridge architecture and always uses the PKG RAPL domain. That's why you get equal values for the PKG and DRAM domains.

You can download the new version here: http://ftp.fau.de/pub/likwid/

With the 3.1.3 version you will get additional power measurements and normally also DRAM values. But the RAPL counters can be deactivated by mainboard vendors. This could also be the reason why you do not get values from the DRAM domain.

Greetings,
Thomas

[Sonal Aggarwal]

Thanks Thomas

Sonal Aggarwal

Computer Science

Stony Brook University

+1-(631)-504-7609

[Sonal Aggarwal]

Hi Thomas,

I got the following output with Likwid 3.1.3.

Domain: PKG

Energy consumed: 278.791 Joules

Power consumed: 74.9784 Watts

Domain: DRAM

Energy consumed: 0 Joules

Power consumed: 0 Watts

Domain: PP0

Energy consumed: 217.879 Joules

Power consumed: 58.5966 Watts

The DRAM values are still zero. Are the RAPL counters deactivated for this group? Is there any workaround for this?

Sonal Aggarwal

Computer Science

Stony Brook University

+1-(631)-504-7609

[Thomas Röhl]

Hi Sonal,

as I wrote in my previous posts it could be deactivated by the mainboard vendor of your system.

I tested the 3.1.3 version on two of our SandyBridge EP systems, both with different configurations. One of them reported DRAM values, the other one also returns only 0 for DRAM.

You could try rapl-read from Vince Weaver (http://web.eece.maine.edu/~vweaver/projects/rapl/rapl-read.c) if it returns any values for DRAM. Also the newer versions of PAPI are able to measure power but this could be depended to your kernel version because PAPI uses the perf_event kernel interface. I never tried power measurements with PAPI.

Some vendors have a BIOS option to activate DRAM power measurements. In the Intel forums, I also read about new BIOS versions after the problem was reported back to the board vendor.

I hope this gives you some starting points for your further investigation.

Thomas

You received this message because you are subscribed to the Google Groups "likwid-developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email to likwid-develop...@googlegroups.com.