Analyzers
Talk to fellow users of Intel Analyzer tools (Intel VTune™ Profiler, Intel Advisor)

Messuring Bandwidth

twingle123
Beginner
588 Views
Hi

Can anyone help me to find the bandwidth using vtune.

I know that using "Intel Microarchitectural Code Name Sandy Bridge - Bandwidth" anlaysis shows the bandwidth on graph. Is there any equation to get the bandwidth that shows on graph. The graph goes up and down, is there any way to get the average of the data in graph or data in graph.

Thanks in advance.
0 Kudos
5 Replies
Rob5
New Contributor II
588 Views

I may need to know more about what you would like to accomplish and the architecture you are working with. It sounds like you are looking for raw data to manually manipulate. However, Intel VTune Amplifier XE can be used to identify bandwidth issues in an application by exploring areas in the timeline. For example, hover over a bar with high bandwidth value to learn how much data was read from or written to DRAM through the on-chip memory controller.

Then use time-filtering context menu options to filter in a specific range of time during which bandwidth is notable or areas of interest. Then, switch to the core-based events that correlate with bandwidth in the grid below to determine what specific code is inducing all the bandwidth.

Some information about interpreting bandwidth data and microarchitecture differences can be found here. This document has numerous jump points to other informative bandwidth topics.

You may also find the article here and here of interest.

Can you provide more information about what you wish to accomplish or usage model?

- Rob

0 Kudos
twingle123
Beginner
588 Views
Hi Rob,

thanks for the reply.

I need information regarding how frequent an application is trying to access memory, whichis something relates to used memory bandwidth GB/sec of entire system or specific application to complete a task.

My board Specs:

Architecture: x86_64
sandy Bridge Desktop Board (Los Lunas)
0 Kudos
Rob5
New Contributor II
588 Views

Thank you for your comments. I have added this Forum thread to an existing enhancement for additional functionality in this area. We will update this forum thread when this functionality is implemented in a future release of Intel VTune Amplifier XE.

- Rob

0 Kudos
Surya_Narayanan_N_
588 Views

Are these functionalities implemented yet? IS there a way to read raw memory bandwidth data?

In xeon-phi, I see 2 ways of computing memory bandwidth. 

1. using the Formula by collecting the hardware counter parameters.

2. By using the knc-bandwidth option which collects from uncore. (this is a time varying trend)

How to correlate the results of 1 and 2? (I feel 1 gives the peak bandwidth on overall system and 2 gives time varying bandwidth)

0 Kudos
Peter_W_Intel
Employee
588 Views

knc-bandwidth analysis is ready in latest update, however the report still doesn't show data for timeline report in command line.

You have to open result on GUI, in bottom-up report to view peak Bandwidth GB/sec at any specific time (but no average value).

UNC_* event came with knc-bandwidth analysis, these counts indicates counts of memory read / write, you can summarize of them to know average, for my example:

memory read, average = 64 * Sum(UNC_F_CH?_NORMAL_READ[UNIT?]) / Elapsed time  (Read GB per second)

Elapse time = CPU_CLK_UNHALTED / cpu_frequency (seconds)   

[root@prc-mic01 peter]# amplxe-cl -report summary
amplxe: Using result path `/home/peter/r001bw'
amplxe: Executing actions 50 % Generating a report

General Exploration Metrics
---------------------------
Parameter            r001bw
-------------------  --------------
CPU Time             2.679
Clockticks           2920000000.000
 CPU_CLK_UNHALTED    2920000000.000
Cache Usage          0.0
Vectorization Usage  0.0
TLB Usage            0.0

Collection and Platform Info
----------------------------
Parameter                 r001bw                                                
------------------------  ---------------------------------------------------------------
Application Command Line  ./sampleC01                                           
User Name                 root                                                  
Operating System          Intel MIC Platform Software Stack (Built by Poky 7.0) 3.1 \n \l
Computer Name             prc-mic01-mic0                                        
Result Size               2249257                                               

CPU
---
Parameter          r001bw
-----------------  -----------------------------------------
Name               Intel(R) Xeon(R) / Core i7 980X Processor
Frequency          1090000000
Logical CPU Count  244

Summary
-------
Elapsed Time:  1.872
CPU Usage:     1.197

Event summary
-------------
Hardware Event Type  Hardware Event Count:Self  Hardware Event Sample Count:Self  Events Per Sample
-------------------  -------------------------  --------------------------------  -----------------
CPU_CLK_UNHALTED                    2920000000                               292  10000000

Uncore Event summary
--------------------
Hardware Event Type            Hardware Event Count:Self
-----------------------------  -------------------------
UNC_F_CH0_NORMAL_WRITE[UNIT0]                      89430
UNC_F_CH0_NORMAL_WRITE[UNIT1]                      91204
UNC_F_CH0_NORMAL_WRITE[UNIT2]                      92531
UNC_F_CH0_NORMAL_WRITE[UNIT3]                      91662
UNC_F_CH0_NORMAL_WRITE[UNIT4]                      91337
UNC_F_CH0_NORMAL_WRITE[UNIT5]                      91530
UNC_F_CH0_NORMAL_WRITE[UNIT6]                      89342
UNC_F_CH0_NORMAL_WRITE[UNIT7]                      89278
UNC_F_CH0_NORMAL_READ[UNIT0]                      270245
UNC_F_CH0_NORMAL_READ[UNIT1]                      276991
UNC_F_CH0_NORMAL_READ[UNIT2]                      283183
UNC_F_CH0_NORMAL_READ[UNIT3]                      290350
UNC_F_CH0_NORMAL_READ[UNIT4]                      286246
UNC_F_CH0_NORMAL_READ[UNIT5]                      280344
UNC_F_CH0_NORMAL_READ[UNIT6]                      271939
UNC_F_CH0_NORMAL_READ[UNIT7]                      274661
UNC_F_CH1_NORMAL_WRITE[UNIT0]                      91273
UNC_F_CH1_NORMAL_WRITE[UNIT1]                      88160
UNC_F_CH1_NORMAL_WRITE[UNIT2]                      88479
UNC_F_CH1_NORMAL_WRITE[UNIT3]                      88407
UNC_F_CH1_NORMAL_WRITE[UNIT4]                      87980
UNC_F_CH1_NORMAL_WRITE[UNIT5]                      88773
UNC_F_CH1_NORMAL_WRITE[UNIT6]                      89068
UNC_F_CH1_NORMAL_WRITE[UNIT7]                      91035
UNC_F_CH1_NORMAL_READ[UNIT0]                      277193
UNC_F_CH1_NORMAL_READ[UNIT1]                      255347
UNC_F_CH1_NORMAL_READ[UNIT2]                      267978
UNC_F_CH1_NORMAL_READ[UNIT3]                      271965
UNC_F_CH1_NORMAL_READ[UNIT4]                      264233
UNC_F_CH1_NORMAL_READ[UNIT5]                      274355
UNC_F_CH1_NORMAL_READ[UNIT6]                      274653
UNC_F_CH1_NORMAL_READ[UNIT7]                      284945
amplxe: Executing actions 100 % done

 

0 Kudos
Reply