Deeper look at CPU utilization : TMAM Example

This is a followup to my previous post on Deeper look at CPU utilization : The power of PMU events .

So let’s go back to my previous example using the General Exploration View of Intel VTune  :

One of the highlighted metric is “DTLB Overhead” and it’s an estimation of the performance penalty paid for missing the first-level data TLB (DTLB) that includes hitting in the second-level data TLB (STLB) as well as performing a hardware page walk on an STLB miss.

One of the recommendation is to use the famous Large Page sizes  ! So let’s do it !

After restarting my database instance with Large Page this time and rerunning my program this is what we get :

 

Using Large Page size has significantly reduced the “DTLB Overhead” metric and it’ is no more highlighted ! We also have slightly  improved our CPI  from 0.553 to 0.532 and reduced our execution time from 5.37 to 5.20.

This was just a very simple example of the utilization of TMAM !

That’s it 😀

REF:

• Clockticks Vs. Pipeline Slots Based Metrics
• Understanding How General Exploration Works
• General Exploration View
• Tuning Guides and Performance Analysis Papers
• DTLB Overhead