Analyzers
Support for Analyzers (Intel VTune™ Profiler, Intel Advisor, Intel Inspector)
Announcements
This community is designed for sharing of public information. Please do not share Intel or third-party confidential information here.
4682 Discussions

trace cache miss performance impart

jrzhou
Beginner
222 Views
Could anyone explain why the trace cache miss performance impact is defined as "((Trace Cache Misses*20) / Clockticks)*100"? Where is the formula coming from?

If I want to computer trace cache miss rate, should I use "trace cache misses / instruction retired" or "trace cache misses / uops retired"?
0 Kudos
1 Reply
david_a_levinthal
222 Views
The tuning assistant tries to estimate the performance impact due to that particular event using a simplistic model. Basically, if the trace cache occasionally has to operate from build mode (because of a trace cache miss), what fraction of the cycles (performance impact) are due to this particular situation, assuming trace cache miss is the sole cause of the performance impact.

As for the trace cache miss rate question. The answer depends on how you want to interpret the data. The more important question is what are you trying to determine and what are you going to do with the information. I suspect that the rate of micro-ops/op averaged over any program doesn't vary much so it may not matter.
If trace cache misses are estimated to be a signifcant issue for running your program I would reccomend identifying modules with relative high rates and recompiling them with profile guided feedback as that will allow the compiler to feed the processor a binary that is organized to better flow through your logical branching structure than what is being created by default
Reply