How to Make Things Faster

Chapter 21. Creating a Profile

A profile is a tabular aggregation of a trace stream. It’s a lot of detail, summarized in a form that’s easy for a person to understand. To create a profile, your first step is to select the grouping key by which you want to aggregate. The aggregation is a mechanical process that I can illustrate with an example using a simple Microsoft Excel spreadsheet.

Imagine a tracing feature that records each execution of events called A, B, and C for a program that is important for you to look at. So then, every time your program executes an event, the event’s name and duration are written to a trace file. Of course, in real life, you’d want a timestamp and some other attributes for each line, but I’ll keep it simple for the example.

So, imagine that the resulting trace stream is stored in a spreadsheet, so that column A contains event names and column B contains durations, like this:

	A	B
1	Event	Duration
2	C	0.088
3	B	0.096
4	A	0.075
5	...	...

Let’s say that you want to group this trace stream by event name. You could use the following Excel functions to create a profile by event:

	D	E	F	G
1	Event	Duration	Executions	Mean dur/exec
2	A	=SUMIF(A:A,D2,B:B)	=COUNTIF(A:A,D2)	=E2/F2
3	B	=SUMIF(A:A,D3,B:B)	=COUNTIF(A:A,D3)	=E3/F3
4	C	=SUMIF(A:A,D4,B:B)	=COUNTIF(A:A,D4)	=E4/F4
5	Total	=SUM(E2:E4)	=SUM(F2:F4)	=E5/F5

This is the kind of work that a profiler does for you. Of course, it’s trickier than this in real life, where profilers ...