Profiling tools
Learn to profile and find bottlenecks with Xdebug and XHProf.
What is profiling for?
Profiling measures where running code spends time. It is not debugging — it is bottleneck hunting. Xdebug gives step tracing and stack traces during development, while XHProf produces a function call graph with low overhead, suitable for production-like runs.
The output lists each function's call count (ct), wall time (wt) and CPU time (cpu). The hot path is the chain of calls that consumes the most time; you read a profile by following it.
Call graph and hot path
A profile report can be turned into a call graph that shows how functions call each other. Nodes are functions, edges are the called from relation. The hot path is the root-to-leaf chain with the highest total time.
XHProf's raw data is enough to rebuild this graph; functions with high call count (ct) but low wall time (wt) are not the target — the chain where wt accumulates is. Sort by total wt to avoid chasing the wrong target.
Exercise Check XHProf availability
Check 'xhprof_enable' with function_exists. If present call xhprof_enable(), run str_repeat('a', 1000), collect $data = xhprof_disable() and print echo count($data) . "\n". Otherwise print echo "XHProf not installed\n".
Show example solution
If function_exists('xhprof_enable') is true call xhprof_enable(); str_repeat('a', 1000); $data = xhprof_disable(); echo count($data) . "\n";. In the else branch echo "XHProf not installed\n";.
Example solution:
<?php
if (function_exists('xhprof_enable')) {
xhprof_enable();
str_repeat('a', 1000);
$data = xhprof_disable();
echo count($data) . "\n";
} else {
echo "XHProf not installed\n";
}Test yourself
-
Complete the xhprof_enable call
____(XHPROF_FLAGS_CPU | XHPROF_FLAGS_MEMORY); $data = xhprof_disable();Show answer
xhprof_enable(XHPROF_FLAGS_CPU | XHPROF_FLAGS_MEMORY); $data = xhprof_disable(); -
Why does XHProf's low overhead matter?
- So it can run in production-like conditions
- To make debugging easier
- To write less code
- To hide stack traces
Show answer
So it can run in production-like conditions — Correct. Low overhead lets you profile in production-like conditions.
-
Match the profiling concepts
- ct
- Call count
- wt
- Wall time
- hot path
- Slowest call chain
- Xdebug
- Step tracing