PHP Profiling

Install Blackfire

# Install Blackfire agent
curl -fsSL https://get.blackfire.io | sudo bash

# Install PHP probe
blackfire agent:install

# Verify installation
blackfire doctor

Profile with Blackfire

# Profile a URL
blackfire curl https://magento.example.com/

# Profile with custom options
blackfire --title="Homepage" curl https://magento.example.com/

# Compare two profiles
blackfire compare profile1.prof profile2.prof

Blackfire in Code

Blackfire\Profile::start();

// Code to profile
$product = $this->productRepository->get('TEST-001');
$products = $this->productCollectionFactory->create()->load();

Blackfire\Profile::stop();

Key Points

• Blackfire shows wall time, CPU, memory, I/O
• Compare profiles before/after optimization
• Use thresholds to detect regressions
• Integration with CI/CD for automated profiling

Enable Profiling

; php.ini
[xdebug]
xdebug.mode=profile
xdebug.output_dir=/tmp/xdebug
xdebug.profiler_output_name=profile.%p-%t

Analyze with KCacheGrind

# Convert to callgrind format
php vendor/bin/pprof --convert callgrind /tmp/xdebug/profile.xdebug

# Open in KCacheGrind
kcachegrind /tmp/xdebug/profile.xdebug

# Or use QCachegrind (macOS)
qcacheegrind /tmp/xdebug/profile.xdebug

Key Metrics

• Self time: time spent in function itself
• Inclusive time: time including sub-calls
• Call count: number of times function called
• Memory: peak memory usage

Key Points

• Xdebug profiler is free and built-in
• Shows function-level performance data
• Use callgrind format for visualization
• Disable in production (performance impact)

Generate Flame Graphs

# Using phpspy
php vendor/bin/phprof /tmp/xdebug/profile.xdebug > /tmp/stacks.txt

# Generate flame graph
./flamegraph.pl /tmp/stacks.txt > /tmp/flamegraph.svg

# Open in browser
open /tmp/flamegraph.svg

Read Flame Graphs

• X-axis: percentage of time
• Y-axis: call stack depth
• Wide blocks = more time spent
• Color: random (not meaningful)

Key Points

• Flame graphs visualize call stack overhead
• Wide blocks indicate optimization targets
• Use for identifying hot paths
• Interactive in browser (zoom, search)

Common Bottlenecks

// 1. Database queries
// - N+1 queries
// - Missing indexes
// - Slow queries

// 2. Memory usage
// - Large collections
// - Memory leaks
// - Unbounded data loading

// 3. CPU-intensive code
// - Complex calculations
// - String processing
// - Regex operations

// 4. I/O operations
// - File system access
// - External API calls
// - Network requests

Profiling Workflow

// 1. Identify hot path
// 2. Measure baseline
// 3. Optimize
// 4. Measure improvement
// 5. Repeat

use Magento\Framework\Profiler;

Profiler::start('optimization_test');
// ... code ...
Profiler::stop('optimization_test');

Key Points

• Profile before optimizing
• Focus on biggest bottlenecks first
• Measure impact of each change
• Avoid premature optimization