PHP Compilation Pipeline

Also Known As

TL;DR

Explanation

PHP execution pipeline: (1) Lexer: source text → token stream (T_FUNCTION, T_STRING, T_WHITESPACE). (2) Parser: tokens → Abstract Syntax Tree using LALR(1) grammar. (3) Compiler: AST → opcode array (zend_op structs — ZEND_ADD, ZEND_CALL, ZEND_RETURN). (4) Zend VM: executes the opcode array. OPcache intercepts after step 3: serialises compiled opcodes to shared memory — subsequent requests fetch opcodes directly, skipping steps 1-3. JIT (PHP 8.0+): OPcache's JIT further compiles hot opcodes to native x86/ARM machine code.

Common Misconception

Why It Matters

Common Mistakes

• Assuming PHP re-lexes and re-parses files on every request without OPcache
• Expecting syntax errors to appear lazily at the line they are on — errors appear at compile time
• Disabling OPcache in production — the single biggest PHP performance regression possible
• Thinking OPcache opcodes are the same as JIT-compiled native code — they are different stages

Code Examples

// Expecting lazy evaluation:
if (false) {
    $x = 1 +; // Syntax error on this line
}
// Assumption: no error — dead branch
// Reality: PHP compiles the entire file first
// Fatal error: parse error, unexpected ';' — before any code executes

// OPcache eliminates recompilation:
// Request 1 (cold): lex → parse → compile → store in shared memory
// Request 2 (warm): fetch opcodes from shared memory → execute directly
// Speedup: skip ~90% of compilation work per request

// Verify OPcache is working:
$status = opcache_get_status();
// Healthy production: hit rate > 99%
// opcache_hit_rate = hits / (hits + misses)

References

• ↗ https://www.php.net/manual/en/internals2.opcodes.php

Tags

Related Terms