Perf: typed int **104 / pow(n,104) chained smul omit llvm.pow.f64 (#36386)

[PurHur]

Summary

• Lower compile-time integer ** 104 / pow($n, 104) via (hundredthird×n) chained smul.with.overflow in JitPowIntegerEmitHigh (omit llvm.pow.f64).
• Even exponent: (-1)**104 === 1. Append-only (no new spine TU).
• Part of Performance: v2.0 targets and the codegen roadmap to reach them — typed code ≥ 3x Zend, untyped ≤ 1.5x slower, web request ≥ php-fpm: inline caches for property/method access, escape analysis for arrays/strings, guarded speculation, per-function O1 by default (lib/JIT.php, lib/JIT/Builtin/Type/Object_.php, lib/JIT/Analyzer.php) #36386 (lane C; after Perf: typed int **103 / pow(n,103) chained smul omit llvm.pow.f64 (#36386) #37691 landed **103; closed duplicate Perf: typed int **103 / pow(n,103) chained smul omit llvm.pow.f64 (#36386) #37692).

php-src: Zend/zend_operators.c pow_function / zend_pow / mul_function; ext/standard/math.c PHP_FUNCTION(pow).

Test plan

• ./script/phpunit.sh --filter NoThrowPowHundredfourthMulAotTest → 4/4 (23 assertions)
• script/differential-sweep.sh --aot --repeat 3 (single case) → 1/1 match Zend
• ./script/aot-smoke.sh → 9/9 + size gate OK

Made with Cursor