Perf: printf/sprintf %b of floats and overflow longs match Zend (#36386)

[PurHur]

Summary

• 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; Focus: performance / typed fast paths). Overflow: CI: php-llvm-structgep-assert.patch fails to apply — blocks north-star5-verify-fast and all bootstrap gates #36143–VM: 1M-iteration loops exit 255 silently — restore compute-suite headroom on bin/vm.php #36148 / RunForge north star (Sep 2026): trust, build speed, performance — automation tracker #36142 children closed.
• printf/sprintf %b of overflow-promoted ints and large floats printed 0 while %d/%o/%x/echo matched Zend — leftover of Perf: lazy f64 overflow slot for native-long ±/× (#36386) #37051 lazy ±/× materialize and Perf: printf/sprintf %d of floats and overflow longs match Zend (#36386) #37075 (which fixed %d/%u/%x but left %b on the non-integer snprintf arm).
• Treat %b as an integer conversion (extractSnprintfLongArg + %llb promote), teach SprintfSnprintfRuntime the same, and materialize overflowable native-long argv for non-const formats via __compiler_sprintf.

php-src: ext/standard/formatted_print.c %b (unsigned machine-word bit pattern) / zval_get_long; Zend/zend_operators.h zend_dval_to_lval.

Test plan

• ./script/phpunit.sh --filter 'SprintfPercentBAotTest|SprintfPercentDFloatAotTest|SprintfOverflowableNativeLongAotTest|EchoOverflowableNativeLongAotTest' → 18/18 OK (90 assertions)
• ./script/aot-smoke.sh → 9/9 + size gate OK
• make north-star5-verify-fast → OK (spine 8353/8353)
• Not run: full script/differential-sweep.sh --aot --repeat 3 (sprintf %b coercion only; peer of Perf: printf/sprintf %d of floats and overflow longs match Zend (#36386) #37075)

Closes nothing (Part of #36386).

Made with Cursor