aot
/
compiler
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
TypePHP 编译器 https://swoole.com/aot/
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
1829 Commits
2 Branches
17 Tags
66 MiB
 
 
Tag: Branch: Tree: 136b08fe72
Branches Tags
${ item.name }
Create tag ${ searchTerm }
Create branch ${ searchTerm }
from '136b08fe72'
${ noResults }
compiler/benchmark/property-method-call
yavon007 a1efda0f7f
fix(compiler): preserve signatures for typed property method calls (#104) 		4 weeks ago
..
README.md fix(compiler): preserve signatures for typed property method calls (#104) 4 weeks ago
main.php fix(compiler): preserve signatures for typed property method calls (#104) 4 weeks ago
project.yml fix(compiler): preserve signatures for typed property method calls (#104) 4 weeks ago

README.md

Typed property method calls

This benchmark calls a final class method through a declared object property 1,000,000 times, returning checksum 7000000. It isolates receiver/property and method-call overhead; it is not a whole-application performance estimate.

Build from the repository root with a matching PHPX runtime:

php bin/tpc.php benchmark/property-method-call/project.yml --no-progress \
  --build-dir /tmp/typephp-property-method-build -o /tmp/property-method-call
/tmp/property-method-call

For a before/after comparison, use the same PHPX, compiler flags and workload in both checkouts, separate build directories, and alternate the two binaries after a warmup. Verify equal output before comparing elapsed time.

An initial Linux ARM64/PHP 8.5.10 ZTS/PHPX 4b3a472 O2 comparison against TypePHP a1782233 measured five-run median process times of 29.97 ms before and 22.63 ms after (24.5% lower elapsed time). Timings include process startup. The normal 22-command dungeon scenario improved only about 2.4%, and its exception-heavy 28-command scenario showed no reliable improvement. These results demonstrate a local call-path improvement, not a claim that all applications become 24.5% faster.