New issue 1822: JIT compiles tons of bridges https://bitbucket.org/pypy/pypy/issue/1822/jit-compiles-tons-of-bridges
Remi Meier: code from http://rosettacode.org/wiki/Perlin_noise#Python see attached file Nightly pypy (pypy-c-jit-72476-dbdcc057562b-linux64) is slower with jit than with "--jit off". Also, current STM builds don't have this problem. **Look at number of bridges:** JIT summary of PyPy-STM: ``` 0# [46643b20651b8] {jit-summary 0# Tracing: 29 0.078599 0# Backend: 27 0.019512 0# TOTAL: 0.456509 0# ops: 50153 0# recorded ops: 11775 0# calls: 885 0# guards: 6755 0# opt ops: 2654 0# opt guards: 637 0# forcings: 96 0# abort: trace too long: 2 0# abort: compiling: 0 0# abort: vable escape: 0 0# abort: bad loop: 0 0# abort: force quasi-immut: 0 0# nvirtuals: 4165 0# nvholes: 614 0# nvreused: 2888 0# Total # of loops: 10 0# Total # of bridges: 18 0# Freed # of loops: 0 0# Freed # of bridges: 0 0# [46643b20f1765] jit-summary} ``` JIT summary of pypy nightly: ``` [466341fc5ad18] {jit-summary Tracing: 921 4.593343 Backend: 920 0.672270 TOTAL: 9.669084 ops: 6136322 recorded ops: 1571845 calls: 56491 guards: 196164 opt ops: 397249 opt guards: 57832 forcings: 942 abort: trace too long: 1 abort: compiling: 0 abort: vable escape: 0 abort: bad loop: 0 abort: force quasi-immut: 0 nvirtuals: 1278724 nvholes: 591642 nvreused: 556151 Total # of loops: 19 Total # of bridges: 902 Freed # of loops: 0 Freed # of bridges: 0 [466341fcecc7b] jit-summary} ``` _______________________________________________ pypy-issue mailing list [email protected] https://mail.python.org/mailman/listinfo/pypy-issue
