New submission from Stefan M. <pypy@stefan-marr.de>: I observed reproducibly on PyPy 2.2.1 that the Richards benchmark (https://raw.githubusercontent.com/smarr/Classic- Benchmarks/2e7f912f9da1c50c4a83cd4b09a1ba105e154beb/richards.py) becomes slower with every iteration (after 3 warmup iterations). It does not seem to reach a stable state. I executed the benchmark with: `pypy richards.py 1000 0 100` (that's 1000 runs, 0 suppressed warmup runs, and 100 inner iterations of the benchmark) The OS is a Ubuntu 13.10, was all updates. ---------- messages: 6815 nosy: pypy-issue, smarr priority: bug release: 2.2 status: unread title: PyPy getting gradually slower on Richards Benchmark with every iteration ________________________________________ PyPy bug tracker <tracker@bugs.pypy.org> <https://bugs.pypy.org/issue1747> ________________________________________