Nightly Builds: Python Environment Setup
Our benchmarks are executed on dual socket platforms hosting 2 Broadwell CPUs (E5-2699-v4) with 22 cores/55MB cache running @ 2.2GHz.
Other configuration details:
- CPU hyper-threading disabled from BIOS for reducing run to run variations
- fixed CPU frequency for reducing run to run variations: 2.2GHz (Turbo disabled from BIOS and CPU frequency fixed at 2.2GHz at OS level)
- 800GB Intel SSD DC S3510
- 8x16GB DDR4 2133MHz (all 4 memory channels populated on both sockets)
Linux OS: Ubuntu Server 16.04.2 LTS,
Kernel version: 4.4.0-62-generic x86_64 GNU/Linux
We build cpython 2 and 3 using gcc 5.4.0 with both default parameters and pgo:
./configure --prefix=/the/build/folder make -j 44
./configure --prefix=/the/build/folder make profile-opt -j 44
We are using The Python Benchmark Suite (v 0.5.4) python/performance, an opensource benchmark suite from which we run the following workload groups:
- apps: "High-level" applicative benchmarks (2to3, Chameleon, Tornado HTTP)
- calls: Microbenchmarks on function and method calls
- math: Float and integers
- regex: Collection of regular expression benchmarks
- serialize: Benchmarks on pickle and json modules
- startup: Collection of microbenchmarks focused on Python interpreter start-up time.
- template: Templating libraries
The run option used is -r ( --rigorous Spend longer running tests to get more accurate results ) The average value and relative standard deviation (standard deviation / average) are computed.
Intel technologies' features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system configuration.