Benchmarks.md

Benchmarks

Below is a comparison between the default Python MIME parser and flanker.mime in Time, Memory, and Function calls as well as other benchmarks relating to Flanker performance.

Test Specification:

• 2x Intel Core i7-3520M CPU @ 2.90GHz
• 2 GB RAM
• Results were averaged over 50 runs

Time

Full Message:

File Size	email.parser (seconds)	flanker.mime (seconds)	Speedup
11 MB	0.601	0.073	8x
1.8 MB	0.113	0.011	10x
16 KB	0.001	0.003	0.33x

Headers Only:

File Size	email.parser (seconds)	flanker.mime (seconds)	Speedup
11 MB	0.601	0.029	20x
1.8 MB	0.113	0.009	12x
16 KB	0.001	0.002	0.5x

Peak Memory Usage (including interpreter)

Full Message:

File Size	email.parser (MB)	flanker.mime (MB)	Memory Difference
11 MB	59.51	46.27	0.77x
1.8 MB	30.55	28.52	0.93x
16 KB	28.45	28.51	0x

Headers Only:

File Size	email.parser (MB)	flanker.mime (MB)	Memory Difference
11 MB	59.51	28.07	0.47x
1.8 MB	30.55	28.52	0.93x
16 KB	28.45	28.52	0x

Function Calls

Full Message:

File Size	email.parser (#)	flanker.mime (#)	Speedup
11 MB	914,741	2,011	454x
1.8 MB	195,062	11,046	17x
16 KB	3,216	3,5842	-1.1x

Headers Only:

File Size	email.parser (#)	flanker.mime (#)	Speedup
11 MB	914,741	1,253	730x
1.8 MB	195,062	9,741	20x
16 KB	3,216	2,476	1.2x

Pinned Regex

Even though master does not use a pinned version of regex, we strongly recommend due to the 6x speedup compared to the latest version on PyPi (as of 4/15/14). It may not be an issue for someone that handles many small files, but is an issue for large files.

File Size	regex.1.20110315 (seconds)	regex 2014.04.10 (seconds)	Speedup
11 MB	0.0720	0.4652	6x