Benchmarking vim's utf8 implementation

Tries to get an idea for how good vim's utf8 implementation is and if it can be improved by swapping it out with an alternative implementation.

To run:

# fill in the compilers you want to test with
$ edit bench.sh
# run the benchmark
$ ./bench.sh

Implementations

• vim's
• Björn Höhrmann's versions
• Taylor R. Campbell's versions

Results

I have an early 2011 Macbook Pro with a 2GHz core i7 chip (and 16GB RAM).

Clang 3.4

file	function	runtime	bytes/codepoint	mb/sec
text/utf8.txt	vim_count	22.3 ms	2.3	281.3
text/utf8.txt	campbell_branch_count	26.3 ms	2.3	238.8
text/utf8.txt	bjorn_regular_count	29.6 ms	2.3	211.6
text/utf8.txt	bjorn_reduced_count	29.9 ms	2.3	210.1
text/utf8.txt	campbell_arith_count	42.1 ms	2.3	149.1

file	function	runtime	bytes/codepoint	mb/sec
text/ascii.txt	campbell_branch_count	100.1 ms	1.0	439.0
text/ascii.txt	bjorn_reduced_count	184.5 ms	1.0	238.3
text/ascii.txt	bjorn_regular_count	192.4 ms	1.0	228.5
text/ascii.txt	vim_count	243.2 ms	1.0	180.8
text/ascii.txt	campbell_arith_count	268.0 ms	1.0	164.0

file	function	runtime	bytes/codepoint	mb/sec
text/enwik8.txt	campbell_branch_count	222.0 ms	1.0	429.5
text/enwik8.txt	bjorn_regular_count	404.5 ms	1.0	235.7
text/enwik8.txt	bjorn_reduced_count	411.3 ms	1.0	231.9
text/enwik8.txt	vim_count	524.9 ms	1.0	181.7
text/enwik8.txt	campbell_arith_count	592.2 ms	1.0	161.1

Gcc 4.9

file	function	runtime	bytes/codepoint	mb/sec
text/utf8.txt	bjorn_regular_count	20.4 ms	2.3	307.2
text/utf8.txt	bjorn_reduced_count	22.8 ms	2.3	275.3
text/utf8.txt	vim_count	27.0 ms	2.3	232.3
text/utf8.txt	campbell_branch_count	28.9 ms	2.3	217.4
text/utf8.txt	campbell_arith_count	29.8 ms	2.3	210.5

file	function	runtime	bytes/codepoint	mb/sec
text/ascii.txt	campbell_branch_count	106.8 ms	1.0	411.6
text/ascii.txt	bjorn_regular_count	142.5 ms	1.0	308.5
text/ascii.txt	bjorn_reduced_count	155.8 ms	1.0	282.1
text/ascii.txt	campbell_arith_count	215.0 ms	1.0	204.5
text/ascii.txt	vim_count	330.1 ms	1.0	133.2

file	function	runtime	bytes/codepoint	mb/sec
text/enwik8.txt	campbell_branch_count	245.4 ms	1.0	388.7
text/enwik8.txt	bjorn_regular_count	290.0 ms	1.0	328.9
text/enwik8.txt	bjorn_reduced_count	324.7 ms	1.0	293.7
text/enwik8.txt	campbell_arith_count	469.5 ms	1.0	203.1
text/enwik8.txt	vim_count	733.4 ms	1.0	130.0