ref_count_test's Introduction

ref_count_test

An incomplete comparison of alternative reference counting mechanisms. Currently more of a rough guestimate of their performance than actual realistic comparison.

Results

Run on (16 X 3600 MHz CPU s) 2017-08-11 21:25:06 WARNING CPU scaling is enabled, the benchmark real time measurements may be noisy and will incur extra overhead.

Benchmark	Time	CPU	Iterations
BM_LocalIncrementAndDecrement/threads:16	0 ns	2 ns
BM_ThreadLocalIncrementAndDecrement/threads:16	0 ns	5 ns	148346736
BM_ThreadLocalArrayIncrementAndDecrement/threads:16	0 ns	3 ns	215744864
BM_UncontestedAtomicIncrementAndDecrement/threads:16	1 ns	14 ns	48595360
BM_ContestedAtomicIncrementAndDecrement/threads:16	27 ns	423 ns	1600000
BM_UnorderedMapInsertAndErase/threads:16	3 ns	51 ns	14268080
BM_ThreadLocalUnorderedMapInsertAndErase/threads:16	3 ns	51 ns	13661520

Benchmark	Time	CPU	Iterations
BM_LocalIncrementAndDecrement/threads:16	0 ns	2 ns	321130640
BM_ThreadLocalIncrementAndDecrement/threads:16	0 ns	5 ns	148380032
BM_ThreadLocalArrayIncrementAndDecrement/threads:16	0 ns	4 ns	192874208
BM_UncontestedAtomicIncrementAndDecrement/threads:16	1 ns	14 ns	48584400
BM_ContestedAtomicIncrementAndDecrement/threads:16	25 ns	383 ns	1600000
BM_UnorderedMapInsertAndErase/threads:16	4 ns	47 ns	14119104
BM_ThreadLocalUnorderedMapInsertAndErase/threads:16	3 ns	49 ns	14724080

Interesting Observations

BM_ThreadLocalIncrementAndDecrement is slower than BM_ThreadLocalArrayIncrementAndDecrement
BM_ContestedAtomicIncrementAndDecrement: the number of iterations is often exacly 1600000

Recommend Projects