Performance

Performance of Boost.Coroutine was measured on the platforms shown in the following table. Performance measurements were taken using rdtsc and boost::chrono::high_resolution_clock, with overhead corrections, on x86 platforms. In each case, cache warm-up was accounted for, and the one running thread was pinned to a single CPU.

Table 1.1. Performance of asymmetric coroutines

Platform	switch	construction (protected stack-allocator)	construction (preallocated stack-allocator)	construction (standard stack-allocator)
i386 (AMD Athlon 64 DualCore 4400+, Linux 32bit)	49 ns / 50 cycles	51 µs / 51407 cycles	14 µs / 15231 cycles	14 µs / 15216 cycles
x86_64 (Intel Core2 Q6700, Linux 64bit)	12 ns / 39 cycles	16 µs / 41802 cycles	6 µs / 10350 cycles	6 µs / 18817 cycles

Table 1.2. Performance of symmetric coroutines

Platform	switch	construction (protected stack-allocator)	construction (preallocated stack-allocator)	construction (standard stack-allocator)
i386 (AMD Athlon 64 DualCore 4400+, Linux 32bit)	47 ns / 49 cycles	27 µs / 28002 cycles	98 ns / 116 cycles	319 ns / 328 cycles
x86_64 (Intel Core2 Q6700, Linux 64bit)	10 ns / 33 cycles	10 µs / 22828 cycles	42 ns / 710 cycles	135 ns / 362 cycles

Boost C++ Libraries

Performance