Bitwise math for speed

[gammazero]

No description provided.

[gammazero]

Without modular math

> go test -bench .
PASS
BenchmarkQueueSerial-8      10000000           317 ns/op
BenchmarkQueueGet-8         100000000           13.8 ns/op
BenchmarkQueueTickTock-8    50000000            29.6 ns/op

With modular math

> go test -bench .
PASS
BenchmarkQueueSerial-8      10000000           311 ns/op
BenchmarkQueueGet-8         300000000            5.05 ns/op
BenchmarkQueueTickTock-8    100000000           12.9 ns/op

Results were consistent across a number of runs for both.

While the optimizer can optimize multiplication by powers of 2, it probably cannot optimize to do bitwise modulus because it does not know apriori if it is operating on numbers that are powers of 2.

[eapache]

Thanks!