A new Fibonacci benchmark demonstrates the improvement by using both
addition and subtraction in each iteration of the loop, like #200.
Before:
test fib2_100 ... bench: 4,558 ns/iter (+/- 3,357)
test fib2_1000 ... bench: 62,575 ns/iter (+/- 5,200)
test fib2_10000 ... bench: 2,898,425 ns/iter (+/- 207,973)
After:
test fib2_100 ... bench: 1,973 ns/iter (+/- 102)
test fib2_1000 ... bench: 41,203 ns/iter (+/- 947)
test fib2_10000 ... bench: 2,544,272 ns/iter (+/- 45,183)
This splits the main arithmetic loops from the final carry/borrow
propagation, and also normalizes the slice lengths before iteration.
This lets the optimizer generate better code for these functions.
If a benchmark takes very long to run, it's harder to iterate on changes
to see their effect. Even reduced to 100, this pow_bench takes around 8
seconds on my machine, and still shows meaningful optimization effects.
Now, arithmetic binary operator traits take operands by value,
but non-copyable types such as `BigUint` should not always moved out when
applying operators.
This commit implements these operators not only for bare structs also
these references.
By-value implementations are forwarded to by-reference implementations
for now. In the future, by-value implementations may be replaced with
more efficient implementations (for example, the implementation that
reuses moved buffers.)
mathematcs.
Import numeric traits and functions that were removed from the standard library.
This also implements `AdditiveIterator` and `MultiplicativeIterator` for
`BigUint` and `BigInt`.
closes#28
Josh Stone
str4d
Nicolas Kirchner
David Hewson
Kent Overstreet
Alex Crichton
kennytm
Darin Morrison
Joseph Crail
Sven Nilsen
gifnksm
Aaron Turon