Let's iterate over smaller number to pass the test on slow computers

Slightly more robust parsing code

Adjusting the versions so the build succeeds with retrolambda 2.1.0

Tests for interfaces with default methods implemented by lamda functions

Switching to most recent version of retrolamda: 2.1.0

Depend on jbox2d-osgi module

Allow access to rt.jar classes

Speeding up array access

Adding JBox2D sample into our benchmarks

[maven-release-plugin] prepare for next development iteration

[maven-release-plugin] copy for tag release-0.17

[maven-release-plugin] prepare release release-0.17

Encapsulating access to lower 32-bits of a long number

Speeding up sieve with long numbers by directly resolving high32

Direct access to compare64 as well

neg64 is statically resolved now

Turning shifts into direct operations

Direct reimplementation of mod64, and64, or64, xor64 functions

Providing direct resolution for add64, sub64, mul64, div64 functions

Introducing Number.prototype.__bit64 with all 64-bit operations

Providing accessors for 64-bit arithmetic operations

Restrict the result of integer negation to 32-bits

Access the static fields via the access method

Extracting direct references to 64-bit operations

Each launcher should be used only once

JavaScript in a browser can be at most three times slower than HotSpot when computing five or ten thousands of prime numbers

The changes to number arithmetic deserve new magic header

Speeding up sieve by directly resolving all 32-bit number operations

Some operations are faster when included in the generated code rather than dispatched to Number.prototype

Resolving field access during compilation time