Rebalance benchmarks

[phischu]

• [ ] Prevent iterate_increment from being constant-folded
• [ ] Prevent factorial_accumulator from being constant-folded
• [ ] Change input sizes to more evenly distribute time
• [ ] Make input sizes round numbers
• [x] Add generator
• [x] Add handler sieve
• [ ] Add checkpointing

[phischu]

@mattisboeckle it would be great if you could figure out why iterate_increment and factorial_accumulator are constant-folded (in the llvm backend), and how we can cheaply prevent this (perhaps by using different arithmetic operations involving a magical modulus?)

[mattisboeckle]

@mattisboeckle it would be great if you could figure out why iterate_increment and factorial_accumulator are constant-folded (in the llvm backend), and how we can cheaply prevent this (perhaps by using different arithmetic operations involving a magical modulus?)

I was able to prevent constant folding in the mentioned programs by doing a modulo operation on the iteration parameter, which llvm does not seem to be able to analyze.

[mattisboeckle]

I've adjusted the magic mod value to 2^63-1 and added a comment explaining, why we do this.