`claripy.Extract` uses exponentation instead of a binary shift

[LukeSerne]

While profiling some code that used some rather large claripy.BVVs, I found that a lot of time was spent in claripy.Extract. I found that the main culprit is that the use of exponentiation to construct a bitmask instead of a shift. I used the below script to check that changing this exponentiation by a shift can result in a big increase in performance when the parameter values to claripy.Extract are large.

Current code

claripy/claripy/bv.py

Lines 290 to 291 in d8eeba6

	def Extract(f, t, o):
	return BVV((o.value >> t) & (2**(f+1) - 1), f-t+1)

Proposed change

def Extract(f, t, o):
    return BVV((o.value >> t) & ((1 << (f + 1)) - 1), f + 1 - t)

Test

main.py:

import claripy

def main():
    data = claripy.BVV(0, 25165824)
    for _ in range(100):
        claripy.Extract(25165823, 25133056, data)

if __name__ == "__main__":
    main()

With the current implementation, running this script takes 7.377 seconds using the command time python main.py. With the proposed improvement, running the same command takes only 0.480 seconds.

[ltfish]

Awesome! Thank you. Do you want to send a PR?