Integer/long swaps: shaving off a few cycles

dell_jockey · Post by **dell_jockey** » Mon Jan 28, 2008 1:57 pm

Deleted, see posts below....

Trond · Post by **Trond** » Mon Jan 28, 2008 2:06 pm

1. Don't do speed testing with the debugger on.
2. A normal "Swap a, b" is twice as fast as your XOR swap.

dell_jockey · Post by **dell_jockey** » Mon Jan 28, 2008 2:15 pm

Trond wrote:1. Don't do speed testing with the debugger on.
2. A normal "Swap a, b" is twice as fast as your XOR swap.

1: not even to compare the relative merit of two methods? I know the durations measured will not be correct by a large margin, but surely that will be true for both methods tested?

2: I stand corrected here, I didn't even know PB had a swap built-in

Thanks for pointing this out! I had downloaded a bit of code that had a swap macro included and assumed that PB doesn't have one. This set off my thought train. Should have checked the manual first.... So, without much ado, I deleted the post above....

Trond · Post by **Trond** » Mon Jan 28, 2008 2:18 pm

dell_jockey wrote:
Trond wrote:1. Don't do speed testing with the debugger on.
2. A normal "Swap a, b" is twice as fast as your XOR swap.
1: not even to compare the relative merit of two methods?

No. Because native commands aren't slowed down by the same magnitude as your own with the debugger on.