On Thu, 11 Aug 2005, Qingqing Zhou wrote:
> It is said that the BufferBlockPointers is used to speedup the
> BufferGetBlock() macro. I compared three ways of getting block pointers.
> I.e., existing method (arrary method), calculating block pointer by adding
> base addr and offset*blockid method (mul method) and optimizing mul method
> by using bit shift (shift method). All of them calculate the block pointer
> 80000 times (i.e., the BufferBlockPointers array is of size 80000), and each
> take 3 rounds.
>
> The result is:
>
> SunOS/gcc 3.2
> duration round 1 of array method: 4.179 ms
> duration round 2 of array method: 4.160 ms
> duration round 3 of array method: 4.143 ms
> duration round 1 of mul method: 3.311 ms
> duration round 2 of mul method: 3.233 ms
> duration round 3 of mul method: 3.233 ms
> duration round 1 of shift method: 3.554 ms
> duration round 2 of shift method: 3.235 ms
> duration round 3 of shift method: 3.233 ms
>
> Linux/gcc 3.2
> duration round 1 of array method: 0.422 ms
> duration round 2 of array method: 0.324 ms
> duration round 3 of array method: 0.354 ms
> duration round 1 of mul method: 0.271 ms
> duration round 2 of mul method: 0.248 ms
> duration round 3 of mul method: 0.304 ms
> duration round 1 of shift method: 0.322 ms
> duration round 2 of shift method: 0.239 ms
> duration round 3 of shift method: 0.265 ms
>
> We can conclude that:
> (1) mul or shift are definitely better than array method;
> (2) mul and shift are comparable;
Do you have results for more recent gcc releases?
Thanks,
Gavin