Re: [PATCH v3 2/5] benchtests: Add memset zero fill benchtest

[Wilco Dijkstra via Libc-alpha <libc-alpha@sourceware.org>]

Hi Lucas,

> Sorry but I suppose don't understood your suggestion completely.  The
> memset_value array will hold patterns like [0,0], [0,1] or [1,1],
> right?  If so, this will not work to measure the zero-to-one pattern for
> example, as it will be mixing zero-to-one with one-to-zero calls. In
> order to measure just an specific patter the buffer must be loaded
> previously of the timing loop.

The original idea was to add more tests for memset of zero and check
whether writing zero is optimized and/or writing zero over zero. There is
an equal number of 0->1 and 1->0 transitions in a pattern, so you can't
easily differentiate between them, but you can tell whether they are the
same or faster than 1->1 transitions.

For 0->0 you can run different patterns with a varying number of transitions
but the same number of zeroes and ones: eg. 0000000011111111 (7 times 0->0)
vs 0011001100110011 (4 times 0->0) vs 0101010101010101 (no 0->0).

Cheers,
Wilco