Re: X much slower in 2.0.24 than in 1.2.13

Keith Rohrer (kwrohrer@uiuc.edu)
Wed, 06 Nov 1996 16:36:37 -0600

Ray Van Tassle-CRV004 wrote:
> Keith wrote:
> > Has anyone tried
> > optimizing the ordering for cache friendliness, or is it the way it is
> > for some other reason?
> I played around with struct buffer_head (in fs.h), running heavy I/O
> benchmark programs, and changing around the order of the members. Somewhere
> around 1.2.86, another entry got added neat the top, completely destroying
> sct's careful ordering of frequently-accessed entries.
> My results: (TA-DA!!!) *None* *NADA*, absolutely no change in the overall
> elapsed time, nor the system time, no matter how I shifted things around.
> Not even when I tried to un-optimize things, and cross cache-line boundaries.
What, besides, the struct sock things, was stressing the cache while you
were testing? Did you try X benchmarks, which were the way the loss was
originally detected?

Keith

-- 
The priests and the friars/Behold me in dread
Because I still love you,/My love, and you're dead.
    ---Dead Can Dance, "I Am Stretched On Your Grave", based on King/S.
	O'Connor's rewrite of "The Unquiet Grave", trad. Irish folk.