Re: ATA identify retries exceeded - kernel panic

["Thomas E. Spanjaard" <tgen@xxxxxxxxxxxxx>]

Matthew Dillon wrote:
> +		if (ocflags & M_ZERO) {
> +			if (oc->simple_objsize)
> +				bzero(obj, oc->simple_objsize);
> +			else
> +				panic("objcache_get(): M_ZERO illegal here");
> +		}

After some discussion with Jeffrey Hsu and Simon Schubert, it was 
concluded that this incurs a penalty on the hot path of objcache_get(), 
which should be avoided. Best to have the consumers handle any zero'ing, 
in order to not nerf the consumers who rely on the speed. In that case, 
it pollutes the instruction cache, and could worsen performance 
significantly in the case of branch misprediction. Especially on the 
long pipelines CPUs have these days (though with the Core 2 family, 
Intel has cut back on the 30+ stages of the Prescott/Presler...).

Cheers,
--
        Thomas E. Spanjaard
        tgen@netphreax.net

Attachment: signature.asc
Description: OpenPGP digital signature