On Wed, 7 Mar 2001, Chris Gilbert wrote:

> I suppose you could claim efficiency for other cpu's by having an arm7
> specific handler in a kernel for multiple cpu versions.  but it's not needed,
> and would just add unneeded bloat to the kernel.

	Bloat in the GENERIC case, but better performance in the
	tuned kernel case :)

	IIRC cgd made a suggestion on the mips list of having some
	of the per CPU trap or cache functions be a padded block
	which initially just calls through a function pointer, but
	during initialisation the appropriate routine would be
	copied into the padded area avoiding the function call
	overhead.

		David/absolute		-- www.netbsd.org: No hype required --