1994-07-11 - Re: Bit counting

Header Data

From: gtoal@an-teallach.com (Graham Toal)
To: cypherpunks@toad.com
Message Hash: 9268deb70afe293d8cfaa63472b2fbdb0a373473058d2d683ea62ccfc78c9369
Message ID: <199407111228.NAA21528@an-teallach.com>
Reply To: N/A
UTC Datetime: 1994-07-11 12:29:23 UTC
Raw Date: Mon, 11 Jul 94 05:29:23 PDT

Raw message

From: gtoal@an-teallach.com (Graham Toal)
Date: Mon, 11 Jul 94 05:29:23 PDT
To: cypherpunks@toad.com
Subject: Re: Bit counting
Message-ID: <199407111228.NAA21528@an-teallach.com>
MIME-Version: 1.0
Content-Type: text/plain


	Again, if its speed you want, you can't beat look up tables no matter how
	hard you try.

Ray, you've missed the point of some of the explanations; VERY FAST cpu's
as unbelievably fast as long as they are executing *on-chip* - as soon as
they have to go to RAM for a table lookup, they suffer a performance hit
equivalent to executing large amounts of in-line instructions - one array
lookup might be worth 200 straight opcodes.  Hence why an in-line scheme
has to be found for critical inner-loop stuff.  The precise details vary
according to the pipelining and the amount of on-cpu instruction cache.

G
PS We'd seen the Escrow Officer Trading Card skit on t.p.c when it first
was posted a couple of months ago ;-)





Thread