On 13 Nov, Peter Seebach wrote:
>>I wonder what would happen if you had a program that legitimately had
>>some data that just happened to match the pattern.
> 
> My assumption would be that the system would be smart enough to only
> complain about executing blocks of code, but I'm not sure.

Well, the pattern is -407680 as an IEEE float, which is not an
"unreasonable" number.  Constants used as immediate operands live in
the .text segment, don't they?

I saw a brief explanation of the bug on comp.arch (by Chris Torek?)
suggesting that it could be worked around by locking one of the trap
handlers into the first level cache.  If that's the case, then you
would get _some_ performance penalty, but perhaps not as much as
turning it off altogether.

-- 
Andrew

"The steady state of disks is full."
				-- Ken Thompson