PureBytes Links
Trading Reference Links
|
Stack overflow or floating point overflow?
Floating point overflow can be caused when a large numerator is divided by a VERY small denominator, which does not have to be zero. The common test for zero (in the example below) ...
if DENOM <> 0 then value1 = NUMER/DENOM
is not as good as the following test ....
if absvalue(DENOM) > 0.0000001 then value1 = NUMER/DENOM
OTOH, stack overflow can occur when the number of embedded function calls exceeds the stack's ability to keep track of it all. This can occur when a function is being called recursively too many times, or too many functions are calling too many others.
Either way, even a CRAY computer can get both types of overflow if the user's program is poorly written.
- Mark Jurik
----------
From: phil
Sent: Tuesday, April 16, 2002 11:28 AM
To: omega-list@xxxxxxxxxx
Subject: No clue
Hi List,
TS products are a piece of ......... Today I got two different floating
point errors that knocked me out both times. The first was a stack overflow
error and the second was a stack under error. What gives?? Any suggestion on
how to prevent this in the future.
Omega ought to have to pay for any losses suffered while in a trade and then
not being able to execute an order due to "downtime".
Phil
|