KB> I make an educated guess when determining the stack size 
KB> for a new thread.  Does anyone have any suggestion as to a 
KB> more accurate method?  That is, are there any tools that 
KB> will analyze a thread at runtime to determine its maximum 
KB> use?  For example, would the profiler do something like 
KB> this?
KB> 
KB> I realize that the stack usage may differ from run to run 
KB> based on the conditions at runtime, but this would 
KB> nevertheless be useful.

The only place I think this matters is for thread 1, where the OS will
fully commit all pages of the allocated stack space.  For threads you
create, it shouldn't be a big deal, since the pages aren't committed
until a guard page exception causes the stack to grow downward.  So
using a large value for those threads should be fine.  Also, I believe
that PM's stack requirements are documented somewhere, and aren't that
big.

John
--- Maximus/2 2.01wb