cost of -fnon-call-exceptions

[Ian Lance Taylor]

Kenny Simpson <theonetruekenny@yahoo.com> writes:

> Is the only cost of using -fnon-call-exceptions some extra data in the
> .eh_frame section, does it impede the optimizer in some way?
> (i.e. should I expect the .text section to be identical)

It does impede the optimizer in various subtle ways.  It causes most
memory references to possibly generate exceptions.  That means that many
more basic blocks have exception edges.  That causes knock-on effects
for the optimization passes.

In particular, the scheduler avoids moving instructions which can throw
exceptions.

Ian