inter-class optimizations

Thu Feb 19 10:21:00 GMT 2004

Ranjit Mathew writes:
 > Per Bothner wrote:
 > > Also, -fomit-frame-pointer is now the default (at least on x86).
 > 
 > It's not been committed yet, has it?
Even when this patch is committed, it won't be the default for gcj
until we've converted to using unwinder data for stack traces.
 > This thread:
 > 
 > http://gcc.gnu.org/ml/gcc-patches/2004-02/threads.html#01085
 > 
 > seems to conclude that we put in -fno-omit-frame-pointer
 > anyways for libjava since we depend on backtrace( ), which
 > in older libc-s (and on MinGW, BTW) uses the frame-pointer
 > to walk the call frames.
That's right. And whatever is the default for gcc as a whole, we can
continue to relay on stack frames with a frame pointer until we're
ready to switch.
Andrew.