GCC trunk SPEC2000 performance

[law@redhat.com]

In message <200206202001.QAA26566@makai.watson.ibm.com>, David Edelsohn writes:
 > >>>>> law  writes:
 > 
 > Jeff> Unknown as Jan sent a patch without a description of the underlying pr
 > oblem.
 > 
 > 	Please re-read the discussion if you missed the purpose of the
 > patch. 
I didn't recall seeing any discussion of a technical nature.  It's certainly
possible I missed it though.

 > 
 > apsi.f: In subroutine `uvset':
 > apsi.f:4327: internal compiler error: RTL flag check: REG_POINTER used with 
 > unexpected rtx code `mem' in copy_rtx_and_substitute, at integrate.c:2065
 > 
 > 	Jan's patch ensures that REG_POINTER() only is called with a REG
 > rtx.  If the rtx is not a REG, the rest of the conditional does not make
 > sense, so it seems like a reasonable fix.
But how/why did we have a non-REG thingie in the register replacement array?
That's key to knowing if Jan's patch is correct or just papering over a problem
elsewhere.

jeff