Re: [ tree-ssa] PRE problem

[law at redhat dot com]

In message <BBFA38DD-63CD-11D8-ABAD-000A95DA505C@dberlin.org>, Daniel Berlin wr
ites:
 >
 >On Feb 20, 2004, at 12:50 PM, Daniel Berlin wrote:
 >
 >>> Seems reasonable.
 >>>
 >>> The problem is we do this substitution (reverse copy propagation) 
 >>> *without*
 >>> fixing up the annotation for method_signature.3.  Specifically, if we 
 >>> do this
 >>> kind of substitution we need to fix is_dereferenced_{load,store} and
 >>> type_mem_tag in method_signature.3's annotation.
 >>>
 >>
 >> This is not copy propagation code.
 >> It's verifying the safety of a rename.
 >> We don't actually use the generated expressions except to check their 
 >> versions in the part you are quoting.
 >
 >We do, however, insert these expressions later if their safety has been 
 >verified
And thus you have to make sure the annotations are updated.  Otherwise
you can get incorrect code, segfaults, etc.  It is absolutely critical
that the annotations be updated.

 >And if we are generating and inserting an illegal expression , please 
 >provide me a preprocessed testcase and i'll be happy to make sure the 
 >rename algorithm doesn't mark it as safe.
For the original testcase I was working with the insertion would be
valid if the annotations are properly updated.

Jeff