Re: Your June 7 change to expand_expr

[kenner at vlsi1 dot ultra dot nyu dot edu (Richard Kenner)]

    One option is to make MEM_IN_STRUCT_P ternary (yes: definitely in
    struct, no: definitely not in struct, and don't know).  Then, as
    Kenner points out would be desirable, we would have a conservative
    setting, and yet we could still do the fixed/varying struct/not-struct
    optimization.

That's an interesting idea, but I'm just not sure it's worth the
trouble.  How much is really picked up by doing the struct/non-struct
case?  I'd suspect it's small given "real aliasing".

Why not try an experiment: turn off the code that uses the bit and do
some benchmarks?  If the effect is small, then just rip out this
ancient code.

    Jeff, in the short term, perhaps we should disable the alias checks
    that depend on MEM_IN_STRUCT_P.  I will submit appropriate patches, if
    you think this is the right thing to do.  Unfortunately, Kenner, these
    patches won't be appropriate for GCC2 because they will all be in
    alias.c, which is not yet in GCC2.

In GCC2, there's also a reference in cse.c, though I don't know if
that still applies in EGCS.  The patch to disable it is small, of
course.