[Bug optimization/13875] [tree-ssa] missed jump thread optimization on the tree-level

[law at redhat dot com]

------- Additional Comments From law at redhat dot com  2004-03-03 18:26 -------
Subject: Re:  [tree-ssa] missed jump thread 
 optimization on the tree-level

In message <20040303181405.32100.qmail@sources.redhat.com>, "dann at godzilla d
ot ics dot uci dot edu" writes:
 >
 >------- Additional Comments From dann at godzilla dot ics dot uci dot edu  20
 >04-03-03 18:14 -------
 >As per Jeff's request, a brief look at generate-3.4.ii.t54.vars function:
 >
 >void MODEL_GENERATOR::assumePTLiteral(bool&, bool, const GLITERAL&, GINTERPRE
 >T&)
 >[snip]
 >
 >  T.8294 = (bool)(int)ptl->neg;
 >  if (T.8294 == 0) goto <L18>; else goto <L17>;
 >
 ><L17>:;
 >  if (reallyPT) goto <L21>; else goto <L18>;
 >
 ><L18>:;
 >  T.8297 = !T.8294;
 >  if (T.8297 == 0) goto <L33>; else goto <L20>;
 >
 ><L20>:;
 >  if (reallyPT == 0) goto <L21>; else goto <L33>;
 >
 ><L21>:;
 >  if (TraceLevel > 1) goto <L22>; else goto <L31>;
 >
 >see some jump threading opportunities missed, the  T.8297 variable
 >could be eliminated.
Can you send me the .cddce dump.  If T.8297_XXXX is used more than once,
then that's going to block most/all jump threading opportunities in this
code fragment.

 >
 >Other observations: 
 > --   casts like "if ((bool)(int)(bool) ....)"
>From a .vars dump?  I'm not terribly surprised.  Not particularly important.

It's actually easier to look at the cddce dumps since you've still got SSA
version information which makes it much clearer where/how values are
being used.

jeff




-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13875