|
|
@@ -1,7 +1,7 @@
|
|
|
{
|
|
|
Common subexpression elimination on base blocks
|
|
|
|
|
|
- Copyright (c) 2005 by Florian Klaempfl
|
|
|
+ Copyright (c) 2005-2012 by Florian Klaempfl
|
|
|
|
|
|
This program is free software; you can redistribute it and/or modify
|
|
|
it under the terms of the GNU General Public License as published by
|
|
|
@@ -34,13 +34,9 @@ unit optcse;
|
|
|
{
|
|
|
the function creates non optimal code so far:
|
|
|
- call para nodes are cse barriers because they can be reordered and thus the
|
|
|
- temp. creation can be done too late
|
|
|
- - cse's in chained expressions are not recognized: the common subexpression
|
|
|
- in (a1 and b and c) vs. (a2 and b and c) is not recognized because there is no common
|
|
|
- subtree b and c
|
|
|
+ temp. creation could be done too late
|
|
|
- the cse knows nothing about register pressure. In case of high register pressure, cse might
|
|
|
have a negative impact
|
|
|
- - assignment nodes are currently cse borders: things like a[i,j]:=a[i,j]+1; are not improved
|
|
|
- the list of cseinvariant node types and inline numbers is not complete yet
|
|
|
|
|
|
Further, it could be done probably in a faster way though the complexity can't probably not reduced
|
|
|
@@ -123,7 +119,6 @@ unit optcse;
|
|
|
result:=fen_norecurse_false;
|
|
|
exit;
|
|
|
end;
|
|
|
- { so far, we can handle only nodes being read }
|
|
|
if
|
|
|
{ node possible to add? }
|
|
|
assigned(n.resultdef) and
|
florian