[llvm-commits] [llvm] r41167 - /llvm/trunk/include/llvm/Analysis/Dominators.h
Author: nicholas
Date: Sat Aug 18 09:46:55 2007
New Revision: 41167
URL: http://llvm.org/viewvc/llvm-project?rev=41167&view=rev
Log:
Simplify.
Modified:
llvm/trunk/include/llvm/Analysis/Dominators.h
Modified: llvm/trunk/include/llvm/Analysis/Dominators.h
URL:
http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Analysis/Dominators.h?rev=41167&r1=41166&r2=41167&view=diff
==
--- llvm/trunk/include/llvm/Analysis/Dominators.h (original)
+++ llvm/trunk/include/llvm/Analysis/Dominators.h Sat Aug 18 09:46:55 2007
@@ -379,8 +379,7 @@
void removeBlock(BasicBlock *BB) {
assert(find(BB) != end() && "Block is not in DominanceFrontier!");
for (iterator I = begin(), E = end(); I != E; ++I)
- if (I->second.count(BB))
-I->second.erase(BB);
+ I->second.erase(BB);
Frontiers.erase(BB);
}
___
llvm-commits mailing list
llvm-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
[llvm-commits] [llvm] r41168 - /llvm/trunk/lib/Transforms/Scalar/LICM.cpp
Author: nicholas Date: Sat Aug 18 10:08:56 2007 New Revision: 41168 URL: http://llvm.org/viewvc/llvm-project?rev=41168&view=rev Log: Typo. Modified: llvm/trunk/lib/Transforms/Scalar/LICM.cpp Modified: llvm/trunk/lib/Transforms/Scalar/LICM.cpp URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/LICM.cpp?rev=41168&r1=41167&r2=41168&view=diff == --- llvm/trunk/lib/Transforms/Scalar/LICM.cpp (original) +++ llvm/trunk/lib/Transforms/Scalar/LICM.cpp Sat Aug 18 10:08:56 2007 @@ -265,7 +265,7 @@ // // Traverse the body of the loop in depth first order on the dominator tree so // that we are guaranteed to see definitions before we see uses. This allows - // us to sink instructions in one pass, without iteration. AFter sinking + // us to sink instructions in one pass, without iteration. After sinking // instructions, we perform another pass to hoist them out of the loop. // SinkRegion(DT->getNode(L->getHeader())); ___ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
[llvm-commits] [llvm] r41170 - /llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp
Author: nicholas
Date: Sat Aug 18 18:21:28 2007
New Revision: 41170
URL: http://llvm.org/viewvc/llvm-project?rev=41170&view=rev
Log:
Oops, remove assert that wasn't meant to be committed.
Modified:
llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp
Modified: llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp
URL:
http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp?rev=41170&r1=41169&r2=41170&view=diff
==
--- llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp Sat Aug 18
18:21:28 2007
@@ -712,8 +712,6 @@
}
if (J != E && J->To == n) {
- assert(J->Subtree->dominates(Subtree));
-
edge.LV = static_cast(J->LV & R);
assert(validPredicate(edge.LV) && "Invalid union of lattice
values.");
___
llvm-commits mailing list
llvm-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
[llvm-commits] [llvm] r41169 - /llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp
Author: nicholas
Date: Sat Aug 18 18:18:03 2007
New Revision: 41169
URL: http://llvm.org/viewvc/llvm-project?rev=41169&view=rev
Log:
Never insert duplicate edges.
Modified:
llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp
Modified: llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp
URL:
http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp?rev=41169&r1=41168&r2=41169&view=diff
==
--- llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/PredicateSimplifier.cpp Sat Aug 18
18:18:03 2007
@@ -711,23 +711,27 @@
++J;
}
-if (J != E && J->To == n && J->Subtree->dominates(Subtree)) {
+if (J != E && J->To == n) {
+ assert(J->Subtree->dominates(Subtree));
+
edge.LV = static_cast(J->LV & R);
assert(validPredicate(edge.LV) && "Invalid union of lattice
values.");
- if (edge.LV != J->LV) {
-// We have to tighten any edge beneath our update.
-for (iterator K = I; K->To == n; --K) {
- if (K->Subtree->DominatedBy(Subtree)) {
-LatticeVal LV = static_cast(K->LV & edge.LV);
-assert(validPredicate(LV) && "Invalid union of lattice
values");
-K->LV = LV;
- }
- if (K == B) break;
+ if (edge.LV == J->LV)
+return; // This update adds nothing new.
+ }
+
+if (I != B) {
+ // We also have to tighten any edge beneath our update.
+ for (iterator K = I - 1; K->To == n; --K) {
+if (K->Subtree->DominatedBy(Subtree)) {
+ LatticeVal LV = static_cast(K->LV & edge.LV);
+ assert(validPredicate(LV) && "Invalid union of lattice values");
+ K->LV = LV;
}
-
+if (K == B) break;
}
-}
+ }
// Insert new edge at Subtree if it isn't already there.
if (I == E || I->To != n || Subtree != I->Subtree)
___
llvm-commits mailing list
llvm-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
Re: [llvm-commits] PR1146 patch for llvm-gcc-4.0
Hi Reid, > + bool extend = false; it looks like extend is never assigned another value. Ciao, Duncan. ___ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
Re: [llvm-commits] PR1146 patch for llvm-gcc-4.0
Hi Reid, > - Attrs.push_back(ParamAttrsWithIndex::get(ArgTypes.size(), Attributes)); > + Client.addAttributes(ArgTypes.size(), Attributes); all users of addAttributes seem to be adding a new attribute at the back. Given this, your implementation of addAttributes looks inefficient. Also, does it work correctly? - addAttributes doesn't add the attribute at all if there are no existing attributes for that index, which seems to be exactly the case for every use of it! Ciao, Duncan. ___ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
Re: [llvm-commits] PR1146 patch for llvm-gcc-4.0
Hi Reid, in ConvertFunctionType: > + PAL = 0; so there is no need for callers to set PAL = 0 before calling ConvertFunctionType, as they all do right now. For that matter, it's not clear that this assignment is needed since it looks like PAL is always assigned elsewhere before ConvertFunctionType returns (you could argue that initializing it to zero is defensive programming, defending against future changes). Ciao, Duncan. ___ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
Re: [llvm-commits] PR1146 patch for llvm-gcc-4.0
On Thu, 2007-08-16 at 16:27 +0200, Duncan Sands wrote: > Hi Reid, > > > + bool extend = false; > > it looks like extend is never assigned another value. This doesn't exist in the current patch. > > Ciao, > > Duncan. ___ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
Re: [llvm-commits] PR1146 patch for llvm-gcc-4.0
On Thu, 2007-08-16 at 16:32 +0200, Duncan Sands wrote: > Hi Reid, > > in ConvertFunctionType: > > > + PAL = 0; > > so there is no need for callers to set PAL = 0 before > calling ConvertFunctionType, as they all do right now. > For that matter, it's not clear that this assignment is > needed since it looks like PAL is always assigned > elsewhere before ConvertFunctionType returns (you could > argue that initializing it to zero is defensive programming, > defending against future changes). That is exactly the reason why it is initialized. > > Ciao, > > Duncan. ___ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
Re: [llvm-commits] PR1146 patch for llvm-gcc-4.0
On Thu, 2007-08-16 at 17:56 +0200, Duncan Sands wrote: > Hi Reid, > > > - Attrs.push_back(ParamAttrsWithIndex::get(ArgTypes.size(), > > Attributes)); > > + Client.addAttributes(ArgTypes.size(), Attributes); > > all users of addAttributes seem to be adding a new attribute at > the back. Given this, your implementation of addAttributes looks > inefficient. Also, does it work correctly? - addAttributes doesn't > add the attribute at all if there are no existing attributes for that > index, which seems to be exactly the case for every use of it! Again, this doesn't exist in the current patch. Please see the patch currently attached to PR1146 for the latest details. Thanks for taking a look at this though, I appreciate it. Reid. > > Ciao, > > Duncan. > ___ llvm-commits mailing list llvm-commits@cs.uiuc.edu http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits
