Title: [161776] branches/jsCStack/Source/_javascript_Core
- Revision
- 161776
- Author
- [email protected]
- Date
- 2014-01-11 12:31:54 -0800 (Sat, 11 Jan 2014)
Log Message
FTL tier-up should behave the same with CountExecution's as it does without
https://bugs.webkit.org/show_bug.cgi?id=126822
Not yet reviewed.
CountExecution nodes are inserted by the profiler. Prior to this change, their
presence would change how we inserted CheckTierUp nodes, since a CountExecution
might be inserted before a LoopHint thereby making it so the LoopHint was no
longer at the top of a loop header. This makes the code resilient against such
things. The result is that when you enable the profiler, you get the same OSR
behavior as with the profiler disabled.
* dfg/DFGNode.h:
(JSC::DFG::Node::isSemanticallySkippable):
* dfg/DFGOSREntrypointCreationPhase.cpp:
(JSC::DFG::OSREntrypointCreationPhase::run):
* dfg/DFGTierUpCheckInjectionPhase.cpp:
(JSC::DFG::TierUpCheckInjectionPhase::run):
Modified Paths
Diff
Modified: branches/jsCStack/Source/_javascript_Core/ChangeLog (161775 => 161776)
--- branches/jsCStack/Source/_javascript_Core/ChangeLog 2014-01-11 20:05:17 UTC (rev 161775)
+++ branches/jsCStack/Source/_javascript_Core/ChangeLog 2014-01-11 20:31:54 UTC (rev 161776)
@@ -1,3 +1,24 @@
+2014-01-11 Filip Pizlo <[email protected]>
+
+ FTL tier-up should behave the same with CountExecution's as it does without
+ https://bugs.webkit.org/show_bug.cgi?id=126822
+
+ Not yet reviewed.
+
+ CountExecution nodes are inserted by the profiler. Prior to this change, their
+ presence would change how we inserted CheckTierUp nodes, since a CountExecution
+ might be inserted before a LoopHint thereby making it so the LoopHint was no
+ longer at the top of a loop header. This makes the code resilient against such
+ things. The result is that when you enable the profiler, you get the same OSR
+ behavior as with the profiler disabled.
+
+ * dfg/DFGNode.h:
+ (JSC::DFG::Node::isSemanticallySkippable):
+ * dfg/DFGOSREntrypointCreationPhase.cpp:
+ (JSC::DFG::OSREntrypointCreationPhase::run):
+ * dfg/DFGTierUpCheckInjectionPhase.cpp:
+ (JSC::DFG::TierUpCheckInjectionPhase::run):
+
2014-01-10 Filip Pizlo <[email protected]>
FTL should work with the bytecode profiler
Modified: branches/jsCStack/Source/_javascript_Core/dfg/DFGNode.h (161775 => 161776)
--- branches/jsCStack/Source/_javascript_Core/dfg/DFGNode.h 2014-01-11 20:05:17 UTC (rev 161775)
+++ branches/jsCStack/Source/_javascript_Core/dfg/DFGNode.h 2014-01-11 20:31:54 UTC (rev 161776)
@@ -1,5 +1,5 @@
/*
- * Copyright (C) 2011, 2012, 2013 Apple Inc. All rights reserved.
+ * Copyright (C) 2011, 2012, 2013, 2014 Apple Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
@@ -1186,6 +1186,11 @@
return shouldGenerate();
}
}
+
+ bool isSemanticallySkippable()
+ {
+ return op() == CountExecution;
+ }
unsigned refCount()
{
Modified: branches/jsCStack/Source/_javascript_Core/dfg/DFGOSREntrypointCreationPhase.cpp (161775 => 161776)
--- branches/jsCStack/Source/_javascript_Core/dfg/DFGOSREntrypointCreationPhase.cpp 2014-01-11 20:05:17 UTC (rev 161775)
+++ branches/jsCStack/Source/_javascript_Core/dfg/DFGOSREntrypointCreationPhase.cpp 2014-01-11 20:31:54 UTC (rev 161776)
@@ -1,5 +1,5 @@
/*
- * Copyright (C) 2013 Apple Inc. All rights reserved.
+ * Copyright (C) 2013, 2014 Apple Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
@@ -63,7 +63,10 @@
BasicBlock* block = m_graph.block(blockIndex);
if (!block)
continue;
+ unsigned nodeIndex = 0;
Node* firstNode = block->at(0);
+ while (firstNode->isSemanticallySkippable())
+ firstNode = block->at(++nodeIndex);
if (firstNode->op() == LoopHint
&& firstNode->codeOrigin == CodeOrigin(bytecodeIndex)) {
target = block;
Modified: branches/jsCStack/Source/_javascript_Core/dfg/DFGTierUpCheckInjectionPhase.cpp (161775 => 161776)
--- branches/jsCStack/Source/_javascript_Core/dfg/DFGTierUpCheckInjectionPhase.cpp 2014-01-11 20:05:17 UTC (rev 161775)
+++ branches/jsCStack/Source/_javascript_Core/dfg/DFGTierUpCheckInjectionPhase.cpp 2014-01-11 20:31:54 UTC (rev 161776)
@@ -1,5 +1,5 @@
/*
- * Copyright (C) 2013 Apple Inc. All rights reserved.
+ * Copyright (C) 2013, 2014 Apple Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
@@ -61,15 +61,41 @@
if (!block)
continue;
- if (block->at(0)->op() == LoopHint) {
- CodeOrigin codeOrigin = block->at(0)->codeOrigin;
- NodeType nodeType;
- if (level == FTL::CanCompileAndOSREnter && !codeOrigin.inlineCallFrame) {
- nodeType = CheckTierUpAndOSREnter;
- RELEASE_ASSERT(block->bytecodeBegin == codeOrigin.bytecodeIndex);
- } else
- nodeType = CheckTierUpInLoop;
- insertionSet.insertNode(1, SpecNone, nodeType, codeOrigin);
+ for (unsigned nodeIndex = 0; nodeIndex < block->size(); ++nodeIndex) {
+ Node* node = block->at(nodeIndex);
+ if (node->op() != LoopHint)
+ continue;
+
+ // We only put OSR checks for the first LoopHint in the block. Note that
+ // more than one LoopHint could happen in cases where we did a lot of CFG
+ // simplification in the bytecode parser, but it should be very rare.
+
+ CodeOrigin codeOrigin = node->codeOrigin;
+
+ if (level != FTL::CanCompileAndOSREnter || codeOrigin.inlineCallFrame) {
+ insertionSet.insertNode(
+ nodeIndex + 1, SpecNone, CheckTierUpInLoop, codeOrigin);
+ break;
+ }
+
+ bool isAtTop = true;
+ for (unsigned subNodeIndex = nodeIndex; subNodeIndex--;) {
+ if (!block->at(subNodeIndex)->isSemanticallySkippable()) {
+ isAtTop = false;
+ break;
+ }
+ }
+
+ if (!isAtTop) {
+ insertionSet.insertNode(
+ nodeIndex + 1, SpecNone, CheckTierUpInLoop, codeOrigin);
+ break;
+ }
+
+ RELEASE_ASSERT(block->bytecodeBegin == codeOrigin.bytecodeIndex);
+ insertionSet.insertNode(
+ nodeIndex + 1, SpecNone, CheckTierUpAndOSREnter, codeOrigin);
+ break;
}
if (block->last()->op() == Return) {
_______________________________________________
webkit-changes mailing list
[email protected]
https://lists.webkit.org/mailman/listinfo/webkit-changes
