Observably effectful nodes in DFG IR should come last in their bytecode instruction (i.e. forExit section), except for Hint nodes
https://bugs.webkit.org/show_bug.cgi?id=142920
Reviewed by Oliver Hunt, Geoffrey Garen, and Mark Lam.
Observably effectful, n.: If we reexecute the bytecode instruction after this node has
executed, then something other than the bytecode instruction's specified outcome will
happen.
We almost never had observably effectful nodes except at the end of the bytecode
instruction. The exception is a lowered transitioning PutById:
PutStructure(@o, S1 -> S2)
PutByOffset(@o, @o, @v)
The PutStructure is observably effectful: if you try to reexecute the bytecode after
doing the PutStructure, then we'll most likely crash. The generic PutById handling means
first checking what the old structure of the object is; but if we reexecute, the old
structure will seem to be the new structure. But the property ensured by the new
structure hasn't been stored yet, so any attempt to load it or scan it will crash.
Intriguingly, however, none of the other operations involved in the PutById are
observably effectful. Consider this example:
PutByOffset(@o, @o, @v)
PutStructure(@o, S1 -> S2)
Note that the PutStructure node doesn't reallocate property storage; see further below
for an example that does that. Because no property storage is happening, we know that we
already had room for the new property. This means that the PutByOffset is no observable
until the PutStructure executes and "reveals" the property. Hence, PutByOffset is not
observably effectful.
Now consider this:
b: AllocatePropertyStorage(@o)
PutByOffset(@b, @o, @v)
PutStructure(@o, S1 -> S2)
Surprisingly, this is also safe, because the AllocatePropertyStorage is not observably
effectful. It *does* reallocate the property storage and the new property storage pointer
is stored into the object. But until the PutStructure occurs, the world will just think
that the reallocation didn't happen, in the sense that we'll think that the property
storage is using less memory than what we just allocated. That's harmless.
The AllocatePropertyStorage is safe in other ways, too. Even if we GC'd after the
AllocatePropertyStorage but before the PutByOffset (or before the PutStructure),
everything could be expected to be fine, so long as all of @o, @v and @b are on the
stack. If they are all on the stack, then the GC will leave the property storage alone
(so the extra memory we just allocated would be safe). The GC will not scan the part of
the property storage that contains @v, but that's fine, so long as @v is on the stack.
The better long-term solution is probably bug 142921.
But for now, this:
- Fixes an object materialization bug, exemplified by the two tests, that previously
crashed 100% of the time with FTL enabled and concurrent JIT disabled.
- Allows us to remove the workaround introduced in r174856.
* dfg/DFGByteCodeParser.cpp:
(JSC::DFG::ByteCodeParser::handlePutById):
* dfg/DFGConstantFoldingPhase.cpp:
(JSC::DFG::ConstantFoldingPhase::emitPutByOffset):
* dfg/DFGFixupPhase.cpp:
(JSC::DFG::FixupPhase::insertCheck):
(JSC::DFG::FixupPhase::indexOfNode): Deleted.
(JSC::DFG::FixupPhase::indexOfFirstNodeOfExitOrigin): Deleted.
* dfg/DFGInsertionSet.h:
(JSC::DFG::InsertionSet::insertOutOfOrder): Deleted.
(JSC::DFG::InsertionSet::insertOutOfOrderNode): Deleted.
* tests/stress/materialize-past-butterfly-allocation.js: Added.
(bar):
(foo0):
(foo1):
(foo2):
(foo3):
(foo4):
* tests/stress/materialize-past-put-structure.js: Added.
(foo):
git-svn-id: http://svn.webkit.org/repository/webkit/trunk@181817 268f45cc-cd09-0410-ab3c-d52691b4dbfc
diff --git a/Source/JavaScriptCore/dfg/DFGFixupPhase.cpp b/Source/JavaScriptCore/dfg/DFGFixupPhase.cpp
index 8bc912b..b36fd91 100644
--- a/Source/JavaScriptCore/dfg/DFGFixupPhase.cpp
+++ b/Source/JavaScriptCore/dfg/DFGFixupPhase.cpp
@@ -89,33 +89,6 @@
m_insertionSet.execute(block);
}
- inline unsigned indexOfNode(Node* node, unsigned indexToSearchFrom)
- {
- unsigned index = indexToSearchFrom;
- while (index) {
- if (m_block->at(index) == node)
- break;
- index--;
- }
- ASSERT(m_block->at(index) == node);
- return index;
- }
-
- inline unsigned indexOfFirstNodeOfExitOrigin(CodeOrigin& originForExit, unsigned indexToSearchFrom)
- {
- unsigned index = indexToSearchFrom;
- ASSERT(m_block->at(index)->origin.forExit == originForExit);
- while (index) {
- index--;
- if (m_block->at(index)->origin.forExit != originForExit) {
- index++;
- break;
- }
- }
- ASSERT(m_block->at(index)->origin.forExit == originForExit);
- return index;
- }
-
void fixupNode(Node* node)
{
NodeType op = node->op();
@@ -1782,19 +1755,7 @@
void insertCheck(unsigned indexInBlock, Node* node)
{
observeUseKindOnNode<useKind>(node);
- CodeOrigin& checkedNodeOrigin = node->origin.forExit;
- CodeOrigin& currentNodeOrigin = m_currentNode->origin.forExit;
- if (currentNodeOrigin == checkedNodeOrigin) {
- // The checked node is within the same bytecode. Hence, the earliest
- // position we can insert the check is right after the checked node.
- indexInBlock = indexOfNode(node, indexInBlock) + 1;
- } else {
- // The checked node is from a preceding bytecode. Hence, the earliest
- // position we can insert the check is at the start of the current
- // bytecode.
- indexInBlock = indexOfFirstNodeOfExitOrigin(currentNodeOrigin, indexInBlock);
- }
- m_insertionSet.insertOutOfOrderNode(
+ m_insertionSet.insertNode(
indexInBlock, SpecNone, Check, m_currentNode->origin, Edge(node, useKind));
}
