AuroraMiddleware/v8
1
0
Fork 0
Code Issues 2 Pull Requests Releases Wiki Activity

Fix aliasing problem in inlined stores on x64 and ia32. The receiver

Browse Source 
and the value can share a register. We need to remove this aliasing
before modifying the registers.

I haven't managed to generate a stand-alon test case for this
yet. I'll do that as a separate change. This was found while loading
Wave.

Review URL: http://codereview.chromium.org/3039025

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5127 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This commit is contained in:
ager@chromium.org 2010-07-23 11:55:03 +00:00
parent e8c0a459a9
commit 79e332010a
2 changed files with 18 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

13
src/ia32/codegen-ia32.cc
View File

@ -8968,16 +8968,21 @@ Result CodeGenerator::EmitNamedStore(Handle<String> name, bool is_contextual) {
// Allocate scratch register for write barrier.
Result scratch = allocator()->Allocate();
ASSERT(scratch.is_valid() &&
result.is_valid() &&
receiver.is_valid() &&
value.is_valid());
ASSERT(scratch.is_valid());
// The write barrier clobbers all input registers, so spill the
// receiver and the value.
frame_->Spill(receiver.reg());
frame_->Spill(value.reg());
// If the receiver and the value share a register allocate a new
// register for the receiver.
if (receiver.reg().is(value.reg())) {
receiver = allocator()->Allocate();
ASSERT(receiver.is_valid());
__ mov(receiver.reg(), Operand(value.reg()));
}
// Update the write barrier. To save instructions in the inlined
// version we do not filter smis.
Label skip_write_barrier;

13
src/x64/codegen-x64.cc
View File

@ -8103,16 +8103,21 @@ Result CodeGenerator::EmitNamedStore(Handle<String> name, bool is_contextual) {
// Allocate scratch register for write barrier.
Result scratch = allocator()->Allocate();
ASSERT(scratch.is_valid() &&
result.is_valid() &&
receiver.is_valid() &&
value.is_valid());
ASSERT(scratch.is_valid());
// The write barrier clobbers all input registers, so spill the
// receiver and the value.
frame_->Spill(receiver.reg());
frame_->Spill(value.reg());
// If the receiver and the value share a register allocate a new
// register for the receiver.
if (receiver.reg().is(value.reg())) {
receiver = allocator()->Allocate();
ASSERT(receiver.is_valid());
__ movq(receiver.reg(), value.reg());
}
// Update the write barrier. To save instructions in the inlined
// version we do not filter smis.
Label skip_write_barrier;