Record a correct deoptimization environment for slow-path code.

Record a correct deoptimization environment for slow-path code.

Slow-path code is not marked as a call, so the deoptimization environment
built by the register allocator can contain unspilled registers.  However,
all registers are spilled before the call.  Compute the spill slot for all
registers and change the ones in the deoptimization environment to be
spilled.

BUG=https://code.google.com/p/dart/issues/detail?id=13995
R=fschneider@google.com

Committed: https://code.google.com/p/dart/source/detail?r=29239

[Kevin Millikin (Google), 2013-10-14 14:49:21]

I'm investigating how to reproduce the failure, so as to create a test.

[Kevin Millikin (Google), 2013-10-14 16:12:52]

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
File runtime/vm/flow_graph_compiler.cc (right):

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:499: assembler()->Comment("You betcha!");
Ignore this.

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:506:
it.SetCurrentLocation(Location::DoubleStackSlot(index));
As discussed offline, I'm not sure that this index is correct.  We may need to
subtract two here.

I'll verify with a test case.

[Florian Schneider, 2013-10-17 14:38:52]

LGTM as discussed off-line with a test case added.

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
File runtime/vm/flow_graph_compiler.cc (right):

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:496: ASSERT(index >= 0);
Shouldn't this be even stronger?

ASSERT(index >= 0 && index >= StackSize());

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:501: ASSERT(index >= 0);
ASSERT(index >= 0 && index >= StackSize());

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:510:
it.SetCurrentLocation(Location::QuadStackSlot(index));
Try to also have a test case for the quad-word case here.

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/locations.h
File runtime/vm/locations.h (right):

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/locations.h#new...
runtime/vm/locations.h:371: intptr_t CpuRegisterCount() const { return
RegisterCount(cpu_registers_); }
I don't see that function used anywhere. Do you anticipate a use for it?

[Kevin Millikin (Google), 2013-10-25 10:17:26]

Committed patchset #3 manually as r29239 (presubmit successful).

[srdjan, 2013-12-03 22:42:10]

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
File runtime/vm/flow_graph_compiler.cc (right):

https://codereview.chromium.org/26823006/diff/6001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:499: assembler()->Comment("You betcha!");
On 2013/10/14 16:12:52, kmillikin wrote:
> Ignore this.

But it was submitted. Can we replace it with something more meaningful :-)?