Initial implementation of on-stack replacement (OSR).

Initial implementation of on-stack replacement (OSR).

Add profiling support to select OSR candidates and launch the compiler
for OSR, followed by entry to the function at the OSR entry point.

Implemented only on IA32 and X64.  The initial implementation can be
improved in various ways --- specifically: tuning of profiling
parameters and incorporation of feedback about the actual values seen
at OSR entry.

R=fschneider@google.com

Committed: https://code.google.com/p/dart/source/detail?r=24024

[srdjan, 2013-06-11 17:12:02]

Initial comments. Do you think we need additional tests for OSR, or you feel
confident that the test suite already covers it well?

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_ia32.h
File runtime/vm/assembler_ia32.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_ia32....
runtime/vm/assembler_ia32.h:718: // The frame layout is a normal dart frame, but
the frame is partially set
s/dart/Dart/

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_x64.h
File runtime/vm/assembler_x64.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_x64.h...
runtime/vm/assembler_x64.h:739: // The frame layout is a normal dart frame, but
the frame is partially set
s/dart/Dart/

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/code_generator....
runtime/vm/code_generator.cc:1297: if (optimized_code.raw() !=
function.unoptimized_code()) {
You could check 'optimized_code.is_optimized()'

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_allo...
File runtime/vm/flow_graph_allocator.cc (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_allo...
runtime/vm/flow_graph_allocator.cc:566: if (defn->IsParameter() &&
range->spill_slot().stack_index() >= 0) {
Add parentheses.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
File runtime/vm/flow_graph_builder.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
runtime/vm/flow_graph_builder.h:176: // The deopt id of the OSR entry or
Isolate::kNoDeoptId if not compiling
Add this comment also to the constructor.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
runtime/vm/flow_graph_builder.h:178: intptr_t osr_id_;
const?

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/intermediate_la...
runtime/vm/intermediate_language.h:5923: bool is_loop() const { return is_loop_;
}
Maybe call it 'in_loop' instead?

[Florian Schneider, 2013-06-12 08:25:24]

LGTM.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
File runtime/vm/flow_graph_builder.cc (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
runtime/vm/flow_graph_builder.cc:3480: Bailout("try/catch when compiling for
OSR");
Add a TODO to enable OSR for such functions. What is missing to enable it now?

[Kevin Millikin (Google), 2013-06-14 10:10:42]

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_ia32.h
File runtime/vm/assembler_ia32.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_ia32....
runtime/vm/assembler_ia32.h:718: // The frame layout is a normal dart frame, but
the frame is partially set
On 2013/06/11 17:12:02, srdjan wrote:
> s/dart/Dart/

Done.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_x64.h
File runtime/vm/assembler_x64.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/assembler_x64.h...
runtime/vm/assembler_x64.h:739: // The frame layout is a normal dart frame, but
the frame is partially set
On 2013/06/11 17:12:02, srdjan wrote:
> s/dart/Dart/

Done.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/code_generator.cc
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/code_generator....
runtime/vm/code_generator.cc:1297: if (optimized_code.raw() !=
function.unoptimized_code()) {
On 2013/06/11 17:12:02, srdjan wrote:
> You could check 'optimized_code.is_optimized()'

Done.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_allo...
File runtime/vm/flow_graph_allocator.cc (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_allo...
runtime/vm/flow_graph_allocator.cc:566: if (defn->IsParameter() &&
range->spill_slot().stack_index() >= 0) {
On 2013/06/11 17:12:02, srdjan wrote:
> Add parentheses.

Done.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
File runtime/vm/flow_graph_builder.cc (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
runtime/vm/flow_graph_builder.cc:3480: Bailout("try/catch when compiling for
OSR");
On 2013/06/12 08:25:24, Florian Schneider wrote:
> Add a TODO to enable OSR for such functions. What is missing to enable it now?

I think they might just work but generate unreachable code in the case of
unreachable catch entries.

We have a similar problem with unreachable code elimination which can eliminate
try/catch.  I'll tackle both issues as a separate change.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
File runtime/vm/flow_graph_builder.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
runtime/vm/flow_graph_builder.h:176: // The deopt id of the OSR entry or
Isolate::kNoDeoptId if not compiling
On 2013/06/11 17:12:02, srdjan wrote:
> Add this comment also to the constructor.

Done.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/flow_graph_buil...
runtime/vm/flow_graph_builder.h:178: intptr_t osr_id_;
On 2013/06/11 17:12:02, srdjan wrote:
> const?

Done.

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/16693006/diff/2001/runtime/vm/intermediate_la...
runtime/vm/intermediate_language.h:5923: bool is_loop() const { return is_loop_;
}
On 2013/06/11 17:12:02, srdjan wrote:
> Maybe call it 'in_loop' instead?

Done.

[Kevin Millikin (Google), 2013-06-14 10:11:17]

Committed patchset #2 manually as r24024 (presubmit successful).