Simplify handling of IC and megamorphic cache misses.

Simplify and improve handling of IC and megamorphic cache miss handling.

Instead of going back and forth from stub code to C++, perform
only the lookup in C++ and call target functions only from
stub code.

noSuchMethod and implicit closure invocations are now also work with
the megamorphic cache. Before they would go slow-case in the megamorphic case.

This CL eliminates the InstanceFunctionLookup stub that was previously
used to handle noSuchMethod and implicit closure invocations.

R=srdjan@google.com

Committed: https://code.google.com/p/dart/source/detail?r=34774

[srdjan, 2014-04-03 18:30:22]

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.cc:626: const ICData& ic_data) {
This method does not compile a function as the name implies.

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.cc:893: //   Returns: target function to call or null
if the
truncated comment

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.cc:923: arguments.SetReturn(target_function);
Who compiles the target_function? The caller of this routine expects
target_function to have been compiled.

[Florian Schneider, 2014-04-04 09:31:27]

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.cc:626: const ICData& ic_data) {
On 2014/04/03 18:30:23, srdjan wrote:
> This method does not compile a function as the name implies.

Done. Removed this helper since it is only 3 lines and used in 1 place.

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.cc:893: //   Returns: target function to call or null
if the
On 2014/04/03 18:30:23, srdjan wrote:
> truncated comment

Done.

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.cc:923: arguments.SetReturn(target_function);
On 2014/04/03 18:30:23, srdjan wrote:
> Who compiles the target_function? The caller of this routine expects
> target_function to have been compiled.

The lazy-compile stub compile the target function on invocation. There is no
need anymore to check if a function is compiled or not. There are probably still
places that explicitly check, and that still works for example if we want to
invoke a function directly from C++ code.

[srdjan, 2014-04-04 15:36:24]

lgtm

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
File runtime/vm/code_generator.cc (right):

https://codereview.chromium.org/221173011/diff/20001/runtime/vm/code_generato...
runtime/vm/code_generator.cc:923: arguments.SetReturn(target_function);
On 2014/04/04 09:31:27, Florian Schneider wrote:
> On 2014/04/03 18:30:23, srdjan wrote:
> > Who compiles the target_function? The caller of this routine expects
> > target_function to have been compiled.
> 
> The lazy-compile stub compile the target function on invocation. There is no
> need anymore to check if a function is compiled or not. There are probably
still
> places that explicitly check, and that still works for example if we want to
> invoke a function directly from C++ code.

It seems that only ARM64 is not installing the lazy compile stub.

[Florian Schneider, 2014-04-07 09:28:09]

Committed patchset #3 manually as r34774 (presubmit successful).

[zra, 2014-04-07 14:32:12]

https://codereview.chromium.org/221173011/diff/40001/runtime/vm/stub_code_arm.cc
File runtime/vm/stub_code_arm.cc (left):

https://codereview.chromium.org/221173011/diff/40001/runtime/vm/stub_code_arm...
runtime/vm/stub_code_arm.cc:446: void
StubCode::GenerateInstanceFunctionLookupStub(Assembler* assembler) {
Now that I've started on the arm64 port, it would be helpful to add/remove
stubs, etc., there as well, even though everything is UNIMPLEMENTED(), just so I
can keep it building and we can stay consistent across the architectures.