Improve smi shift operations and avoid repeated deoptimizations.

Improve smi shift operations and avoid repeated deoptimizations.

Smi left shifts do not contain a call now. Only after deoptimizing
at a smi operation we we generate mint code or a generic call to
avoid repeated deoptimization.

Added a test case for the code that was not covered by existing tests.
Committed: https://code.google.com/p/dart/source/detail?r=14709

[srdjan, 2012-11-08 21:37:27]

LGTM

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
File runtime/vm/deopt_instructions.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
runtime/vm/deopt_instructions.cc:272: }
This could be:
uword pc = code.GetPcForDeoptId(deopt_id_, PcDescriptors::kIcCall)
ic_data = CodePatcher::GetInstanceCallIcDataAt(pc);

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
File runtime/vm/deopt_instructions.h (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
runtime/vm/deopt_instructions.h:78: DeoptReasonId deopt_reason_;
const

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
File runtime/vm/flow_graph_optimizer.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
runtime/vm/flow_graph_optimizer.cc:693: if (ic_data.deopt_reason() ==
kDeoptBinaryMintOp)  return false;
one space too much before 'return false'.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
runtime/vm/flow_graph_optimizer.cc:706: if (ic_data.deopt_reason() ==
kDeoptBinarySmiOp)  return false;
ditto

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
runtime/vm/flow_graph_optimizer.cc:744: // of an overflow.
Why is test FlowGraphCompiler::SupportsUnboxedMints() removed?

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language_ia32.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_ia32.cc:1828: if (value < 0 || value >=
kCountLimit) {
Add parenthesis.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_ia32.cc:1829: __ jmp(deopt);
Make a note that this may not be known at instruction creation time (because of
constant propagation, inlining, ...)

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language_x64.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_x64.cc:1703: if (value < 0 || value >=
kCountLimit) {
Same comments as from ia32.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/raw_object.h
File runtime/vm/raw_object.h (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/raw_object.h#ne...
runtime/vm/raw_object.h:928: intptr_t deopt_reason_;     // Last deoptimization
reason.
uint8_t

[Florian Schneider, 2012-11-08 22:03:00]

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
File runtime/vm/deopt_instructions.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
runtime/vm/deopt_instructions.cc:272: }
On 2012/11/08 21:37:27, srdjan wrote:
> This could be:
> uword pc = code.GetPcForDeoptId(deopt_id_, PcDescriptors::kIcCall)
> ic_data = CodePatcher::GetInstanceCallIcDataAt(pc);

Done.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
File runtime/vm/deopt_instructions.h (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/deopt_instructi...
runtime/vm/deopt_instructions.h:78: DeoptReasonId deopt_reason_;
On 2012/11/08 21:37:27, srdjan wrote:
> const

Done.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
File runtime/vm/flow_graph_optimizer.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
runtime/vm/flow_graph_optimizer.cc:693: if (ic_data.deopt_reason() ==
kDeoptBinaryMintOp)  return false;
On 2012/11/08 21:37:27, srdjan wrote:
> one space too much before 'return false'.

Done.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
runtime/vm/flow_graph_optimizer.cc:706: if (ic_data.deopt_reason() ==
kDeoptBinarySmiOp)  return false;
On 2012/11/08 21:37:27, srdjan wrote:
> ditto

Done.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/flow_graph_opti...
runtime/vm/flow_graph_optimizer.cc:744: // of an overflow.
On 2012/11/08 21:37:27, srdjan wrote:
> Why is test FlowGraphCompiler::SupportsUnboxedMints() removed?

Instead of testing it here in many places, I moved it down for less code
duplication.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language_ia32.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_ia32.cc:1829: __ jmp(deopt);
On 2012/11/08 21:37:27, srdjan wrote:
> Make a note that this may not be known at instruction creation time (because
of
> constant propagation, inlining, ...)

Done.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language_x64.cc (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_x64.cc:1703: if (value < 0 || value >=
kCountLimit) {
On 2012/11/08 21:37:27, srdjan wrote:
> Same comments as from ia32.

Done.

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/raw_object.h
File runtime/vm/raw_object.h (right):

https://codereview.chromium.org/11363141/diff/1012/runtime/vm/raw_object.h#ne...
runtime/vm/raw_object.h:928: intptr_t deopt_reason_;     // Last deoptimization
reason.
On 2012/11/08 21:37:27, srdjan wrote:
> uint8_t

Done.