Flow graph SIMD changes

Flow graph SIMD changes

Committed: https://code.google.com/p/dart/source/detail?r=20945

[Cutch, 2013-04-02 22:52:04]

Changes to allocator, ParallelMoveResolver::EmitMove, and
ParallelMoveResolver::EmitSwap (IA32/x64) to support unboxed SIMD values.

[Cutch, 2013-04-02 22:55:03]

$ ./tools/test.py -m all -a all
Test configurations: none_vm_debug_ia32 none_vm_release_ia32 none_vm_debug_x64
none_vm_release_x64 none_vm_debug_simarm none_vm_release_simarm
none_vm_debug_simmips none_vm_release_simmips
[00:00 |  --% | +    0 | -    0]
[09:33 | 100% | +27263 | -    0]

[Vyacheslav Egorov (Google), 2013-04-03 14:25:45]

lgtm

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_allocat...
File runtime/vm/flow_graph_allocator.cc (right):

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_allocat...
runtime/vm/flow_graph_allocator.cc:1609: location =
Location::Float32x4StackSlot(cpu_spill_slot_count_ + slot_idx,
moving this duplicated expression: cpu_spill_slot_count_ + slot_idx up and
storing it in a variable.

you can even make slot_idx to be equal to 

cpu_spill_slot_count_ + idx * ...

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_allocat...
runtime/vm/flow_graph_allocator.cc:1610: range->representation());
I find it confusing that we still pass representation to a location that
obviously has fixed representation. 

Consider making it fixed internally.

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_compile...
File runtime/vm/flow_graph_compiler.cc (right):

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_compile...
runtime/vm/flow_graph_compiler.cc:418: const intptr_t kFpuRegisterSpillFactor =
you can kill this variable. it serves no purpose now.

[Cutch, 2013-04-04 22:24:06]

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_allocat...
File runtime/vm/flow_graph_allocator.cc (right):

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_allocat...
runtime/vm/flow_graph_allocator.cc:1609: location =
Location::Float32x4StackSlot(cpu_spill_slot_count_ + slot_idx,
On 2013/04/03 14:25:45, Vyacheslav Egorov (Google) wrote:
> moving this duplicated expression: cpu_spill_slot_count_ + slot_idx up and
> storing it in a variable.
> 
> you can even make slot_idx to be equal to 
> 
> cpu_spill_slot_count_ + idx * ...

Done.

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_allocat...
runtime/vm/flow_graph_allocator.cc:1610: range->representation());
On 2013/04/03 14:25:45, Vyacheslav Egorov (Google) wrote:
> I find it confusing that we still pass representation to a location that
> obviously has fixed representation. 
> 
> Consider making it fixed internally.

Done.

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_compile...
File runtime/vm/flow_graph_compiler.cc (right):

https://codereview.chromium.org/13471013/diff/1/runtime/vm/flow_graph_compile...
runtime/vm/flow_graph_compiler.cc:418: const intptr_t kFpuRegisterSpillFactor =
On 2013/04/03 14:25:45, Vyacheslav Egorov (Google) wrote:
> you can kill this variable. it serves no purpose now.

It makes the for loop on line 423 easier to read and fit within 80 columns.

[srdjan, 2013-04-04 23:03:32]

LGTM

https://codereview.chromium.org/13471013/diff/7001/runtime/vm/flow_graph_comp...
File runtime/vm/flow_graph_compiler.cc (right):

https://codereview.chromium.org/13471013/diff/7001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:419:
FlowGraphAllocator::kFpuRegisterSpillFactor;
indent 4 spaces

[Cutch, 2013-04-04 23:06:43]

https://codereview.chromium.org/13471013/diff/7001/runtime/vm/flow_graph_comp...
File runtime/vm/flow_graph_compiler.cc (right):

https://codereview.chromium.org/13471013/diff/7001/runtime/vm/flow_graph_comp...
runtime/vm/flow_graph_compiler.cc:419:
FlowGraphAllocator::kFpuRegisterSpillFactor;
On 2013/04/04 23:03:32, srdjan wrote:
> indent 4 spaces

Done.

[Cutch, 2013-04-04 23:23:38]

Committed patchset #3 manually as r20945 (presubmit successful).