Reland r17365.

Reland r17365.

Introduce InvokeMathCFunction that can be used to directly invoke mathematical
function provided by runtime.

Use it to unconditionally inline _Double.pow.

Use it to inline floor, ceil, round, truncate, round when SSE4.1 is not
available.

Perform representation selection phase after constant propagation to minimize
boxing.

Add support for enter instruction in the x64 disassembler.

Add test for optimized pow and fix compilation on windows.

R=fschneider@google.com
BUG=dart:8002
Committed: https://code.google.com/p/dart/source/detail?r=17393

[Florian Schneider, 2013-01-22 11:22:34]

LGTM.

https://codereview.chromium.org/12038013/diff/14/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language_x64.cc (right):

https://codereview.chromium.org/12038013/diff/14/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language_x64.cc:2427: __
ReserveAlignedFrameSpace(kDoubleSize * InputCount());
Should be ReserveAlignedFrameSpace(0).

https://codereview.chromium.org/12038013/diff/14/tests/language/math_vm_test....
File tests/language/math_vm_test.dart (right):

https://codereview.chromium.org/12038013/diff/14/tests/language/math_vm_test....
tests/language/math_vm_test.dart:47: for (int i = 0; i < 200; i++) {
Not sure if 200 is enough to get it optimized reliably. Maybe use 2000 instead.

[Vyacheslav Egorov (Google), 2013-01-22 11:36:31]

Thanks for the review. Relanding.

https://codereview.chromium.org/12038013/diff/14/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language_x64.cc (right):

https://codereview.chromium.org/12038013/diff/14/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language_x64.cc:2427: __
ReserveAlignedFrameSpace(kDoubleSize * InputCount());
On 2013/01/22 11:22:34, Florian Schneider wrote:
> Should be ReserveAlignedFrameSpace(0).

Done.

https://codereview.chromium.org/12038013/diff/14/tests/language/math_vm_test....
File tests/language/math_vm_test.dart (right):

https://codereview.chromium.org/12038013/diff/14/tests/language/math_vm_test....
tests/language/math_vm_test.dart:47: for (int i = 0; i < 200; i++) {
On 2013/01/22 11:22:34, Florian Schneider wrote:
> Not sure if 200 is enough to get it optimized reliably. Maybe use 2000
instead.

This tests comes with  VMOptions=--optimization_counter_threshold=100

[srdjan, 2013-01-22 21:19:41]

DBC

https://codereview.chromium.org/12038013/diff/14/runtime/vm/flow_graph_optimi...
File runtime/vm/flow_graph_optimizer.cc (right):

https://codereview.chromium.org/12038013/diff/14/runtime/vm/flow_graph_optimi...
runtime/vm/flow_graph_optimizer.cc:1294: MethodRecognizer::Kind recognized_kind)
{
Indent 4 spaces

https://codereview.chromium.org/12038013/diff/14/runtime/vm/flow_graph_optimi...
runtime/vm/flow_graph_optimizer.cc:1380: if
(!CPUFeatures::double_truncate_round_supported()) {
I prefer to test for positive instead of for negative as IMHO it makes code more
readable.

https://codereview.chromium.org/12038013/diff/14/runtime/vm/intermediate_lang...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/12038013/diff/14/runtime/vm/intermediate_lang...
runtime/vm/intermediate_language.h:4067: : inputs_(inputs), locs_(NULL),
recognized_kind_(recognized_kind) {
indent 4 spaces