VM: Improve receiver class check in polymorphic inlining.

VM: Improve receiver class check in polymorphic inlining.

Avoid reloading the class-id at the entry to last inlined variant at
inlined polymorphic call sites. For this I added a new check instruction
CheckClassIdInstr to deoptimize on integer-comparison.

In polymorphic CheckClass with small dense ranges of class-ids now generate
max. 2 branch instructions by doing a range check and computing a test mask
instead of a sequence of compare and branch instructions.

R=johnmccutchan@google.com

Committed: https://code.google.com/p/dart/source/detail?r=38116

[Cutch, 2014-02-27 21:58:37]

lgtm

[Florian Schneider, 2014-03-03 15:49:01]

On 2014/02/27 21:58:37, Cutch wrote:
> lgtm

I didn't land this yet because I'm investigating a performance regression with
this CL before landing. It seems to have to do with our cid propagation.

[srdjan, 2014-03-03 18:23:49]

DBC and question: do you see any performance improvement?

https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
runtime/vm/intermediate_language.h:6908: GrowableArray<intptr_t> cids_;
Please add comment that the cids_ are sorted, lowest first.

[Florian Schneider, 2014-05-05 15:01:47]

Please take another look. I ported it to all platforms.

https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
File runtime/vm/intermediate_language.h (right):

https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
runtime/vm/intermediate_language.h:6908: GrowableArray<intptr_t> cids_;
On 2014/03/03 18:23:49, srdjan wrote:
> Please add comment that the cids_ are sorted, lowest first.

Done.

[Cutch, 2014-06-20 14:50:07]

On 2014/05/05 15:01:47, Florian Schneider wrote:
> Please take another look. I ported it to all platforms.
> 
>
https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
> File runtime/vm/intermediate_language.h (right):
> 
>
https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
> runtime/vm/intermediate_language.h:6908: GrowableArray<intptr_t> cids_;
> On 2014/03/03 18:23:49, srdjan wrote:
> > Please add comment that the cids_ are sorted, lowest first.
> 
> Done.

Is this still live?

[Florian Schneider, 2014-06-23 09:35:13]

On 2014/06/20 14:50:07, Cutch wrote:
> On 2014/05/05 15:01:47, Florian Schneider wrote:
> > Please take another look. I ported it to all platforms.
> > 
> >
>
https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
> > File runtime/vm/intermediate_language.h (right):
> > 
> >
>
https://codereview.chromium.org/181183004/diff/20001/runtime/vm/intermediate_...
> > runtime/vm/intermediate_language.h:6908: GrowableArray<intptr_t> cids_;
> > On 2014/03/03 18:23:49, srdjan wrote:
> > > Please add comment that the cids_ are sorted, lowest first.
> > 
> > Done.
> 
> Is this still live?

Kind of. It still improves code size and generates less branches. At some point
the performance results were mixed though. I have to re-do some measurements.

[Florian Schneider, 2014-07-09 09:43:06]

Reviving this old CL and ported it to arm64. I improved temp register usage and
it gives consistent 1-2% speedup on DeltaBlue.

[Cutch, 2014-07-09 19:22:53]

https://codereview.chromium.org/181183004/diff/110001/runtime/vm/disassembler...
File runtime/vm/disassembler_ia32.cc (right):

https://codereview.chromium.org/181183004/diff/110001/runtime/vm/disassembler...
runtime/vm/disassembler_ia32.cc:1821: }  // NOLINT
Did you add this to avoid the function length lint warning? I wonder if it would
be better to start refactoring the function to shorten its length.

[Florian Schneider, 2014-07-10 09:00:59]

https://codereview.chromium.org/181183004/diff/110001/runtime/vm/disassembler...
File runtime/vm/disassembler_ia32.cc (right):

https://codereview.chromium.org/181183004/diff/110001/runtime/vm/disassembler...
runtime/vm/disassembler_ia32.cc:1821: }  // NOLINT
On 2014/07/09 19:22:53, Cutch wrote:
> Did you add this to avoid the function length lint warning? I wonder if it
would
> be better to start refactoring the function to shorten its length.

Yes, I don't know if the lint warning makes sense here: The reason for this
large function is the huge switch statement, and splitting it does not make it
more readable in my opinion. It does the opposite: Making the reader jump back
and forth between helper functions that are used in exactly one place.

[Florian Schneider, 2014-07-10 11:33:20]

Committed patchset #7 manually as r38116 (presubmit successful).