Enable language tests on SIMARM and mark 14 of them as failing.

Enable language tests on SIMARM and mark 14 of them as failing.
Implement more features on ARM.

R=zra@google.com

Committed: https://code.google.com/p/dart/source/detail?r=23448

[zra, 2013-05-30 23:17:35]

lgtm

https://codereview.chromium.org/16244007/diff/1/runtime/vm/flow_graph_compile...
File runtime/vm/flow_graph_compiler_arm.cc (right):

https://codereview.chromium.org/16244007/diff/1/runtime/vm/flow_graph_compile...
runtime/vm/flow_graph_compiler_arm.cc:1423: __ b(&fall_through, NE);
The code is a bit dense up to here. A comment or two will help me translate to
MIPS =)

https://codereview.chromium.org/16244007/diff/1/runtime/vm/flow_graph_compile...
runtime/vm/flow_graph_compiler_arm.cc:1544: BlockEntryInstr* nan_result =
(true_condition == NE) ?
I'm not familiar with the spec here, but I thought that any comparison with NaN
usually results in false, even NE. Or is this instruction like this?

if(!(float1 op float2)) {} else {}

[regis, 2013-05-31 00:37:04]

Thanks!

https://codereview.chromium.org/16244007/diff/1/runtime/vm/flow_graph_compile...
File runtime/vm/flow_graph_compiler_arm.cc (right):

https://codereview.chromium.org/16244007/diff/1/runtime/vm/flow_graph_compile...
runtime/vm/flow_graph_compiler_arm.cc:1423: __ b(&fall_through, NE);
On 2013/05/30 23:17:35, zra wrote:
> The code is a bit dense up to here. A comment or two will help me translate to
> MIPS =)

:-)
Done and simplified a bit.
It will be easier for you to port the ia32 version.

https://codereview.chromium.org/16244007/diff/1/runtime/vm/flow_graph_compile...
runtime/vm/flow_graph_compiler_arm.cc:1544: BlockEntryInstr* nan_result =
(true_condition == NE) ?
On 2013/05/30 23:17:35, zra wrote:
> I'm not familiar with the spec here, but I thought that any comparison with
NaN
> usually results in false, even NE. Or is this instruction like this?
> 
> if(!(float1 op float2)) {} else {}

In your example, the negation is outside of this instruction. It would be wrong
to propagate it down to the op level and compile a !op.
The special handling applies only to float1 != float2, which is true in the NaN
case (according to our Intel implementation).

[regis, 2013-05-31 00:38:36]

Committed patchset #2 manually as r23448 (presubmit successful).

[kustermann, 2013-06-03 07:54:26]

https://codereview.chromium.org/16244007/diff/7001/tests/language/language.st...
File tests/language/language.status (right):

https://codereview.chromium.org/16244007/diff/7001/tests/language/language.st...
tests/language/language.status:805: *: Skip
Why do you skip all language tests on arm?
I setup a new builder on FYI which builds and runs arm tests on real hardware.
Now it's not doing anything, because you skip all the tests on arm.
What is the reason for that?

[zra, 2013-06-03 15:30:52]

https://codereview.chromium.org/16244007/diff/7001/tests/language/language.st...
File tests/language/language.status (right):

https://codereview.chromium.org/16244007/diff/7001/tests/language/language.st...
tests/language/language.status:805: *: Skip
On 2013/06/03 07:54:26, kustermann wrote:
> Why do you skip all language tests on arm?
> I setup a new builder on FYI which builds and runs arm tests on real hardware.
> Now it's not doing anything, because you skip all the tests on arm.
> What is the reason for that?

I think this was probably a misunderstanding. The stuff from simarm below should
probably be copied up here to the arm section.

[kustermann, 2013-06-03 15:40:39]

https://codereview.chromium.org/16244007/diff/7001/tests/language/language.st...
File tests/language/language.status (right):

https://codereview.chromium.org/16244007/diff/7001/tests/language/language.st...
tests/language/language.status:805: *: Skip
On 2013/06/03 15:30:52, zra wrote:
> On 2013/06/03 07:54:26, kustermann wrote:
> > Why do you skip all language tests on arm?
> > I setup a new builder on FYI which builds and runs arm tests on real
hardware.
> > Now it's not doing anything, because you skip all the tests on arm.
> > What is the reason for that?
> 
> I think this was probably a misunderstanding. The stuff from simarm below
should
> probably be copied up here to the arm section.

In case this was an accident: There a couple of other status files where arm
tests were marked as skip (we don't run any tests at the moment). You might want
to update these as well.

I don't know what your plans are, but if you enable all the tests on arm and
update the status files for failing/crashing tests we could move the arm board
builder from FYI to the main waterfall.

[zra, 2013-06-03 15:44:22]

On 2013/06/03 15:40:39, kustermann wrote:
> In case this was an accident: There a couple of other status files where arm
> tests were marked as skip (we don't run any tests at the moment). You might
want
> to update these as well.
> 
> I don't know what your plans are, but if you enable all the tests on arm and
> update the status files for failing/crashing tests we could move the arm board
> builder from FYI to the main waterfall.

I'll fix-up the status files so that arm == simarm. If it looks good on FYI,
then I think moving over to the main waterfall is a good idea.

Thanks,
Zach