Implement missing features to run Hello world! on simulated ARM.

Implement missing features to run Hello world! on simulated ARM.

Committed: https://code.google.com/p/dart/source/detail?r=21582

[zra, 2013-04-16 21:12:29]

lgtm

Hello, SIMARM!

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language_arm.cc (right):

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_arm.cc:1029: // allocates registers.
This sounds a bit hack-y. How would I know what register to use on MIPS?

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_arm.cc:1043: value_cid_reg = R3;
Same here, I guess?

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.cc
File runtime/vm/stub_code_arm.cc (right):

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:347: Label size_tag_overflow, done;
These labels aren't used.

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:443: // Load closure function code in EAX.
R0?

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:941: // EAX: new object.
R0

[regis, 2013-04-16 21:43:39]

Thanks!

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/intermediate_la...
File runtime/vm/intermediate_language_arm.cc (right):

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_arm.cc:1029: // allocates registers.
On 2013/04/16 21:12:29, zra wrote:
> This sounds a bit hack-y. How would I know what register to use on MIPS?

Agreed. As the comment explains (copied from the Intel side), this is due to the
sharing of locations between optimized and unoptimized code. This is not
specific to ARM or MIPS. So I did not want to do some redesign work here.

My understanding is the following: This 'hack' or implicit use of a register
happens only in unoptimized code, where no register survives between flow graph
instructions, since the stack is used as an expression stack for all temporary
values.
On MIPS, you would pick any register after the first and second one (V0 and V1)
allocated by the allocator, i.e. A0.

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/intermediate_la...
runtime/vm/intermediate_language_arm.cc:1043: value_cid_reg = R3;
On 2013/04/16 21:12:29, zra wrote:
> Same here, I guess?

Yes. A1 on MIPS. As the TODO below indicates, we may have to revisit and had
more such cases.

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.cc
File runtime/vm/stub_code_arm.cc (right):

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:347: Label size_tag_overflow, done;
On 2013/04/16 21:12:29, zra wrote:
> These labels aren't used.

Removed. And removed local scope.

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:443: // Load closure function code in EAX.
On 2013/04/16 21:12:29, zra wrote:
> R0?

Done.

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:941: // EAX: new object.
On 2013/04/16 21:12:29, zra wrote:
> R0

Done.

[regis, 2013-04-16 21:53:15]

Committed patchset #3 manually as r21582 (presubmit successful).

[zra, 2013-04-16 23:17:42]

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.cc
File runtime/vm/stub_code_arm.cc (right):

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:284: __ tst(R2, ShifterOperand(kSmiTagSize));
Just noticed this. Should this be kSmiTagMask? both Mask and Size are 1, so this
will work, but everywhere else it's Mask.

[regis, 2013-04-16 23:22:04]

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.cc
File runtime/vm/stub_code_arm.cc (right):

https://codereview.chromium.org/14153004/diff/5002/runtime/vm/stub_code_arm.c...
runtime/vm/stub_code_arm.cc:284: __ tst(R2, ShifterOperand(kSmiTagSize));
On 2013/04/16 23:17:42, zra wrote:
> Just noticed this. Should this be kSmiTagMask? both Mask and Size are 1, so
this
> will work, but everywhere else it's Mask.

You are right. It is a typo on all architectures, both here and also in
intrinsifier_<arch>.cc

I'll fix that in my next cl. Thanks!