Clean up ARM assembly code loading and storing from/to a large offset.

Clean up ARM assembly code loading and storing from/to a large offset.
Make use of vldm and vstm ARM instructions.
Implement ARM stub printing stop message.
Committed: https://code.google.com/p/dart/source/detail?r=19571

[zra, 2013-03-06 03:30:28]

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc
File runtime/vm/assembler_arm.cc (right):

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc#n...
runtime/vm/assembler_arm.cc:1277: bool
Address::CanHoldLoadOffset(LoadOperandType type,
I just want to double-check this and CanHoldStoreOffset against the spec first
thing tomorrow morning, but otherwise I think these changes are good.

[zra, 2013-03-06 16:41:27]

lgtm with comment

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc
File runtime/vm/assembler_arm.cc (right):

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc#n...
runtime/vm/assembler_arm.cc:1256: uint32_t offset_lo = offset & offset_mask;  //
unsigned
ASSERT(CanHoldLoadOffset(offset_lo))

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc#n...
runtime/vm/assembler_arm.cc:1454: int offset_lo = offset & offset_mask;
For consistency, you might put ASSERT(CanHoldLoadOffset(..., offset_lo)) here
also like you have down in LoadFromOffset. Same in BranchLinkPatchable.

[regis, 2013-03-06 18:42:00]

Thanks

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc
File runtime/vm/assembler_arm.cc (right):

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc#n...
runtime/vm/assembler_arm.cc:1256: uint32_t offset_lo = offset & offset_mask;  //
unsigned
On 2013/03/06 16:41:27, zra wrote:
> ASSERT(CanHoldLoadOffset(offset_lo))

Actually, each instruction already asserts that the offset is in range. So
instead of adding extra asserts here, I prefer to remove those in LoadFromOffset
and friends.

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc#n...
runtime/vm/assembler_arm.cc:1277: bool
Address::CanHoldLoadOffset(LoadOperandType type,
On 2013/03/06 03:30:28, zra wrote:
> I just want to double-check this and CanHoldStoreOffset against the spec first
> thing tomorrow morning, but otherwise I think these changes are good.

OK. I'll revisit if you find a problem.

https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc#n...
runtime/vm/assembler_arm.cc:1454: int offset_lo = offset & offset_mask;
On 2013/03/06 16:41:27, zra wrote:
> For consistency, you might put ASSERT(CanHoldLoadOffset(..., offset_lo)) here
> also like you have down in LoadFromOffset. Same in BranchLinkPatchable.

I am all for consistency. So I'll remove the extra ones, instead of adding
redundant ones.

[zra, 2013-03-06 18:48:06]

On 2013/03/06 18:42:00, regis wrote:
>
https://codereview.chromium.org/12481007/diff/1/runtime/vm/assembler_arm.cc#n...
> runtime/vm/assembler_arm.cc:1277: bool
> Address::CanHoldLoadOffset(LoadOperandType type,
> On 2013/03/06 03:30:28, zra wrote:
> > I just want to double-check this and CanHoldStoreOffset against the spec
first
> > thing tomorrow morning, but otherwise I think these changes are good.
> 
> OK. I'll revisit if you find a problem.

I already looked this morning, and didn't find any problems.

[regis, 2013-03-06 18:49:47]

Committed patchset #3 manually as r19571 (presubmit successful).