Re-assign registers on ARM so PP and CODE_REG are below R7 (FP on iOS).

Re-assign registers on ARM so PP and CODE_REG are below R7 (FP on iOS).

Also update descriptions of register availability in instrinsics.

1) Mechanical replacement in *arm.* files except the simulator and constants_arm.h (with sed).

R5 -> R9
R6 -> R8
R8 -> R10
R9 -> R5
R10 -> R6

2) Change register aliases.

CTX = R10 -> CTX = R6
PP = R9 -> PP = R5
ICREG = R5 -> ICREG = R9
CODE_REG = R10 -> CODE_REG = R6
THR = R8 -> THR = R10

3) Fixup users of ldrd/strd to have valid register pairs.

bigint_lsh
bigint_absadd
bigint_abssub
bigint_sqradd
TYPED_ARRAY_ALLOCATION
InlineArrayAllocation
AllocateArrayStub
AllocateContextStub

4) Fix PushList in MegamorphicMissStub to keep runtime call arguments in the right order.

BUG=http://dartbug.com/24683
R=fschneider@google.com

Committed: https://github.com/dart-lang/sdk/commit/dd7250323d671248fa03c79bf0718fa100fd6e79

[rmacnak, 2015-10-23 19:59:35]

rmacnak@google.com changed reviewers:
+ fschneider@google.com, srdjan@google.com, zra@google.com

[rmacnak, 2015-10-23 19:59:35]

Depends on https://codereview.chromium.org/1418813005/

[Florian Schneider, 2015-10-26 17:12:31]

lgtm

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/instructions...
File runtime/vm/instructions_arm.cc (right):

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/instructions...
runtime/vm/instructions_arm.cc:183: if ((instr & 0xffff0000) == 0xe5950000) { 
// ldr reg, [pp, #+offset]
I remember these as easy to forget to change, but forgot to fix it then. Should
we maybe use PP here as well (and in the cases below)? i.e. 0xe59 | (PP << 16)

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/intrinsifier...
File runtime/vm/intrinsifier_arm.cc (left):

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/intrinsifier...
runtime/vm/intrinsifier_arm.cc:317: __ SmiUntag(R0);  // Untags R6. We only want
result shifted by one.
This comments seems out-of-sync already here. s/R6/R0/

[rmacnak, 2015-10-26 18:01:50]

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/instructions...
File runtime/vm/instructions_arm.cc (right):

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/instructions...
runtime/vm/instructions_arm.cc:183: if ((instr & 0xffff0000) == 0xe5950000) { 
// ldr reg, [pp, #+offset]
On 2015/10/26 17:12:31, Florian Schneider wrote:
> I remember these as easy to forget to change, but forgot to fix it then.
Should
> we maybe use PP here as well (and in the cases below)? i.e. 0xe59 | (PP << 16)

Sounds fine.

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/intrinsifier...
File runtime/vm/intrinsifier_arm.cc (left):

https://codereview.chromium.org/1419223003/diff/20001/runtime/vm/intrinsifier...
runtime/vm/intrinsifier_arm.cc:317: __ SmiUntag(R0);  // Untags R6. We only want
result shifted by one.
On 2015/10/26 17:12:31, Florian Schneider wrote:
> This comments seems out-of-sync already here. s/R6/R0/

Done.

[rmacnak, 2015-10-26 18:05:36]

Committed patchset #3 (id:40001) manually as
dd7250323d671248fa03c79bf0718fa100fd6e79.