[arm][stubs] Fix d16-d31 preservation on stub failure

[arm/arm64][stubs] Fix d16-d31 preservation on stub failure

Reading the registers' values back from the FrameDescription
should use the same offset computation as storing them into it.
The offsets must also match what the deoptimizer expects, which
is rx at offset rx.code() * kDoubleSize, even if some registers
are not saved (leaving gaps).

BUG=v8:4800
LOG=n
R=danno@chromium.org

Committed: https://crrev.com/32b3d3e931be1ba5540c02f656d78b3bbe11fccd
Cr-Commit-Position: refs/heads/master@{#34633}

[Jakob Kummerow, 2016-03-09 10:59:13]

Description was changed from

==========
[arm][stubs] Fix d16-d31 preservation on stub failure

Reading the registers' values back from the FrameDescription
should use the same offset computation as storing them into it.

BUG=v8:4800
LOG=n
R=danno@chromium.org
==========

to

==========
[arm/arm64][stubs] Fix d16-d31 preservation on stub failure

Reading the registers' values back from the FrameDescription
should use the same offset computation as storing them into it.
The offsets must also match what the deoptimizer expects, which
is rx at offset rx.code() * kDoubleSize, even if some registers
are not saved (leaving gaps).

BUG=v8:4800
LOG=n
R=danno@chromium.org
==========

[Jakob Kummerow, 2016-03-09 11:07:25]

Danno, PTAL.

MIPS/PPC folks: I believe your ports are correct, but you may want to
double-check, especially in case the repro triggers failures ;-)

https://codereview.chromium.org/1769833006/diff/40001/src/arm/deoptimizer-arm.cc
File src/arm/deoptimizer-arm.cc (right):

https://codereview.chromium.org/1769833006/diff/40001/src/arm/deoptimizer-arm...
src/arm/deoptimizer-arm.cc:268: for (int i = 0; i <
config->num_allocatable_double_registers(); ++i) {
cf. lines 194ff above. Both loops should follow the same logic.

https://codereview.chromium.org/1769833006/diff/40001/src/arm64/deoptimizer-a...
File src/arm64/deoptimizer-arm64.cc (right):

https://codereview.chromium.org/1769833006/diff/40001/src/arm64/deoptimizer-a...
src/arm64/deoptimizer-arm64.cc:176: (reg.code() * kDoubleSize);
The change here is s/i/reg.code()/. We must not use compact storage; non-saved
registers must leave gaps in the FrameDescription in order for
Deoptimizer::CopyDoubleRegisters (lines 79ff above) to work correctly. This is
consistent with other platforms.

[danno, 2016-03-09 17:05:00]

lgtm

[Jakob Kummerow, 2016-03-09 17:10:03]

The CQ bit was checked by jkummerow@chromium.org

[commit-bot: I haz the power, 2016-03-09 17:10:07]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1769833006/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1769833006/40001

[commit-bot: I haz the power, 2016-03-09 17:35:28]

Description was changed from

==========
[arm/arm64][stubs] Fix d16-d31 preservation on stub failure

Reading the registers' values back from the FrameDescription
should use the same offset computation as storing them into it.
The offsets must also match what the deoptimizer expects, which
is rx at offset rx.code() * kDoubleSize, even if some registers
are not saved (leaving gaps).

BUG=v8:4800
LOG=n
R=danno@chromium.org
==========

to

==========
[arm/arm64][stubs] Fix d16-d31 preservation on stub failure

Reading the registers' values back from the FrameDescription
should use the same offset computation as storing them into it.
The offsets must also match what the deoptimizer expects, which
is rx at offset rx.code() * kDoubleSize, even if some registers
are not saved (leaving gaps).

BUG=v8:4800
LOG=n
R=danno@chromium.org
==========

[commit-bot: I haz the power, 2016-03-09 17:35:29]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2016-03-09 17:36:14]

Description was changed from

==========
[arm/arm64][stubs] Fix d16-d31 preservation on stub failure

Reading the registers' values back from the FrameDescription
should use the same offset computation as storing them into it.
The offsets must also match what the deoptimizer expects, which
is rx at offset rx.code() * kDoubleSize, even if some registers
are not saved (leaving gaps).

BUG=v8:4800
LOG=n
R=danno@chromium.org
==========

to

==========
[arm/arm64][stubs] Fix d16-d31 preservation on stub failure

Reading the registers' values back from the FrameDescription
should use the same offset computation as storing them into it.
The offsets must also match what the deoptimizer expects, which
is rx at offset rx.code() * kDoubleSize, even if some registers
are not saved (leaving gaps).

BUG=v8:4800
LOG=n
R=danno@chromium.org

Committed: https://crrev.com/32b3d3e931be1ba5540c02f656d78b3bbe11fccd
Cr-Commit-Position: refs/heads/master@{#34633}
==========

[commit-bot: I haz the power, 2016-03-09 17:36:15]

Patchset 3 (id:??) landed as
https://crrev.com/32b3d3e931be1ba5540c02f656d78b3bbe11fccd
Cr-Commit-Position: refs/heads/master@{#34633}