A64: Add support for --optimize-for-size in PushMultipleTimes

A64: Add support for --optimize-for-size in PushMultipleTimes

BUG=none
R=rodolph.perfetta@arm.com, ulan@chromium.org
LOG=n

Committed: https://code.google.com/p/v8/source/detail?r=19164

[ulan, 2014-02-06 14:53:12]

lgtm, consider the optimization below.

https://codereview.chromium.org/140893009/diff/1/src/a64/macro-assembler-a64.cc
File src/a64/macro-assembler-a64.cc (right):

https://codereview.chromium.org/140893009/diff/1/src/a64/macro-assembler-a64....
src/a64/macro-assembler-a64.cc:681: PushHelper(1, size, src, NoReg, NoReg,
NoReg);
Optimization: push 4 values at once (count / 4) times, and then set count =
(count % 4) and fall through to the old code.

[Rodolph Perfetta (ARM), 2014-02-06 15:14:35]

https://codereview.chromium.org/140893009/diff/1/src/a64/macro-assembler-a64.cc
File src/a64/macro-assembler-a64.cc (right):

https://codereview.chromium.org/140893009/diff/1/src/a64/macro-assembler-a64....
src/a64/macro-assembler-a64.cc:677: if (FLAG_optimize_for_size && count > 4) {
The code below is 4 instructions long, PushMultipleTimes can handle 8 push in 4
instructions (using stp) so the break even point should be 8.

https://codereview.chromium.org/140893009/diff/1/src/a64/macro-assembler-a64....
src/a64/macro-assembler-a64.cc:681: PushHelper(1, size, src, NoReg, NoReg,
NoReg);
if count is even you can push two at a time with no extra memory cost.

https://codereview.chromium.org/140893009/diff/1/src/a64/macro-assembler-a64....
src/a64/macro-assembler-a64.cc:681: PushHelper(1, size, src, NoReg, NoReg,
NoReg);
4 at a time will be faster but bigger, is that what we want?

On 2014/02/06 14:53:12, ulan wrote:
> Optimization: push 4 values at once (count / 4) times, and then set count =
> (count % 4) and fall through to the old code.

[jochen (gone - plz use gerrit), 2014-02-06 15:26:34]

PTAL

I agree that not falling through will be smaller, but not a lot. I think pushing
2 values at once it's a nice compromise, so at least we can use stp instead of
str

[Rodolph Perfetta (ARM), 2014-02-06 15:37:15]

lgtm

[jochen (gone - plz use gerrit), 2014-02-06 15:38:14]

Committed patchset #3 manually as r19164 (presubmit successful).