MIPS; Improve checked load/store operators for constant offset and length.

MIPS; Improve checked load/store operators for constant offset and length.

TEST=
BUG=
R=balazs.kilvady@imgtec.com, paul.lind@imgtec.com

Committed: https://chromium.googlesource.com/v8/v8/+/f62ab8d061f3dc61f24252c9676ff4446dc07b1b

[dusmil.imgtec, 2014-12-04 13:40:30]

dusan.milosavljevic@imgtec.com changed reviewers:
+ akos.palfi@imgtec.com, balazs.kilvady@imgtec.com, paul.lind@imgtec.com

[balazs.kilvady, 2014-12-04 15:32:14]

lgtm with a question.

[balazs.kilvady, 2014-12-04 15:33:41]

lgtm with a note.

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
File src/compiler/mips/code-generator-mips.cc (right):

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
src/compiler/mips/code-generator-mips.cc:179: __ Branch(ool->entry(), ls,
i.InputRegister(1), Operand(offset)); \
As I understand the operands were replaced in the Branch so the Condition should
be negated.
I think !hs == lo not ls. For the following Branches too. And for 64 bits too.

[dusmil.imgtec, 2014-12-04 15:57:04]

No, there is no negation of condition it is the same condition, the idea is to
have
right operand in branch as immediate so operands are reversed.

a >= b   <==>  b <= a

[paul.l..., 2014-12-04 16:05:56]

lgtm, but see comments.

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
File src/compiler/mips/code-generator-mips.cc (right):

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
src/compiler/mips/code-generator-mips.cc:174: __ Branch(ool->entry(), hs,
offset, i.InputOperand(1));           \
micro-optimization possible by putting the addu in the branch-delay-slot. The
change to 'at' won't matter if the branch is taken. 3 more spots in the below
macros. And in mips64.

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
src/compiler/mips/code-generator-mips.cc:210: auto value =
i.Input##width##Register(2);                    \
If you do use the delay slot of the branch, you might want to move this variable
declaration above the branch, so relation of Branch() and addu() is very clear,

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips64/code-gen...
File src/compiler/mips64/code-generator-mips64.cc (right):

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips64/code-gen...
src/compiler/mips64/code-generator-mips64.cc:176: __ Daddu(at,
i.InputRegister(2), offset);                         \
If you use delay slot, might want to make this add non-macro version for
clarity: daddu()

[dusmil.imgtec, 2014-12-04 17:25:23]

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
File src/compiler/mips/code-generator-mips.cc (right):

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
src/compiler/mips/code-generator-mips.cc:174: __ Branch(ool->entry(), hs,
offset, i.InputOperand(1));           \
Very nice, I missed it.

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
src/compiler/mips/code-generator-mips.cc:174: __ Branch(ool->entry(), hs,
offset, i.InputOperand(1));           \
On 2014/12/04 16:05:56, paul.l... wrote:
> micro-optimization possible by putting the addu in the branch-delay-slot. The
> change to 'at' won't matter if the branch is taken. 3 more spots in the below
> macros. And in mips64.

Done.

https://codereview.chromium.org/782493002/diff/1/src/compiler/mips/code-gener...
src/compiler/mips/code-generator-mips.cc:210: auto value =
i.Input##width##Register(2);                    \
On 2014/12/04 16:05:56, paul.l... wrote:
> If you do use the delay slot of the branch, you might want to move this
variable
> declaration above the branch, so relation of Branch() and addu() is very
clear,

Done.

[dusmil.imgtec, 2014-12-04 17:28:14]

Committed patchset #2 (id:20001) manually as
f62ab8d061f3dc61f24252c9676ff4446dc07b1b (presubmit successful).