MIPS: Truncate booleans to 0/1 in truncating t-to-i.

MIPS: Truncate booleans to 0/1 in truncating t-to-i.

Port r17166 (0a4bf790)

Original commit message:
Truncate booleans to 0/1 in truncating t-to-i.

Thanks to weiliang.lin2@gmail.com for discovering the issue.

BUG=
R=plind44@gmail.com

Committed: https://code.google.com/p/v8/source/detail?r=17173

[palfia, 2013-10-11 20:46:16]

PTAL.

[Paul Lind, 2013-10-11 21:18:30]

LGTM, but please update per comments about delay slots. 

You should probably run a subset of tests again, just to be sure, at least the
new test test/mjsunit/bitwise-operations-bools.js

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips.cc
File src/mips/lithium-codegen-mips.cc (right):

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips...
src/mips/lithium-codegen-mips.cc:4911: __ Branch(&done);
We should probably use the delay slot here (and 2 more place below) to save an
instruction.
__ Branch(USE_DELAY_SLOT, &done);
__ mov(input_reg, zero_reg);  // In delay slot.

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips...
src/mips/lithium-codegen-mips.cc:4916: __ li(input_reg, Operand(1));
li() of 1 is always single instruction, use delay slot here also.

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips...
src/mips/lithium-codegen-mips.cc:4923: __ Branch(&done);
Use delay slot.

[palfia, 2013-10-11 23:21:51]

Thanks for optimizing this! I've uploaded the new patch.

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips.cc
File src/mips/lithium-codegen-mips.cc (right):

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips...
src/mips/lithium-codegen-mips.cc:4911: __ Branch(&done);
On 2013/10/11 21:18:30, Paul Lind wrote:
> We should probably use the delay slot here (and 2 more place below) to save an
> instruction.
> __ Branch(USE_DELAY_SLOT, &done);
> __ mov(input_reg, zero_reg);  // In delay slot.

Done.

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips...
src/mips/lithium-codegen-mips.cc:4916: __ li(input_reg, Operand(1));
On 2013/10/11 21:18:30, Paul Lind wrote:
> li() of 1 is always single instruction, use delay slot here also.

Done.

https://codereview.chromium.org/26742006/diff/1/src/mips/lithium-codegen-mips...
src/mips/lithium-codegen-mips.cc:4923: __ Branch(&done);
On 2013/10/11 21:18:30, Paul Lind wrote:
> Use delay slot.

Done.

[Paul Lind, 2013-10-11 23:26:14]

LGTM, with a comment-only nit. Go ahead and fix on the dcommit push if you want.

https://codereview.chromium.org/26742006/diff/12001/src/mips/lithium-codegen-...
File src/mips/lithium-codegen-mips.cc (right):

https://codereview.chromium.org/26742006/diff/12001/src/mips/lithium-codegen-...
src/mips/lithium-codegen-mips.cc:4911: __ mov(input_reg, zero_reg);
nit: I really do like to see comment "// In delay slot." when there is room to
put it there. Just an extra reminder when we are reading the code. 3 places.

[palfia, 2013-10-11 23:31:07]

https://codereview.chromium.org/26742006/diff/12001/src/mips/lithium-codegen-...
File src/mips/lithium-codegen-mips.cc (right):

https://codereview.chromium.org/26742006/diff/12001/src/mips/lithium-codegen-...
src/mips/lithium-codegen-mips.cc:4911: __ mov(input_reg, zero_reg);
On 2013/10/11 23:26:14, Paul Lind wrote:
> nit: I really do like to see comment "// In delay slot." when there is room to
> put it there. Just an extra reminder when we are reading the code. 3 places.

Done.

[Paul Lind, 2013-10-11 23:32:08]

LGTM, thanks for accommodating my pickiness!

[palfia, 2013-10-11 23:38:05]

Committed patchset #3 manually as r17173.