MIPS: Enable Math rounding operations call reduction optimization in TF.

MIPS: Enable Math rounding operations call reduction optimization in TF.

TEST=
BUG=
R=paul.lind@imgtec.com

Committed: https://chromium.googlesource.com/v8/v8/+/8c4b7d272ac71340394b96835255fee45c227aea

[dusmil.imgtec, 2014-12-18 14:48:05]

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

[dusmil.imgtec, 2014-12-18 15:39:53]

PTAL

[paul.l..., 2014-12-18 18:07:41]

https://codereview.chromium.org/791473003/diff/40001/src/compiler/mips/code-g...
File src/compiler/mips/code-generator-mips.cc (right):

https://codereview.chromium.org/791473003/diff/40001/src/compiler/mips/code-g...
src/compiler/mips/code-generator-mips.cc:289: __ Branch(USE_DELAY_SLOT, &done,
hs, at, Operand(1075));               \
Hmmm, would prefer we don't use so many magic numbers here. There are constants:
HeapNumber::kExponentShift, HeapNumber::kExponentBits, and your
1075 = HeapNumber::kExponentBias + HeapNumber::kMantissaBits

Even with those defs, it is probably worth a comment to mention that there is no
fractional part when the exponent is > number of mantissa bits, so your intent
is clear.

Same for mips64 version of this.

[dusmil.imgtec, 2014-12-22 14:34:49]

https://codereview.chromium.org/791473003/diff/40001/src/compiler/mips/code-g...
File src/compiler/mips/code-generator-mips.cc (right):

https://codereview.chromium.org/791473003/diff/40001/src/compiler/mips/code-g...
src/compiler/mips/code-generator-mips.cc:289: __ Branch(USE_DELAY_SLOT, &done,
hs, at, Operand(1075));               \
On 2014/12/18 18:07:41, paul.l... wrote:
> Hmmm, would prefer we don't use so many magic numbers here. There are
constants:
> HeapNumber::kExponentShift, HeapNumber::kExponentBits, and your
> 1075 = HeapNumber::kExponentBias + HeapNumber::kMantissaBits
> 
> Even with those defs, it is probably worth a comment to mention that there is
no
> fractional part when the exponent is > number of mantissa bits, so your intent
> is clear.
> 
> Same for mips64 version of this.

Done.

[paul.l..., 2014-12-22 16:40:35]

LGTM, thanks!

[commit-bot: I haz the power, 2014-12-22 17:37:28]

Patchset 4 (id:??) landed as
https://crrev.com/8c4b7d272ac71340394b96835255fee45c227aea
Cr-Commit-Position: refs/heads/master@{#25931}

[dusmil.imgtec, 2014-12-22 17:37:28]

Committed patchset #4 (id:60001) manually as
8c4b7d272ac71340394b96835255fee45c227aea (presubmit successful).