MIPS64: Fix TruncatingDiv.

MIPS64: Fix TruncatingDiv.

- Fix the Macroassembler::TruncatingDiv to compute
  the correct result even when the magic number is negative.
- Fix the MULT instruction in the simulator.

TEST=mjsunit/compiler/division-by-constant
BUG=
R=dusan.milosavljevic@imgtec.com

Committed: https://code.google.com/p/v8/source/detail?r=24475

[akos.palfi.imgtec, 2014-10-08 08:34:51]

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

[akos.palfi.imgtec, 2014-10-08 08:34:51]

PTAL.

[dusmil.imgtec, 2014-10-08 09:56:40]

https://codereview.chromium.org/637273002/diff/1/src/mips64/macro-assembler-m...
File src/mips64/macro-assembler-mips64.cc (right):

https://codereview.chromium.org/637273002/diff/1/src/mips64/macro-assembler-m...
src/mips64/macro-assembler-mips64.cc:6070:
base::MagicNumbersForDivision<uint32_t> mag =
Can we instantiate template class with int32_t parameter here?
Will it cause to much trouble to change this?

[Sven Panne, 2014-10-08 10:44:13]

svenpanne@chromium.org changed reviewers:
+ svenpanne@chromium.org

[Sven Panne, 2014-10-08 10:44:13]

https://codereview.chromium.org/637273002/diff/1/src/mips64/macro-assembler-m...
File src/mips64/macro-assembler-mips64.cc (right):

https://codereview.chromium.org/637273002/diff/1/src/mips64/macro-assembler-m...
src/mips64/macro-assembler-mips64.cc:6070:
base::MagicNumbersForDivision<uint32_t> mag =
On 2014/10/08 09:56:40, dusmil.imgtec wrote:
> Can we instantiate template class with int32_t parameter here?
> Will it cause to much trouble to change this?

Nope, the template parameter *must* be unsigned, otherwise all this stuff is
wrong. With full C++11 support I would have added

   static_assert(std::is_unsigned<T>, "T must be unsigned");

to base::MagicNumbersForDivision, but we're not there yet, so there's only a
comment. Furthermore, we only have only instantiations for uint32_t and
uint64_t, so you would get a linker error reminding you of that fact. :-)

[dusmil.imgtec, 2014-10-08 11:19:14]

Ok, thanks for clarifying this.

[dusmil.imgtec, 2014-10-08 11:19:35]

Lgtm.

[akos.palfi.imgtec, 2014-10-08 16:15:38]

Committed patchset #1 (id:1) manually as 24475 (presubmit successful).