Use TRUNC.W instead of CVT.W on mips to convert from double to int as to not

Use TRUNC.W instead of CVT.W on mips to convert from double to int as to not
depend on the current rounding mode (fixes issue 25900).
Add support for TRUNC.W in assembler, disassembler, and simulator.
Fix disassembler to understand CVT.S.
Remove unused instructions.
Update co19 status file.

R=rmacnak@google.com

Committed: https://github.com/dart-lang/sdk/commit/6219753e18b7bbf0df8554012492dd475461acad

[regis, 2016-03-03 20:38:25]

regis@google.com changed reviewers:
+ rmacnak@google.com

[rmacnak, 2016-03-03 22:12:03]

Looking good on the hardware so far, but it will take a few hours to finish the
test suite.

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/assembler_mips_t...
File runtime/vm/assembler_mips_test.cc (right):

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/assembler_mips_t...
runtime/vm/assembler_mips_test.cc:2008: 
Add one of the inf/nan/overflow cases.

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/simulator_mips.cc
File runtime/vm/simulator_mips.cc (right):

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/simulator_mips.c...
runtime/vm/simulator_mips.cc:1896: if (isnan(fs_dbl) || isinf(fs_dbl) || (fs_dbl
> INT_MAX) ||
INT_MAX isn't necessarily kInt32Max (although ILP64 systems are rare)

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/simulator_mips.c...
runtime/vm/simulator_mips.cc:1898: fs_int = INT_MIN;
kMaxInt32

"Otherwise, the default result, 2^31–1, is written to fd"

[regis, 2016-03-03 23:14:48]

Thanks!

PTAL

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/assembler_mips_t...
File runtime/vm/assembler_mips_test.cc (right):

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/assembler_mips_t...
runtime/vm/assembler_mips_test.cc:2008: 
On 2016/03/03 22:12:02, rmacnak wrote:
> Add one of the inf/nan/overflow cases.

Done.

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/simulator_mips.cc
File runtime/vm/simulator_mips.cc (right):

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/simulator_mips.c...
runtime/vm/simulator_mips.cc:1896: if (isnan(fs_dbl) || isinf(fs_dbl) || (fs_dbl
> INT_MAX) ||
On 2016/03/03 22:12:03, rmacnak wrote:
> INT_MAX isn't necessarily kInt32Max (although ILP64 systems are rare)

OK, changed to kMaxInt32 and kMinInt32. But note that we will have many other
locations to fix if we use an ILP64 platform one day.

https://codereview.chromium.org/1765623002/diff/1/runtime/vm/simulator_mips.c...
runtime/vm/simulator_mips.cc:1898: fs_int = INT_MIN;
On 2016/03/03 22:12:02, rmacnak wrote:
> kMaxInt32
> 
> "Otherwise, the default result, 2^31–1, is written to fd"

Good catch. Fixed. I had simply recycled the existing code for CVT.W, which was
also wrong.

[rmacnak, 2016-03-04 00:52:06]

LGTM

[regis, 2016-03-04 01:05:12]

Description was changed from

==========
Use TRUNC.W instead of CVT.W on mips to convert from double to int as to not
depend on the current rounding mode (fixes issue 25900).
Add support for TRUNC.W in assembler, disassembler, and simulator.
Fix disassembler to understand CVT.S.
Remove unused instructions.
Update co19 status file.
==========

to

==========
Use TRUNC.W instead of CVT.W on mips to convert from double to int as to not
depend on the current rounding mode (fixes issue 25900).
Add support for TRUNC.W in assembler, disassembler, and simulator.
Fix disassembler to understand CVT.S.
Remove unused instructions.
Update co19 status file.

R=rmacnak@google.com

Committed:
https://github.com/dart-lang/sdk/commit/6219753e18b7bbf0df8554012492dd475461acad
==========

[regis, 2016-03-04 01:05:14]

Committed patchset #2 (id:20001) manually as
6219753e18b7bbf0df8554012492dd475461acad (presubmit successful).