Add frintp instruction.

Add FRINTP (round towards positive infinity) instruction.

Macro Assember, assembler and simulator for ARM64 were missing FRINTP.

R=rodolph.perfetta@arm.com, ulan@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=24800

[sigurds, 2014-10-21 15:33:39]

sigurds@chromium.org changed required reviewers:
+ rodolph.perfetta@arm.com, ulan@chromium.org

[sigurds, 2014-10-21 15:33:41]

Please take a look if this is OK. Please also advise on how to test properly.

[Rodolph Perfetta (ARM), 2014-10-21 19:22:00]

For tests, have a look at frinta or others in
test/cctest/test-assembler-arm64.cc, test/cctest/test-disasm-arm64.cc.

Out of curiosity, what is the use case in V8?

https://codereview.chromium.org/669923002/diff/1/src/arm64/simulator-arm64.cc
File src/arm64/simulator-arm64.cc (right):

https://codereview.chromium.org/669923002/diff/1/src/arm64/simulator-arm64.cc...
src/arm64/simulator-arm64.cc:2777: int_result = ceil(value);
Is ceil of a value in the range ]-1, -0] guaranteed to return -0 on all
supported platform?

[Sven Panne, 2014-10-22 06:34:39]

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

[Sven Panne, 2014-10-22 06:34:40]

https://codereview.chromium.org/669923002/diff/1/src/arm64/simulator-arm64.cc
File src/arm64/simulator-arm64.cc (right):

https://codereview.chromium.org/669923002/diff/1/src/arm64/simulator-arm64.cc...
src/arm64/simulator-arm64.cc:2777: int_result = ceil(value);
On 2014/10/21 19:22:00, Rodolph Perfetta (ARM) wrote:
> Is ceil of a value in the range ]-1, -0] guaranteed to return -0 on all
> supported platform?

According to http://stackoverflow.com/questions/5986875/c-ceil-and-negative-zero
it should, but I fail to find the quoted part of the C standard in my local
copy. :-/

[sigurds, 2014-10-22 09:53:30]

Thanks for the comments! I've added tests, including checking  correct behavior
on some values from the range ]-1, -0].

The use-case for V8 is to use frintp to implement Math.ceil in turbofan, which
currently uses a runtime call to implement Math.ceil. I hope to upload the
corresponding CL later this week. I'd like to add you as a reviewer for the
ARM64 parts, if that's ok.

https://codereview.chromium.org/669923002/diff/1/src/arm64/simulator-arm64.cc
File src/arm64/simulator-arm64.cc (right):

https://codereview.chromium.org/669923002/diff/1/src/arm64/simulator-arm64.cc...
src/arm64/simulator-arm64.cc:2777: int_result = ceil(value);
On 2014/10/22 06:34:40, Sven Panne wrote:
> On 2014/10/21 19:22:00, Rodolph Perfetta (ARM) wrote:
> > Is ceil of a value in the range ]-1, -0] guaranteed to return -0 on all
> > supported platform?
> 
> According to
http://stackoverflow.com/questions/5986875/c-ceil-and-negative-zero
> it should, but I fail to find the quoted part of the C standard in my local
> copy. :-/

F.9.6.1 of the standard defines ceil(). Annex F begins with stating that 
"An implementation that defines __STDC_IEC_559__ shall conform to the
specifications in this annex. Where a binding between the C language and IEC
60559 is indicated, the IEC 60559-specified behavior is adopted by reference,
unless stated otherwise."

I interpret this to mean: Every implementation defining __STDC_IEC_559__
implements IEEE behavior for ceil.

As Sven pointed out in personal discussion, if ceil would not conform to IEEE,
then we'd have problem in the implementation of DoubleToInteger in
conversions-inl.h:78, where we assume IEEE behavior.

[Rodolph Perfetta (ARM), 2014-10-22 10:16:11]

LGTM.

I am happy to review your incoming CL.

[ulan, 2014-10-22 10:27:27]

lgtm

[sigurds, 2014-10-22 11:06:36]

Committed patchset #3 (id:40001) manually as 24800 (presubmit successful).