Implement Math.tanh using fdlibm port.

Implement Math.tanh using fdlibm port.

Contributed by Raymond Toy: http://rtoy.github.io/fdlibm-js/

R=jkummerow@chromium.org
BUG=v8:3495
LOG=N

Committed: https://crrev.com/47c9e1c904050f82c4e527c5f86fc287856cdd3c
Cr-Commit-Position: refs/heads/master@{#31335}

[Jakob Kummerow, 2015-10-16 08:35:05]

lgtm

[Yang, 2015-10-16 09:54:56]

The CQ bit was checked by yangguo@chromium.org

[Yang, 2015-10-16 09:54:56]

The patchset sent to the CQ was uploaded after l-g-t-m from
jkummerow@chromium.org
Link to the patchset: https://codereview.chromium.org/1407213002/#ps20001
(title: "simplified some comments")

[commit-bot: I haz the power, 2015-10-16 09:55:05]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1407213002/20001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1407213002/20001

[commit-bot: I haz the power, 2015-10-16 09:57:25]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-10-16 09:57:25]

Try jobs failed on following builders:
  v8_linux_nodcheck_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux_nodcheck_rel/build...)
  v8_linux_rel on tryserver.v8 (JOB_FAILED,
http://build.chromium.org/p/tryserver.v8/builders/v8_linux_rel/builds/10619)

[Yang, 2015-10-16 11:40:47]

The CQ bit was checked by yangguo@chromium.org

[Yang, 2015-10-16 11:40:48]

The patchset sent to the CQ was uploaded after l-g-t-m from
jkummerow@chromium.org
Link to the patchset: https://codereview.chromium.org/1407213002/#ps40001
(title: "delete the actual old implementation")

[commit-bot: I haz the power, 2015-10-16 11:40:56]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1407213002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1407213002/40001

[commit-bot: I haz the power, 2015-10-16 12:56:23]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-10-16 12:56:43]

Patchset 3 (id:??) landed as
https://crrev.com/47c9e1c904050f82c4e527c5f86fc287856cdd3c
Cr-Commit-Position: refs/heads/master@{#31335}

[Raymond Toy, 2015-10-16 17:07:24]

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

[Raymond Toy, 2015-10-16 17:07:25]

Thanks so much for doing this!

https://codereview.chromium.org/1407213002/diff/40001/src/third_party/fdlibm/...
File src/third_party/fdlibm/fdlibm.js (right):

https://codereview.chromium.org/1407213002/diff/40001/src/third_party/fdlibm/...
src/third_party/fdlibm/fdlibm.js:869: define TWO_M55 =
2.77555756156289135105e-17;  // 2^-55, empty lower half
A bit late, but this comment is incorrect.  Since 2^-55 is a power of two, the
lower half is always empty.  The comment here implies to me that you've forcibly
made the lower half zero, but it was already zero anyway.

[Yang, 2015-10-16 18:06:40]

On 2015/10/16 17:07:25, Raymond Toy wrote:
> Thanks so much for doing this!
> 
>
https://codereview.chromium.org/1407213002/diff/40001/src/third_party/fdlibm/...
> File src/third_party/fdlibm/fdlibm.js (right):
> 
>
https://codereview.chromium.org/1407213002/diff/40001/src/third_party/fdlibm/...
> src/third_party/fdlibm/fdlibm.js:869: define TWO_M55 =
> 2.77555756156289135105e-17;  // 2^-55, empty lower half
> A bit late, but this comment is incorrect.  Since 2^-55 is a power of two, the
> lower half is always empty.  The comment here implies to me that you've
forcibly
> made the lower half zero, but it was already zero anyway.

What I meant is exactly that: the lower half of 2^-55 is already empty, which is
why we define it as number literal and not part of the constant array, as
loading the constant into memory on 32 bit systems is cheap. A little bit above
we have the same definition and comment for 2^-28.

[Raymond Toy, 2015-10-20 15:55:11]

On 2015/10/16 at 18:06:40, yangguo wrote:
> On 2015/10/16 17:07:25, Raymond Toy wrote:
> > Thanks so much for doing this!
> > 
> >
https://codereview.chromium.org/1407213002/diff/40001/src/third_party/fdlibm/...
> > File src/third_party/fdlibm/fdlibm.js (right):
> > 
> >
https://codereview.chromium.org/1407213002/diff/40001/src/third_party/fdlibm/...
> > src/third_party/fdlibm/fdlibm.js:869: define TWO_M55 =
> > 2.77555756156289135105e-17;  // 2^-55, empty lower half
> > A bit late, but this comment is incorrect.  Since 2^-55 is a power of two,
the
> > lower half is always empty.  The comment here implies to me that you've
forcibly
> > made the lower half zero, but it was already zero anyway.
> 
> What I meant is exactly that: the lower half of 2^-55 is already empty, which
is why we define it as number literal and not part of the constant array, as
loading the constant into memory on 32 bit systems is cheap. A little bit above
we have the same definition and comment for 2^-28.

I don't understand why the lower half being zero makes defining it as a literal
better.  But no worries. I don't have to know as long as it makes sense to you
and the reviewers.