tweak of Math.abs in its x64 stub

tweak of Math.abs in its x64 stub

[haitao.feng, 2013-08-06 07:49:06]

Why not introduce SmiMathAbs in the X64 macro assembler?

[Jakob Kummerow, 2013-08-06 12:56:17]

Unifying lithium codegen and the stub sounds reasonable. Maybe you can do a
microbenchmark to figure out whether the branchless stub code or the conditional
"neg" in lithium is faster in practice.

[weiliang.lin2, 2013-08-06 13:05:48]

On 2013/08/06 12:56:17, Jakob wrote:
> Unifying lithium codegen and the stub sounds reasonable. Maybe you can do a
> microbenchmark to figure out whether the branchless stub code or the
conditional
> "neg" in lithium is faster in practice.

Actually I did an experiment before. The branchless stub code was faster than
the conditional "neg". One of my previous idea is to port branchless code to
lithium. The issue is that LMathAbs only has one available register. To use the
branchless code, you need to change the LMathAbs  instruction to assign a
temporary register.

[Jakob Kummerow, 2013-08-07 16:43:47]

In light of the discussion, LGTM.

Haitao, could you please land this?

[haitao.feng, 2013-08-08 01:04:55]

Sure. Weiliang, could you "add a comment plus a URL describing the bit twiddling
hack" as patchset 2 as Sven suggested?

[weiliang.lin2, 2013-08-08 01:19:53]

On 2013/08/08 01:04:55, haitao.feng wrote:
> Sure. Weiliang, could you "add a comment plus a URL describing the bit
twiddling
> hack" as patchset 2 as Sven suggested?

[haitao.feng, 2013-08-08 02:04:53]

I have committed this Cl but it is weird this issue is not closed. I have
inputed the right coderview password but the dcommit does not think so.
Weiliang, could you close it manually? Sorry for the inconvenience.