[llvm] r177730 - R600: Use legacy (0 * anything = 0) MUL instructions for pow intrinsics

[Rafael Espíndola]

> I know the policy, but it still doesn't make much sense for us covering the
> same code by two different unit testing frameworks.

If you think the policy should be changed, please propose that. Don't
just ignore it.

I don't think it should be changed. As Justin noticed, only a minority
of llvm developers is working on mesa, so whatever tests you have
there are invisible to us. In addition to that, a .ll to assembly test
has some advantages that make the redundancy worth it:

* They are fast.
* They can be run on any platform.
* They make it obvious what they are testing. Consider the test that
was modified in r177736. If I regress this, it is obvious what the
problem is (wrong multiply instruction). Without it, I would get
"wrong lighting in some corner cases" which would require a debug
session to figure out what the problem is.

Given these advantages we normally add tests to llvm/test even when
the problem was catch in llvm's own test-suite!

> Christian.

Cheers,
Rafael