[llvm-commits] [llvm] r150060 - in /llvm/trunk: include/llvm/IntrinsicsX86.td lib/Target/X86/X86ISelLowering.cpp lib/Target/X86/X86InstrSSE.td

[Eric Christopher]

On Feb 8, 2012, at 12:48 PM, David A. Greene <dag at cray.com> wrote:

> Eric Christopher <echristo at apple.com> writes:
> 
>>> Whether generic IR is a "win" isn't the primary issue here.  The issue
>>> is that the user wrote code using intrinsics and expects that exact code
>>> to be in the asm.  Whether or not that is the best performing code
>>> possible doesn't matter.  It's what they want and we have to respect it.
>>> Our customers demand it.
>> 
>> Then make the generic IR match what you think should happen. I see no win
>> for anyone with this.
> 
> There is no way to do that.  It's generic.  By definition passes can
> change it.
> 

Yes.

> You see no benefit even though I have directly stated that their
> existance and use is critically important to us?
> 

I disagree with it. "My customers expect a particular sequence of code generated for _mm_XXX_ZZZ" isn't a compiler problem, it's a frontend problem if you don't want the backend to optimize it.

-eric