[llvm-commits] [llvm] r73009 - in /llvm/trunk: lib/Target/PowerPC/PPCInstrAltivec.td test/CodeGen/PowerPC/vec_shift.ll

[Eli Friedman]

On Fri, Jun 12, 2009 at 4:44 PM, Chris Lattner<clattner at apple.com> wrote:
>
> On Jun 6, 2009, at 6:07 PM, Eli Friedman wrote:
>
>> Author: efriedma
>> Date: Sat Jun  6 20:07:55 2009
>> New Revision: 73009
>>
>> URL: http://llvm.org/viewvc/llvm-project?rev=73009&view=rev
>> Log:
>> PR3628: Add patterns to match SHL/SRL/SRA to the corresponding Altivec
>> instructions.
>
> Nice.  Are there altivec intrinsics for these?  Can we zap the
> intrinsics and have llvm-gcc lower to the LLVM IR node?

The intrinsics are int_ppc_altivec_vslb/h/w, int_ppc_altivec_vsrb/h/w,
and int_ppc_altivec_vsrab/h/w.  We could zap the intrinsics; I don't
really want to write the patch, though, because I couldn't test it
properly.

-Eli