[llvm-commits] [llvm] r124064 - in /llvm/trunk: lib/Transforms/Scalar/ScalarReplAggregates.cpp test/Transforms/ScalarRepl/vector_promote.ll

Chris Lattner clattner at apple.com
Sun Jan 23 18:33:07 PST 2011


On Jan 23, 2011, at 4:49 AM, Duncan Sands wrote:

> Hi Chris,
> 
>> remove an old hack that avoided creating MMX datatypes.  The
>> X86 backend has been fixed.
> 
> does this mean that the special MMX types are going to be removed from the IR?

The special MMX IR type is why it is safe to remove this.  Previously the mid-level optimizer would form a <2 x i32> and all hell would break loose :).  With the MMX type, <2 x i32> isn't mapped onto MMX registers by the x86 backend so this is safe.

-Chris



More information about the llvm-commits mailing list