[PATCH] Allow casts of mixed-size same-arity vectors

[Stephen Canon]

Hal —

Am I understanding this correctly:

	vector4float x;
	vector4int16 y = (vector4int16)x; // this is a conversion
	vector4int32 z = (vector4int32)x; // this is a bitcast

This seems confusing in the extreme.

Thanks,
– Steve

On Sep 11, 2013, at 12:04 PM, Hal Finkel <hfinkel at anl.gov> wrote:

> Hello,
> 
> Please review the attached patch which adds support for casting between vector types with different sizes (but the same number of vector elements). LLVM supports these conversions for the underlying LLVM vector types (just as for the scalar types), and this adds support in Clang, requiring minor modifications in Sema and CodeGen.
> 
> For example, this patch allows the following to work:
> 
> typedef double vector4double __attribute__((__vector_size__(32)));
> typedef float  vector4float  __attribute__((__vector_size__(16)));
> 
> vector4float flt_trunc(vector4double x) {
>  return (vector4float) x;
> }
> 
> On some targets the code generation for these conversions has been optimized (autovectorization can produce these conversions), and accessing these capabilities from Clang is important, for example, when writing intrinsics headers (which, when possible, we prefer to implement in terms of generic IR).
> 
> Thanks in advance,
> Hal
> 
> -- 
> Hal Finkel
> Assistant Computational Scientist
> Leadership Computing Facility
> Argonne National Laboratory
> <mixed-size-vec-casts.patch>_______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits