This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] Add comparison operators for vector types

On Sat, May 30, 2009 at 11:39 AM, Richard Guenther
<> wrote:
> On Sat, May 30, 2009 at 6:17 AM, Chris Lattner <> wrote:
>> On May 29, 2009, at 6:06 PM, Andrew Pinski wrote:
>>> Hi,
>>> ?The C/C++ Language extension for the Cell defines the comparison
>>> operators for vectors. ?The result of these operators is a single
>>> boolean value representing the bitwise AND of all pairwise
>>> comparisons. ?This patch adds support for these operators and a
>>> testcase to test them for both C and C++.
>>> This patch was originally written by Trevor and updated by me for the
>>> tuplifications of tree-vect-generic.c.
>>> OK? Bootstrapped and tested on i686-linux-gnu with no regressions.
>> Is this something you want to do on the Cell alone, or for generic vectors
>> in general? ?Note that OpenCL defines vector comparisons differently than
>> this. ?If GCC aims to support it one day, it would useful to consider what
>> it wants. ?Cell is a particularly useful target for OpenCL.
> It is also a change from what tree.def defines for EQ_EXPR and NE_EXPR.
> You should at least update the docs there. ?Please also update the
> gimple (type) checker to check that the result of vector comparisons is
> of vector type (which would also catch old uses of EQ/NE_EXPR with
> the other suggested semantic).

Doh, I completely misunderstood how the comparison semantics is
(tree.def still needs to be updated though).  So for the other "natural"
semantics (producing vector bit-masks) we would need different
tree codes.  Are the unordered comparisons allowed/implemented
for floating point vectors?


> Richard.
>> -Chris

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]