Fold vector to scalar optimization

[Artem Shinkarov]

This patch was inspired by the previous patch I was working on
http://gcc.gnu.org/ml/gcc-patches/2010-06/msg01511.html and it should
be commited to the trunk only when vector shifting patch would be
commited. In other case the tests would not work.

The patch itself converts an operation "vector1 >> vector2" to
"vector1 >> scalar" if all the elements of the vector2 are the same.
At the moment the patch would not catch chains of variable assignmenrs
like:

b = x;
a = b;
c = a;
vec1 >> {x,a,b,c}

because the vector lowering is happening too early, though the
optimization could be moved down later.

ChangeLog:

2010-06-18  Artem Shinkarov <artyom.shinakroff@gmail.com>

      * passes.c (init_optimization_passes): Move veclower after SSA.
      * tree-vect-generic.c (expand_vector_operations_1): Handle conversion.

      testsuite/
      * gcc.dg/vec-scal-opt.c: New testcase.
      * gcc.dg/vec-scal-opt1.c: New testcase.
      * gcc.dg/vec-scal-opt2.c: New testcase.


bootstrapped and tested on x86_64_unknown-linux

OK?



-- 
Artem Shinkarov
-------------- next part --------------
A non-text attachment was scrubbed...
Name: vec-scal-opt.diff
Type: text/x-diff
Size: 6362 bytes
Desc: not available
URL: <http://gcc.gnu.org/pipermail/gcc-patches/attachments/20100618/a98d3e58/attachment.bin>