This is the mail archive of the
mailing list for the GCC project.
Re: [PATCH] [testsuite+spu] enable vectorizer testuite for the SPU
- From: Dorit Nuzman <DORIT at il dot ibm dot com>
- To: Dorit Nuzman <DORIT at il dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org, Andrew Pinski <pinskia at gmail dot com>, trevor_smigiel at playstation dot sony dot com
- Date: Sun, 3 Dec 2006 23:13:30 +0200
- Subject: Re: [PATCH] [testsuite+spu] enable vectorizer testuite for the SPU
With Janis's general approval (
http://gcc.gnu.org/ml/gcc-patches/2006-12/msg00074.html) for the testuite
part, and Andrew's approval for the spu bits, I went ahead and committed
the patch. In the patch I committed I also enabled the vectorizer tests in
g++.dg (which I forgot to include in the original patch I posted).
Together with the patch in
http://gcc.gnu.org/ml/gcc-patches/2006-12/msg00017.html (committed earlier
today), and with this patch by Trevor -
http://gcc.gnu.org/ml/gcc-patches/2006-12/msg00065.html (in particular the
new subv16qi3 and negv16qi2 patterns - thanks!), all the vectorizer
testcases are passing,
except for the vect-pow-2 testcase which is not a target specific problem,
and should be solved by this patch:
Next on my list is to support the mask-for-load pattern to handle
And then gradually provide the rest of the missing vector idioms (for
widening multiplication, packing/unpacking, etc).
> Andrew Pinski <firstname.lastname@example.org> wrote on 01/12/2006 11:41:53:
> > On Fri, 2006-12-01 at 11:27 +0200, Dorit Nuzman wrote:
> > >
> > > Index: config/spu/spu.md
> > > ===================================================================
> > This part is OK with the appropriate changelog.
> oops. here's the missing changelog entries:
> * config/spu/spu.md (smaxv4sf3, sminv4sf3): New.
> * gcc.dg/vect/vect.exp: Enable tests to run on spu.
> * gcc.dg/vect/no-section-anchors-vect-69.c: Reduce array size.
> * gcc.dg/vect/vect-70.c: Reduce array size.
> * gcc.dg/vect/fast-math-vect-reduc-7.c: Xfail for
> vect_no_compaer_double targets.
> * lib/target-supports.exp: Add spu to vect_no_align,
> vect_float, vect_double, and vect_int.
> > For:
> > > (5) testcases that require double-precision-fp conditional
> > > I
> > > added a keyword for targets that don't support this, and xfailed this
> > > test
> > > for those targets:
> > >
> > > FAIL: gcc.dg/vect/fast-math-vect-reduc-7.c scan-tree-dump-times
> > > vectorized
> > > 3 loops 1
> > There is a vector version of the compare function that is defined by
> > "SIMD Math library for the Cell BEA" so we should be able to use that
> > later on.
> > Thanks,
> > Andrew Pinski