> On Sat, Oct 6, 2012 at 11:34 AM, Jan Hubicka wrote:
> > Hi,
> > I benchmarked the patch moving loop header copying and it is quite
> > noticeable win.
> >
> > Some testsuite updating is needed. In many cases it is just because the
> > optimizations are now happening earlier.
> > There are howev
On Sat, Oct 6, 2012 at 11:34 AM, Jan Hubicka wrote:
> Hi,
> I benchmarked the patch moving loop header copying and it is quite noticeable
> win.
>
> Some testsuite updating is needed. In many cases it is just because the
> optimizations are now happening earlier.
> There are however few testusite
Hi,
I benchmarked the patch moving loop header copying and it is quite noticeable
win.
Some testsuite updating is needed. In many cases it is just because the
optimizations are now happening earlier.
There are however few testusite failures I have torubles to deal with:
./testsuite/gcc/gcc.sum:FA
Hi,
this is the udpated patch I comitted after testing. I suppose we will need to
find way
to make SOC smaller for simple loops - it is way too overestimated currently.
* tree-vectorizer.h (vect_estimate_min_profitable_iters): Remove.
* tree-vect-loop.c (vect_estimate_min_profita
On Thu, 4 Oct 2012, Jan Hubicka wrote:
> > > So SOC cancels out in the runtime check.
> > > I still think we need two formulas - one determining if vectorization is
> > > profitable, other specifying the threshold for scalar path at runtime
> > > (that
> > > will generally give lower values).
> >
> > So SOC cancels out in the runtime check.
> > I still think we need two formulas - one determining if vectorization is
> > profitable, other specifying the threshold for scalar path at runtime (that
> > will generally give lower values).
>
> True, we want two values. But part of the scalar pat
On Mon, 1 Oct 2012, Jan Hubicka wrote:
> > >
> > > So the unvectorized cost is
> > > SIC * niters
> > >
> > > The vectorized path is
> > > SOC + VIC * ((niters-PL_ITERS-EP_ITERS)/VF) + VOC
> > > The scalar path of vectorizer loop is
> > > SIC * niters + SOC
> >
> >
> >
> > So the unvectorized cost is
> > SIC * niters
> >
> > The vectorized path is
> > SOC + VIC * ((niters-PL_ITERS-EP_ITERS)/VF) + VOC
> > The scalar path of vectorizer loop is
> > SIC * niters + SOC
>
> Note that 'th' is used for the runtime profitability check
On Sun, 30 Sep 2012, Jan Hubicka wrote:
> Hi,
> the point of the following patch is to make vectorizer to not vectorize the
> following testcase with profile feedback:
>
> int a[1];
> int i=5;
> int k=2;
> int val;
> __attribute__ ((noinline,noclone))
> test()
> {
> int j;
> for(j
Hi,
the point of the following patch is to make vectorizer to not vectorize the
following testcase with profile feedback:
int a[1];
int i=5;
int k=2;
int val;
__attribute__ ((noinline,noclone))
test()
{
int j;
for(j=0;j VIC * ((niters-PL_ITERS-EP_ITERS)/VF) + VOC(A)
where
10 matches
Mail list logo
