On Mon, 29 Oct 2012, Jan Hubicka wrote:
> > > ICK ...
> > >
> > > Why not sth as simple as
> > >
> > > return num_ssa_operands (stmt, SSA_OP_USE);
> > >
> > > ? a[1][2] and b[2] really have the same cost, variable length
> > > objects have extra SSA operands in ARRAY_REF/COMPONENT_REF for
> > ICK ...
> >
> > Why not sth as simple as
> >
> > return num_ssa_operands (stmt, SSA_OP_USE);
> >
> > ? a[1][2] and b[2] really have the same cost, variable length
> > objects have extra SSA operands in ARRAY_REF/COMPONENT_REF for
> > the size. Thus, stmt cost somehow should reflect th
> ICK ...
>
> Why not sth as simple as
>
> return num_ssa_operands (stmt, SSA_OP_USE);
>
> ? a[1][2] and b[2] really have the same cost, variable length
> objects have extra SSA operands in ARRAY_REF/COMPONENT_REF for
> the size. Thus, stmt cost somehow should reflect the number
> of depe
On Thu, 25 Oct 2012, Jan Hubicka wrote:
> Hi,
> while looking into cunroll issues I noticed that we miss quite lot of
> important unrolling
> at -O2 for EON because we think it will increase code size. This is because
> we do not
> account the fact that making array index constant is good thing
On Thu, 25 Oct 2012, Jan Hubicka wrote:
> > Hmm, also in ASM/call/return? It will definitely make quite a fuzz into
> > the cost metric
> > by making a=b+c to have cost of 3 instead of 1 as it have now. I am not
> > 100% sure if
> > a+b should be more expensive than a+1.
> > I can give that a
On Thu, 25 Oct 2012, Jan Hubicka wrote:
> > > +/* For operands of load/stores estimate cost of the address computations
> > > + involved. */
> > > +
> > > +static int
> > > +estimate_operand_cost (tree op)
> > > +{
> > > + int cost = 0;
> > > + while (handled_component_p (op))
> > > +{
>
> Hmm, also in ASM/call/return? It will definitely make quite a fuzz into the
> cost metric
> by making a=b+c to have cost of 3 instead of 1 as it have now. I am not 100%
> sure if
> a+b should be more expensive than a+1.
> I can give that a try and we will see after re-tunning how well it work
> > +/* For operands of load/stores estimate cost of the address computations
> > + involved. */
> > +
> > +static int
> > +estimate_operand_cost (tree op)
> > +{
> > + int cost = 0;
> > + while (handled_component_p (op))
> > +{
> > + cost += estimate_ref_cost (op);
> > + op = TR
On Thu, 25 Oct 2012, Jan Hubicka wrote:
> Hi,
> while looking into cunroll issues I noticed that we miss quite lot of
> important unrolling
> at -O2 for EON because we think it will increase code size. This is because
> we do not
> account the fact that making array index constant is good thing
Hi,
while looking into cunroll issues I noticed that we miss quite lot of important
unrolling
at -O2 for EON because we think it will increase code size. This is because we
do not
account the fact that making array index constant is good thing.
This tracks back to the fact that estimate_num_ins
10 matches
Mail list logo
