> New optimization flags and new params need documentation in
> gcc/doc/invoke.texi.
>
Thanks. Added description in invoke.texi. The patch is in trunk.
> The description of the --params suggest they provide fixed values - is
> there no way to autodetect sensible values with a cost-model? I
> ha
op_optimize_isl)
+ if (flag_loop_optimize_isl || flag_loop_unroll_jam)
transform_done |= optimize_isl (scop);
return transform_done;
@@ -327,6 +327,7 @@
pbb->schedule = NULL;
pbb->transformed = NULL;
pbb->saved = NULL;
+ pbb->map_sepclass = NULL;
PBB_SCOP (pbb) = sc
age -
> From: "Sven Verdoolaege"
> To: "Mircea Namolaru"
> Cc: gcc-patches@gcc.gnu.org, "Tobias Grosser" , "Richard
> Biener" ,
> "Albert Cohen"
> Sent: Wednesday, November 12, 2014 12:24:40 AM
> Subject: Re: [GRAPHITE, PATCH] L
.c, the separation class
option is set (using
the auxiliary map computed previously) and is added to the other AST build
options.
Mircea
2014-11-12 Mircea Namolaru
* common.opt (flag_loop_unroll_and_jam): New flag for unroll and jam.
* params.def (PARAM_LOOP_UNROLL_JAM_
3, 4) + 3 == HEIGHT && c0 + 3 == HEIGHT)
> > S_4(HEIGHT - 1, c1);
> > }
> > }
> >
> > The "separate" option (set by default for all dimensions for the new isl
> > based code generator)
> > don't succeed to remove the ifs from the loops and generate two loop
> > structures (this would
> > have been highly desirable).
> >
> > As the stage 1 is going to close soon, quick feedback to this patch is
> > greatly appreciated.
> > Many thanks, Mircea Namolaru
>
int
f1(int v[1024][1024], int HEIGHT, int LENGTH)
{
int i, j;
for (i=0; i
=
--- gcc/params.def (revision 217013)
+++ gcc/params.def (working copy)
@@ -847,6 +847,21 @@
"size of tiles for loop blocking",
51, 0, 0)
+/* Size of unrolling factor for unroll-and-jam. */
+
+DEFPARAM (PARAM_LOOP_UNROLL_JAM_SIZE,
+ "loop-unroll-jam-size&
or the new isl based
code generator)
don't succeed to remove the ifs from the loops and generate two loop structures
(this would
have been highly desirable).
As the stage 1 is going to close soon, quick feedback to this patch is greatly
appreciated.
Many thanks, Mircea Namolaru
2014-11-7
> On 08/03/14 17:44, Mircea Namolaru wrote:
> > 2014-08-03 Mircea Namolaru
> >
> > Replacement of isl-int by isl_val
> > * graphite-clast-to-gimple.c: include isl/val.h, isl/val_gmp.h
> > (compute_bounds_for_param): use isl_val instead of isl_int
clear (v);
+ c = isl_constraint_set_constant_val (c, v);
c = isl_constraint_set_coefficient_si (c, isl_dim_param, p, -1);
scop->context = isl_set_add_constraint (scop->context, c);
2014-08-03 Mircea Namolaru
Replacement of isl-int by isl_val
* graphite-clast-to-gimple.c:
main, dimension);
+ isl_val_get_num_gmp (isl_value, low);
+ isl_val_free (isl_value);
+ isl_value = isl_set_max_val (domain, dimension);
+ isl_val_get_num_gmp (isl_value, up);
+ isl_val_free (isl_value);
isl_set_free (domain);
isl_aff_free (dimension);
}
2014-07-06 Mircea Namolaru
future so you don't want to base the correctness of
the code on this assumption.
Mircea
- Original Message -
> From: "Roman Gareev"
> To: gcc-patches@gcc.gnu.org
> Cc: "Tobias Grosser" , "mircea namolaru"
>
> Sent: Monday, March 10, 2014
Hi,
Your patch is fine - even without this bug, introducing NULL pointer checks
before dereferencing a pointer is a good thing.
Mircea
Hi,
Thanks. Here is the updated patch.
2014-02-26 Tobias Grosser
Mircea Namolaru
PR tree-optimization/58028
* graphite-clast-to-gimple.c (set_cloog_options): Don't remove scalar
dimensions.
Index: gcc/graphite-clast-to-gim
Thanks for comments - updated the patch (fixed my e-mail address too :-)).
2014-02-26 Tobias Grosser
Mircea Namolaru
Fix for bug 58028
* graphite-clast-to-gimple.c (set_cloog_options):
Don't remove scalar dimensions.
Index: gcc/graphite-clast-to-gim
Patch for replacement of the isl_int (obsolete) by isl_val.
No regressions for c/c++/fortran on x86-64 Linux.2014-02-09 Mircea Namolaru
Replacement of isl-int by isl_val
* graphite-clast-to-gimple.c: include isl/val.h, isl/val_gmp.h
(compute_bounds_for_param): use
15 matches
Mail list logo
