procyberian/gcc
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

re PR tree-optimization/45764 (wrong code -O2 vs -O3 (problem in vectorizer???))

Browse source 
2010-10-21  Richard Guenther  <rguenther@suse.de>
	Michael Matz  <matz@suse.de>

	PR tree-optimization/45764
	* tree-vect-data-refs.c (vect_compute_data_ref_alignment):
	Adjust initial misalignment for negative DR_STEP.
	(vect_find_same_alignment_drs): Two DRs with different DR_STEP
	do not have the same alignment over the whole iteration domain.

        * gcc.dg/torture/pr45764.c: New testcase.

Co-Authored-By: Michael Matz <matz@suse.de>

From-SVN: r165768
This commit is contained in:
Richard Guenther 2010-10-21 11:10:41 +00:00 • committed by Richard Biener
parent 82c1834a6f
commit 46241ea9d1
4 changed files with 62 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

9
gcc/ChangeLog
View file

@ -1,3 +1,12 @@
2010-10-21 Richard Guenther <rguenther@suse.de>
Michael Matz <matz@suse.de>
PR tree-optimization/45764
* tree-vect-data-refs.c (vect_compute_data_ref_alignment):
Adjust initial misalignment for negative DR_STEP.
(vect_find_same_alignment_drs): Two DRs with different DR_STEP
do not have the same alignment over the whole iteration domain.
2010-10-21 Richard Guenther <rguenther@suse.de>
PR tree-optimization/46111

6
gcc/testsuite/ChangeLog
View file

@ -1,3 +1,9 @@
2010-10-21 Richard Guenther <rguenther@suse.de>
Michael Matz <matz@suse.de>
PR tree-optimization/45764
* gcc.dg/torture/pr45764.c: New testcase.
2010-10-21 Richard Guenther <rguenther@suse.de>
PR tree-optimization/46111

27
gcc/testsuite/gcc.dg/torture/pr45764.c Normal file
View file

@ -0,0 +1,27 @@
/* { dg-do run } */
int result[64][16];
int main()
{
double dbuf[1000] = {0.0};
int ibuf[900];
int i, j;
double d1 = 0.0;
double d2 = 0.0;
for (i = 0; i < 900; ++i) {
ibuf[i] = (int)(d2 - d1);
d1 += dbuf[i];
d2 += dbuf[i + 64];
}
for (i = 0; i < 64; ++i) {
for (j = 0; j < 8; ++j) {
result[i][ j] = ibuf[64 - i + 64 * j];
result[i][15 - j] = ibuf[ i + 64 * j];
}
}
return 0;
}

20
gcc/tree-vect-data-refs.c
View file

@ -900,6 +900,19 @@ vect_compute_data_ref_alignment (struct data_reference *dr)
|| (TREE_CODE (base) == VAR_DECL
&& DECL_ALIGN (base) >= TYPE_ALIGN (vectype)));
/* If this is a backward running DR then first access in the larger
vectype actually is N-1 elements before the address in the DR.
Adjust misalign accordingly. */
if (tree_int_cst_compare (DR_STEP (dr), size_zero_node) < 0)
{
tree offset = ssize_int (TYPE_VECTOR_SUBPARTS (vectype) - 1);
/* DR_STEP(dr) is the same as -TYPE_SIZE of the scalar type,
otherwise we wouldn't be here. */
offset = fold_build2 (MULT_EXPR, ssizetype, offset, DR_STEP (dr));
/* PLUS because DR_STEP was negative. */
misalign = size_binop (PLUS_EXPR, misalign, offset);
}
/* Modulo alignment. */
misalign = size_binop (FLOOR_MOD_EXPR, misalign, alignment);
@ -1934,6 +1947,13 @@ vect_find_same_alignment_drs (struct data_dependence_relation *ddr,
if (DDR_NUM_DIST_VECTS (ddr) == 0)
return;
/* Data-dependence analysis reports a distance vector of zero
for data-references that overlap only in the first iteration
but have different sign step (see PR45764).
So as a sanity check require equal DR_STEP. */
if (!operand_equal_p (DR_STEP (dra), DR_STEP (drb), 0))
return;
loop_depth = index_in_loop_nest (loop->num, DDR_LOOP_NEST (ddr));
FOR_EACH_VEC_ELT (lambda_vector, DDR_DIST_VECTS (ddr), i, dist_v)
{