OpenE2K/gcc
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

re PR tree-optimization/33301 (wrong vectorization factor due to an invariant type-promotion in the loop)

Browse Source 
PR tree-optimization/33301
        * tree-vect-analyze (analyze_operations): Look at the type of the rhs
        when relevant.

From-SVN: r128265
This commit is contained in:
Dorit Nuzman 2007-09-08 09:19:39 +00:00 committed by Dorit Nuzman
parent f1979d7e89
commit dab1260138
4 changed files with 46 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
gcc/ChangeLog
View File

@ -1,3 +1,9 @@
2007-09-08 Dorit Nuzman <dorit@il.ibm.com>
PR tree-optimization/33301
* tree-vect-analyze (analyze_operations): Look at the type of the rhs
when relevant.
2007-09-07 Zdenek Dvorak <ook@ucw.cz>
PR tree-optimization/32183

5
gcc/testsuite/ChangeLog
View File

@ -1,3 +1,8 @@
2007-09-08 Dorit Nuzman <dorit@il.ibm.com>
PR tree-optimization/33301
* gfortran.dg/vect/pr33301.f: New test.
2007-09-07 Jerry DeLisle <jvdelisle@gcc.gnu.org>
PR libfortran/33307

14
gcc/testsuite/gfortran.dg/vect/pr33301.f Normal file
View File

@ -0,0 +1,14 @@
c { dg-do compile }
C Derived from lapack
SUBROUTINE ZGELSX( M, N, NRHS, A, LDA, B, LDB, JPVT, RCOND, RANK,
$ WORK, RWORK, INFO )
COMPLEX(kind=8) WORK( * )
c Following declaration added on transfer to gfortran testsuite.
c It is present in original lapack source
integer rank
DO 20 I = 1, RANK
WORK( ISMAX+I-1 ) = S2*WORK( ISMAX+I-1 )
20 CONTINUE
END
c { dg-final { cleanup-tree-dump "vect" } }

25
gcc/tree-vect-analyze.c
View File

@ -216,21 +216,38 @@ vect_determine_vectorization_factor (loop_vec_info loop_vinfo)
}
else
{
tree operation;
gcc_assert (! STMT_VINFO_DATA_REF (stmt_info)
&& !is_pattern_stmt_p (stmt_info));
/* We set the vectype according to the type of the result (lhs).
/* We generally set the vectype according to the type of the
result (lhs).
For stmts whose result-type is different than the type of the
arguments (e.g. demotion, promotion), vectype will be reset
appropriately (later). Note that we have to visit the smallest
datatype in this function, because that determines the VF.
If the smallest datatype in the loop is present only as the
rhs of a promotion operation - we'd miss it here.
However, in such a case, that a variable of this datatype
does not appear in the lhs anywhere in the loop, it shouldn't
affect the vectorization factor. */
Such a case, where a variable of this datatype does not appear
in the lhs anywhere in the loop, can only occur if it's an
invariant: e.g.: 'int_x = (int) short_inv', which we'd expect
to have been optimized away by invariant motion. However, we
cannot rely on invariant motion to always take invariants out
of the loop, and so in the case of promotion we also have to
check the rhs. */
scalar_type = TREE_TYPE (GIMPLE_STMT_OPERAND (stmt, 0));
operation = GIMPLE_STMT_OPERAND (stmt, 1);
if (TREE_CODE (operation) == NOP_EXPR
|| TREE_CODE (operation) == CONVERT_EXPR
|| TREE_CODE (operation) == WIDEN_MULT_EXPR)
{
tree rhs_type = TREE_TYPE (TREE_OPERAND (operation, 0));
if (TYPE_SIZE_UNIT (rhs_type) < TYPE_SIZE_UNIT (scalar_type))
scalar_type = TREE_TYPE (TREE_OPERAND (operation, 0));
}
if (vect_print_dump_info (REPORT_DETAILS))
{
fprintf (vect_dump, "get vectype for scalar type: ");