diff options
| author | Richard Biener <rguenther@suse.de> | 2023-07-31 14:44:52 +0200 |
|---|---|---|
| committer | Richard Biener <rguenther@suse.de> | 2023-08-03 14:52:11 +0200 |
| commit | 29370f1387274ad5a35a020db6a5d06c0324e6c1 (patch) | |
| tree | e4810a4d72a3a5f89e55383039c607206261131f /gcc/tree-vect-patterns.cc | |
| parent | 13dfb01e5c30c3bd09333ac79d6ff96a617fea67 (diff) | |
| download | gcc-29370f1387274ad5a35a020db6a5d06c0324e6c1.zip gcc-29370f1387274ad5a35a020db6a5d06c0324e6c1.tar.gz gcc-29370f1387274ad5a35a020db6a5d06c0324e6c1.tar.bz2 | |
tree-optimization/110838 - vectorization of widened shifts
The following makes sure to limit the shift operand when vectorizing
(short)((int)x >> 31) via (short)x >> 31 as the out of bounds shift
operand otherwise invokes undefined behavior. When we determine
whether we can demote the operand we know we at most shift in the
sign bit so we can adjust the shift amount.
Note this has the possibility of un-CSEing common shift operands
as there's no good way to share pattern stmts between patterns.
We'd have to separately pattern recognize the definition.
PR tree-optimization/110838
* tree-vect-patterns.cc (vect_recog_over_widening_pattern):
Adjust the shift operand of RSHIFT_EXPRs.
* gcc.dg/torture/pr110838.c: New testcase.
Diffstat (limited to 'gcc/tree-vect-patterns.cc')
| -rw-r--r-- | gcc/tree-vect-patterns.cc | 24 |
1 files changed, 24 insertions, 0 deletions
diff --git a/gcc/tree-vect-patterns.cc b/gcc/tree-vect-patterns.cc index ef806e2..e4ab8c2 100644 --- a/gcc/tree-vect-patterns.cc +++ b/gcc/tree-vect-patterns.cc @@ -3099,9 +3099,33 @@ vect_recog_over_widening_pattern (vec_info *vinfo, tree ops[3] = {}; for (unsigned int i = 1; i < first_op; ++i) ops[i - 1] = gimple_op (last_stmt, i); + /* For right shifts limit the shift operand. */ vect_convert_inputs (vinfo, last_stmt_info, nops, &ops[first_op - 1], op_type, &unprom[0], op_vectype); + /* Limit shift operands. */ + if (code == RSHIFT_EXPR) + { + wide_int min_value, max_value; + if (TREE_CODE (ops[1]) == INTEGER_CST) + ops[1] = wide_int_to_tree (op_type, + wi::bit_and (wi::to_wide (ops[1]), + new_precision - 1)); + else if (!vect_get_range_info (ops[1], &min_value, &max_value) + || wi::ge_p (max_value, new_precision, TYPE_SIGN (op_type))) + { + /* ??? Note the following bad for SLP as that only supports + same argument widened shifts and it un-CSEs same arguments. */ + tree new_var = vect_recog_temp_ssa_var (op_type, NULL); + gimple *pattern_stmt + = gimple_build_assign (new_var, BIT_AND_EXPR, ops[1], + build_int_cst (op_type, new_precision - 1)); + ops[1] = new_var; + gimple_set_location (pattern_stmt, gimple_location (last_stmt)); + append_pattern_def_seq (vinfo, last_stmt_info, pattern_stmt); + } + } + /* Use the operation to produce a result of type OP_TYPE. */ tree new_var = vect_recog_temp_ssa_var (op_type, NULL); gimple *pattern_stmt = gimple_build_assign (new_var, code, |