SRA tends to use VIEW_CONVERT_EXPR when replacing bool fields with unsigned char fields. Those are not handled in vector bool pattern detection causing vector true values to leak. The following fixes this by turning those into b ? 1 : 0 as well.
Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed to trunk. 2020-10-01 Richard Biener <rguent...@suse.de> * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle VIEW_CONVERT_EXPR. * g++.dg/vect/pr97255.cc: New testcase. --- gcc/testsuite/g++.dg/vect/pr97255.cc | 44 ++++++++++++++++++++++++++++ gcc/tree-vect-patterns.c | 8 +++-- 2 files changed, 50 insertions(+), 2 deletions(-) create mode 100644 gcc/testsuite/g++.dg/vect/pr97255.cc diff --git a/gcc/testsuite/g++.dg/vect/pr97255.cc b/gcc/testsuite/g++.dg/vect/pr97255.cc new file mode 100644 index 00000000000..efb7f53fd27 --- /dev/null +++ b/gcc/testsuite/g++.dg/vect/pr97255.cc @@ -0,0 +1,44 @@ +// { dg-require-effective-target c++11 } +// { dg-additional-options "-O3" } + +template<class T, unsigned N> +class Array{ +public: + T& operator[](unsigned x) {return m_arr[x];} +private: + T m_arr[N]; +}; + +int +__attribute__((noipa)) +logicalOr(Array< char, 4 > in1[60], + Array< bool, 4 > out[60]) +{ + for (unsigned k0 = 0u; k0 < 60u; ++k0) { + Array< char, 4 > in1m = in1[k0]; + Array< bool, 4 > x; + for (unsigned k1 = 0u; k1 < 4u; ++k1) { + char in1s = in1m[k1]; + x[k1] = in1s != char(0) || in1s != char(0); + } + out[k0] = x; + } + return out[0][0]; +} + + +int main() +{ + Array< char, 4 > In1[60]{}; + Array< bool, 4 > Out7[60]{}; + + for( int i = 0; i < 60; ++i){ + for( int j = 0; j < 4; ++j){ + In1[i][j] = 240 - i*4 - j; + } + } + + if (logicalOr(In1, Out7) != 1) + __builtin_abort (); + return 0; +} diff --git a/gcc/tree-vect-patterns.c b/gcc/tree-vect-patterns.c index db45740da3c..d626c5f7362 100644 --- a/gcc/tree-vect-patterns.c +++ b/gcc/tree-vect-patterns.c @@ -4028,14 +4028,18 @@ vect_recog_bool_pattern (vec_info *vinfo, var = gimple_assign_rhs1 (last_stmt); lhs = gimple_assign_lhs (last_stmt); + rhs_code = gimple_assign_rhs_code (last_stmt); + + if (rhs_code == VIEW_CONVERT_EXPR) + var = TREE_OPERAND (var, 0); if (!VECT_SCALAR_BOOLEAN_TYPE_P (TREE_TYPE (var))) return NULL; hash_set<gimple *> bool_stmts; - rhs_code = gimple_assign_rhs_code (last_stmt); - if (CONVERT_EXPR_CODE_P (rhs_code)) + if (CONVERT_EXPR_CODE_P (rhs_code) + || rhs_code == VIEW_CONVERT_EXPR) { if (! INTEGRAL_TYPE_P (TREE_TYPE (lhs)) || TYPE_PRECISION (TREE_TYPE (lhs)) == 1) -- 2.26.2