https://gcc.gnu.org/bugzilla/show_bug.cgi?id=116974
Bug ID: 116974
Summary: omp inscan reduction not supported with SLP
Product: gcc
Version: 15.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: rguenth at gcc dot gnu.org
Target Milestone: ---
When forcing SLP vectorization with --param vect-force-slp=1 it shows that
vectorizable_scan_store isn't yet supported for SLP. This results in
FAIL: g++.dg/vect/simd-10.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-2.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-3.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-4.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-5.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-6.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-7.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-8.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
FAIL: g++.dg/vect/simd-9.cc -std=c++17 scan-tree-dump-times vect "vectorized
[1-3] loops" 2
but note that for example gcc.dg/vect/vect-simd-8.c also behaves differently,
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:17:5:
optimized: loop vectorized using 16 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:15:11:
optimized: loop vectorized using 32 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:15:11:
optimized: loop vectorized using 16 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:30:5:
optimized: loop vectorized using 16 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:28:11:
optimized: loop vectorized using 32 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:28:11:
optimized: loop vectorized using 16 byte vectors
vs.
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:15:11:
optimized: loop vectorized using 32 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:15:11:
optimized: loop vectorized using 16 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:28:11:
optimized: loop vectorized using 32 byte vectors
/home/rguenther/src/gcc/gcc/testsuite/gcc.dg/vect/vect-simd-8.c:28:11:
optimized: loop vectorized using 16 byte vectors
note the lack of vectorization of the main loop (we only vectorize the
initialization loop). The testcase doesn't notice because it scans for
"vectorized \[1-3] loops"
