The r16-4558-g1b387bd8978577 change added a check that does not
match what the commit message says. The following fixes this,
resolving the testcases.
Bootstrap and regtest running on x86_64-unknown-linux-gnu. I'm
trying to bootstrap and test on aarch64 as well.
PR tree-optimization/122474
* tree-vect-loop.cc (vectorizable_reduction): Adjust condition
guarding the check for .VEC_SHL_INSERT.
* gcc.target/aarch64/sve2/pr123053.c: New testcase.
* gcc.target/riscv/rvv/pr122474.c: Likewise.
---
.../gcc.target/aarch64/sve2/pr123053.c | 25 +++++++++++++++++++
gcc/testsuite/gcc.target/riscv/rvv/pr122474.c | 15 +++++++++++
gcc/tree-vect-loop.cc | 12 ++++++---
3 files changed, 48 insertions(+), 4 deletions(-)
create mode 100644 gcc/testsuite/gcc.target/aarch64/sve2/pr123053.c
create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/pr122474.c
diff --git a/gcc/testsuite/gcc.target/aarch64/sve2/pr123053.c
b/gcc/testsuite/gcc.target/aarch64/sve2/pr123053.c
new file mode 100644
index 00000000000..9f2fc438a62
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/sve2/pr123053.c
@@ -0,0 +1,25 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-O2 -march=armv9-a" } */
+
+#include <stdint.h>
+
+struct
+{
+ int32_t f3;
+ int64_t f4;
+} g_137, g_138;
+struct
+{
+ int8_t f0;
+} g_174;
+int32_t g_67;
+extern uint32_t g_179[];
+uint16_t func_71()
+{
+ for (; g_174.f0; g_174.f0 -= 1)
+ {
+ g_137.f3 = 0;
+ for (; g_137.f3 <= 4; g_137.f3 += 1)
+ for (g_67 = 3; g_67; g_67 -= 1) g_179[g_67] || (g_138.f4 = 0);
+ }
+}
diff --git a/gcc/testsuite/gcc.target/riscv/rvv/pr122474.c
b/gcc/testsuite/gcc.target/riscv/rvv/pr122474.c
new file mode 100644
index 00000000000..314b0f93d71
--- /dev/null
+++ b/gcc/testsuite/gcc.target/riscv/rvv/pr122474.c
@@ -0,0 +1,15 @@
+/* { dg-do compile } */
+/* { dg-options "-march=rv64gcv_zvl256b -O3 -fwrapv -w" } */
+
+_Bool a;
+void b(unsigned c[][8][8][8][8][8]) {
+ for (int d = 1; d; d += 3)
+ for (int e = 0; e < 11; e += 2)
+ for (short f = 0; f < 1; f = 30482)
+ for (short g = 0; g < 014; g++)
+ a = ({
+ int h = a;
+ int i = c[2][f][d][d][d][d] ^ c[g][g][1][2][g][g];
+ i ? h : i;
+ });
+}
diff --git a/gcc/tree-vect-loop.cc b/gcc/tree-vect-loop.cc
index aa59cd1a39d..d3bb788d866 100644
--- a/gcc/tree-vect-loop.cc
+++ b/gcc/tree-vect-loop.cc
@@ -7653,12 +7653,16 @@ vectorizable_reduction (loop_vec_info loop_vinfo,
/* For double reductions, and for SLP reductions with a neutral value,
we construct a variable-length initial vector by loading a vector
full of the neutral value and then shift-and-inserting the start
- values into the low-numbered elements. */
+ values into the low-numbered elements. This is however not needed
+ if the target can do the reduction using the new optabs, and the initial
+ reduction value matches the neutral value and we have one SLP lane
+ while not having a reduction chain. */
if ((double_reduc || neutral_op)
&& !nunits_out.is_constant ()
- && (SLP_TREE_LANES (slp_node) != 1 && !reduc_chain)
- && (!neutral_op
- || !operand_equal_p (neutral_op,
+ && !(SLP_TREE_LANES (slp_node) == 1
+ && !reduc_chain
+ && neutral_op
+ && operand_equal_p (neutral_op,
vect_phi_initial_value (reduc_def_phi)))
&& !direct_internal_fn_supported_p (IFN_VEC_SHL_INSERT,
vectype_out, OPTIMIZE_FOR_SPEED))
--
2.51.0
