* [pushed] aarch64: Make sure the UF divides the VF [PR105254]
@ 2022-04-13 16:54 Richard Sandiford
0 siblings, 0 replies; only message in thread
From: Richard Sandiford @ 2022-04-13 16:54 UTC (permalink / raw)
To: gcc-patches
In this PR, we were trying to set the unroll factor to a value higher
than the minimum VF (or more specifically, to a value that doesn't
divide the VF). I guess there are two approaches to this: let the
target pick any value it likes and make target-independent code pare
it back to something that makes sense, or require targets to supply
sensible values from the outset. This patch goes for the latter
approach.
Tested on aarch64-linux-gnu & pushed.
Richard
gcc/
PR tree-optimization/105254
* config/aarch64/aarch64.cc
(aarch64_vector_costs::determine_suggested_unroll_factor): Take a
loop_vec_info as argument. Restrict the unroll factor to values
that divide the VF.
(aarch64_vector_costs::finish_cost): Update call accordingly.
gcc/testsuite/
PR tree-optimization/105254
* g++.dg/vect/pr105254.cc: New test.
---
gcc/config/aarch64/aarch64.cc | 12 ++++++++----
gcc/testsuite/g++.dg/vect/pr105254.cc | 26 ++++++++++++++++++++++++++
2 files changed, 34 insertions(+), 4 deletions(-)
create mode 100644 gcc/testsuite/g++.dg/vect/pr105254.cc
diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
index cf62bdd023b..f650abbc4ce 100644
--- a/gcc/config/aarch64/aarch64.cc
+++ b/gcc/config/aarch64/aarch64.cc
@@ -15637,7 +15637,7 @@ private:
unsigned int adjust_body_cost (loop_vec_info, const aarch64_vector_costs *,
unsigned int);
bool prefer_unrolled_loop () const;
- unsigned int determine_suggested_unroll_factor ();
+ unsigned int determine_suggested_unroll_factor (loop_vec_info);
/* True if we have performed one-time initialization based on the
vec_info. */
@@ -16746,7 +16746,8 @@ adjust_body_cost_sve (const aarch64_vec_op_count *ops,
}
unsigned int
-aarch64_vector_costs::determine_suggested_unroll_factor ()
+aarch64_vector_costs::
+determine_suggested_unroll_factor (loop_vec_info loop_vinfo)
{
bool sve = m_vec_flags & VEC_ANY_SVE;
/* If we are trying to unroll an Advanced SIMD main loop that contains
@@ -16760,6 +16761,7 @@ aarch64_vector_costs::determine_suggested_unroll_factor ()
return 1;
unsigned int max_unroll_factor = 1;
+ auto vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
for (auto vec_ops : m_ops)
{
aarch64_simd_vec_issue_info const *vec_issue
@@ -16768,7 +16770,8 @@ aarch64_vector_costs::determine_suggested_unroll_factor ()
return 1;
/* Limit unroll factor to a value adjustable by the user, the default
value is 4. */
- unsigned int unroll_factor = aarch64_vect_unroll_limit;
+ unsigned int unroll_factor = MIN (aarch64_vect_unroll_limit,
+ (int) known_alignment (vf));
unsigned int factor
= vec_ops.reduction_latency > 1 ? vec_ops.reduction_latency : 1;
unsigned int temp;
@@ -16946,7 +16949,8 @@ aarch64_vector_costs::finish_cost (const vector_costs *uncast_scalar_costs)
{
m_costs[vect_body] = adjust_body_cost (loop_vinfo, scalar_costs,
m_costs[vect_body]);
- m_suggested_unroll_factor = determine_suggested_unroll_factor ();
+ m_suggested_unroll_factor
+ = determine_suggested_unroll_factor (loop_vinfo);
}
/* Apply the heuristic described above m_stp_sequence_cost. Prefer
diff --git a/gcc/testsuite/g++.dg/vect/pr105254.cc b/gcc/testsuite/g++.dg/vect/pr105254.cc
new file mode 100644
index 00000000000..34814671a54
--- /dev/null
+++ b/gcc/testsuite/g++.dg/vect/pr105254.cc
@@ -0,0 +1,26 @@
+/* { dg-do compile } */
+/* { dg-additional-options "-fvect-cost-model=dynamic -mcpu=zeus" { target aarch64*-*-* } } */
+
+template <typename>
+struct complex;
+
+template <>
+struct complex<double> {
+ void operator+= (complex r) { v_ += r.v_; }
+ _Complex v_;
+};
+
+template <typename T>
+void
+bar (T);
+
+void
+foo (complex<double> *x)
+{
+ complex<double> s = {0.0};
+
+ for (int i = 0; i < 16; ++i)
+ s += x[i];
+
+ bar<complex<double> > (s);
+}
--
2.25.1
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2022-04-13 16:54 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-04-13 16:54 [pushed] aarch64: Make sure the UF divides the VF [PR105254] Richard Sandiford
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).