The attached patch was originally posted in 2017 and has been carried
around on the OGxx development branch ever since. Here's the original
patch mail:
https://gcc.gnu.org/pipermail/gcc-patches/2017-May/473643.html
I'd like to get this on mainline, finally. It's a near-trivial bug fix
for an OpenACC 2.5 conformance issue; we currently document that GCC
"strives to be compatible" with version 2.6 and the behavior is supposed
to be the same there (it's section 2.9.9).
I retested this in isolation on mainline and the results look good. OK
to commit? I've left the original author attribution in place to give
proper credit for the patch.
-Sandra
From 68c87a64273297467a66dff7177f34ceae1207ec Mon Sep 17 00:00:00 2001
From: Cesar Philippidis <[email protected]>
Date: Tue, 26 Feb 2019 15:55:23 -0800
Subject: [PATCH] Don't mark OpenACC auto loops as independent inside acc
parallel regions
gcc/ChangeLog
* omp-low.cc (lower_oacc_head_mark): Don't mark OpenACC auto
loops as independent inside acc parallel regions.
gcc/testsuite/ChangeLog
* c-c++-common/goacc/loop-auto-1.c: Adjust test case to conform to
the new behavior of the auto clause in OpenACC 2.5.
* c-c++-common/goacc/loop-auto-2.c: Likewise.
* gcc.dg/goacc/loop-processing-1.c: Likewise.
* c-c++-common/goacc/loop-auto-3.c: New test.
* gfortran.dg/goacc/loop-auto-1.f90: New test.
libgomp/ChangeLog
* testsuite/libgomp.oacc-c-c++-common/loop-auto-1.c: Adjust test case
to conform to the new behavior of the auto clause in OpenACC 2.5.
---
gcc/omp-low.cc | 6 +-
.../c-c++-common/goacc/loop-auto-1.c | 50 +++++------
.../c-c++-common/goacc/loop-auto-2.c | 4 +-
.../c-c++-common/goacc/loop-auto-3.c | 78 ++++++++++++++++
.../gcc.dg/goacc/loop-processing-1.c | 2 +-
.../gfortran.dg/goacc/loop-auto-1.f90 | 88 +++++++++++++++++++
.../libgomp.oacc-c-c++-common/loop-auto-1.c | 20 ++---
7 files changed, 208 insertions(+), 40 deletions(-)
create mode 100644 gcc/testsuite/c-c++-common/goacc/loop-auto-3.c
create mode 100644 gcc/testsuite/gfortran.dg/goacc/loop-auto-1.f90
diff --git a/gcc/omp-low.cc b/gcc/omp-low.cc
index 270a9abee33..ba5ee5fb380 100644
--- a/gcc/omp-low.cc
+++ b/gcc/omp-low.cc
@@ -8356,8 +8356,10 @@ lower_oacc_head_mark (location_t loc, tree ddvar, tree clauses,
else
gcc_unreachable ();
- /* In a parallel region, loops are implicitly INDEPENDENT. */
- if (!tgt || is_oacc_parallel_or_serial (tgt))
+ /* In a parallel region, loops without auto and seq clauses are
+ implicitly INDEPENDENT. */
+ if ((!tgt || is_oacc_parallel_or_serial (tgt))
+ && !(tag & (OLF_SEQ | OLF_AUTO)))
tag |= OLF_INDEPENDENT;
/* Loops inside OpenACC 'kernels' decomposed parts' regions are expected to
diff --git a/gcc/testsuite/c-c++-common/goacc/loop-auto-1.c b/gcc/testsuite/c-c++-common/goacc/loop-auto-1.c
index 124befc4002..dcad07f11c8 100644
--- a/gcc/testsuite/c-c++-common/goacc/loop-auto-1.c
+++ b/gcc/testsuite/c-c++-common/goacc/loop-auto-1.c
@@ -10,7 +10,7 @@ void Foo ()
#pragma acc loop seq
for (int jx = 0; jx < 10; jx++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++) {}
}
@@ -20,7 +20,7 @@ void Foo ()
#pragma acc loop auto
for (int jx = 0; jx < 10; jx++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++)
{
#pragma acc loop vector
@@ -51,7 +51,7 @@ void Foo ()
#pragma acc loop vector
for (int jx = 0; jx < 10; jx++)
{
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int kx = 0; kx < 10; kx++) {}
}
@@ -64,27 +64,27 @@ void Foo ()
}
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < 10; ix++)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < 10; jx++)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int kx = 0; kx < 10; kx++) {}
}
}
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < 10; ix++)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < 10; jx++)
{
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int kx = 0; kx < 10; kx++)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int lx = 0; lx < 10; lx++) {}
}
}
@@ -101,7 +101,7 @@ void Gang (void)
#pragma acc loop seq
for (int jx = 0; jx < 10; jx++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++) {}
}
@@ -111,7 +111,7 @@ void Gang (void)
#pragma acc loop auto
for (int jx = 0; jx < 10; jx++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++)
{
#pragma acc loop vector
@@ -142,7 +142,7 @@ void Gang (void)
#pragma acc loop vector
for (int jx = 0; jx < 10; jx++)
{
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int kx = 0; kx < 10; kx++) {}
}
@@ -176,7 +176,7 @@ void Worker (void)
#pragma acc loop seq
for (int jx = 0; jx < 10; jx++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++) {}
}
@@ -186,7 +186,7 @@ void Worker (void)
#pragma acc loop auto
for (int jx = 0; jx < 10; jx++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++)
{
#pragma acc loop vector
@@ -194,20 +194,20 @@ void Worker (void)
}
}
-#pragma acc loop auto
+#pragma acc loop
for (int ix = 0; ix < 10; ix++)
{
-#pragma acc loop auto
+#pragma acc loop
for (int jx = 0; jx < 10; jx++) {}
}
-#pragma acc loop auto
+#pragma acc loop
for (int ix = 0; ix < 10; ix++)
{
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++)
{
-#pragma acc loop auto
+#pragma acc loop
for (int kx = 0; kx < 10; kx++) {}
}
}
@@ -222,17 +222,17 @@ void Vector (void)
#pragma acc loop seq
for (int jx = 0; jx < 10; jx++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++) {}
}
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < 10; ix++) {}
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int ix = 0; ix < 10; ix++)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < 10; jx++) {}
}
}
@@ -240,6 +240,6 @@ void Vector (void)
#pragma acc routine seq
void Seq (void)
{
-#pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int ix = 0; ix < 10; ix++) {}
}
diff --git a/gcc/testsuite/c-c++-common/goacc/loop-auto-2.c b/gcc/testsuite/c-c++-common/goacc/loop-auto-2.c
index af3f0bddf2c..5aa36e93ab8 100644
--- a/gcc/testsuite/c-c++-common/goacc/loop-auto-2.c
+++ b/gcc/testsuite/c-c++-common/goacc/loop-auto-2.c
@@ -72,12 +72,12 @@ void Bad ()
#pragma acc loop tile(*) gang vector
for (int ix = 0; ix < 10; ix++)
{
- #pragma acc loop auto /* { dg-warning "insufficient partitioning" } */
+ #pragma acc loop auto independent /* { dg-warning "insufficient partitioning" } */
for (int jx = 0; jx < 10; jx++)
;
}
-#pragma acc loop tile(*) auto /* { dg-warning "insufficient partitioning" } */
+#pragma acc loop tile(*) auto independent /* { dg-warning "insufficient partitioning" } */
for (int ix = 0; ix < 10; ix++)
{
#pragma acc loop worker
diff --git a/gcc/testsuite/c-c++-common/goacc/loop-auto-3.c b/gcc/testsuite/c-c++-common/goacc/loop-auto-3.c
new file mode 100644
index 00000000000..8f79ead16aa
--- /dev/null
+++ b/gcc/testsuite/c-c++-common/goacc/loop-auto-3.c
@@ -0,0 +1,78 @@
+/* Ensure that the auto clause falls back to seq parallelism when the
+ OpenACC loop is not explicitly independent. */
+
+/* { dg-additional-options "-fopt-info-optimized-omp" } */
+
+void
+test ()
+{
+ int i, j, k, l, n = 100;
+
+#pragma acc parallel loop auto /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ for (i = 0; i < n; i++)
+#pragma acc loop auto independent /* { dg-message "optimized: assigned OpenACC gang loop parallelism" } */
+ for (j = 0; j < n; j++)
+#pragma acc loop worker vector /* { dg-message "optimized: assigned OpenACC worker vector loop parallelism" } */
+ for (k = 0; k < n; k++)
+ ;
+
+#pragma acc parallel loop auto independent /* { dg-message "optimized: assigned OpenACC gang worker loop parallelism" } */
+ for (i = 0; i < n; i++)
+#pragma acc loop auto /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ for (j = 0; j < n; j++)
+#pragma acc loop auto /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ for (k = 0; k < n; k++)
+#pragma acc loop auto independent /* { dg-message "optimized: assigned OpenACC vector loop parallelism" } */
+ for (l = 0; l < n; l++)
+ ;
+
+#pragma acc parallel loop gang /* { dg-message "optimized: assigned OpenACC gang loop parallelism" } */
+ for (i = 0; i < n; i++)
+#pragma acc loop worker /* { dg-message "optimized: assigned OpenACC worker loop parallelism" } */
+ for (j = 0; j < n; j++)
+#pragma acc loop vector /* { dg-message "optimized: assigned OpenACC vector loop parallelism" } */
+ for (k = 0; k < n; k++)
+ {
+#pragma acc loop auto independent /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ /* { dg-warning "insufficient partitioning available to parallelize loop" "" { target *-*-* } .-1 } */
+ for (l = 0; l < n; l++)
+ ;
+#pragma acc loop auto /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ for (l = 0; l < n; l++)
+ ;
+ }
+
+#pragma acc parallel loop /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ /* { dg-warning "insufficient partitioning available to parallelize loop" "" { target *-*-* } .-1 } */
+ for (i = 0; i < n; i++)
+ {
+#pragma acc loop gang worker /* { dg-message "optimized: assigned OpenACC gang worker loop parallelism" } */
+ for (j = 0; j < n; j++)
+#pragma acc loop auto /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ for (k = 0; k < n; k++)
+ {
+#pragma acc loop vector /* { dg-message "optimized: assigned OpenACC vector loop parallelism" } */
+ for (l = 0; l < n; l++)
+ ;
+#pragma acc loop auto independent /* { dg-message "optimized: assigned OpenACC vector loop parallelism" } */
+ for (l = 0; l < n; l++)
+ ;
+ }
+#pragma acc loop worker /* { dg-message "optimized: assigned OpenACC worker loop parallelism" } */
+ for (j = 0; j < n; j++)
+#pragma acc loop vector /* { dg-message "optimized: assigned OpenACC vector loop parallelism" } */
+ for (k = 0; k < n; k++)
+ ;
+ }
+
+#pragma acc parallel loop /* { dg-message "optimized: assigned OpenACC gang loop parallelism" } */
+ for (i = 0; i < n; i++)
+#pragma acc loop /* { dg-message "optimized: assigned OpenACC worker loop parallelism" } */
+ for (j = 0; j < n; j++)
+#pragma acc loop /* { dg-message "optimized: assigned OpenACC seq loop parallelism" } */
+ /* { dg-warning "insufficient partitioning available to parallelize loop" "" { target *-*-* } .-1 } */
+ for (k = 0; k < n; k++)
+#pragma acc loop /* { dg-message "optimized: assigned OpenACC vector loop parallelism" } */
+ for (l = 0; l < n; l++)
+ ;
+}
diff --git a/gcc/testsuite/gcc.dg/goacc/loop-processing-1.c b/gcc/testsuite/gcc.dg/goacc/loop-processing-1.c
index f6e25151e1e..79be012058e 100644
--- a/gcc/testsuite/gcc.dg/goacc/loop-processing-1.c
+++ b/gcc/testsuite/gcc.dg/goacc/loop-processing-1.c
@@ -9,7 +9,7 @@ void vector_1 (int *ary, int size)
{
#pragma acc loop gang
for (int jx = 0; jx < 1; jx++)
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < size; ix++)
ary[ix] = place ();
}
diff --git a/gcc/testsuite/gfortran.dg/goacc/loop-auto-1.f90 b/gcc/testsuite/gfortran.dg/goacc/loop-auto-1.f90
new file mode 100644
index 00000000000..8d600f49c14
--- /dev/null
+++ b/gcc/testsuite/gfortran.dg/goacc/loop-auto-1.f90
@@ -0,0 +1,88 @@
+! Ensure that the auto clause falls back to seq parallelism when the
+! OpenACC loop is not explicitly independent.
+
+! { dg-additional-options "-fopt-info-optimized-omp" }
+
+program test
+ implicit none
+ integer, parameter :: n = 100
+ integer i, j, k, l
+
+ !$acc parallel loop auto ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ do i = 1, n
+ !$acc loop auto independent ! { dg-message "optimized: assigned OpenACC gang loop parallelism" }
+ do j = 1, n
+ !$acc loop worker vector ! { dg-message "optimized: assigned OpenACC worker vector loop parallelism" }
+ do k = 1, n
+ end do
+ end do
+ end do
+
+ !$acc parallel loop auto independent ! { dg-message "optimized: assigned OpenACC gang worker loop parallelism" }
+ do i = 1, n
+ !$acc loop auto ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ do j = 1, n
+ !$acc loop auto ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ do k = 1, n
+ !$acc loop auto independent ! { dg-message "optimized: assigned OpenACC vector loop parallelism" }
+ do l = 1, n
+ end do
+ end do
+ end do
+ end do
+
+ !$acc parallel loop gang ! { dg-message "optimized: assigned OpenACC gang loop parallelism" }
+ do i = 1, n
+ !$acc loop worker ! { dg-message "optimized: assigned OpenACC worker loop parallelism" }
+ do j = 1, n
+ !$acc loop vector ! { dg-message "optimized: assigned OpenACC vector loop parallelism" }
+ do k = 1, n
+ !$acc loop auto independent ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ ! { dg-warning "insufficient partitioning available to parallelize loop" "" { target *-*-* } .-1 }
+ do l = 1, n
+ end do
+ !$acc loop auto ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ do l = 1, n
+ end do
+ end do
+ end do
+ end do
+
+
+ !$acc parallel loop ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ ! { dg-warning "insufficient partitioning available to parallelize loop" "" { target *-*-* } .-1 }
+ do i = 1, n
+ !$acc loop gang worker ! { dg-message "optimized: assigned OpenACC gang worker loop parallelism" }
+ do j = 1, n
+ !$acc loop auto ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ do k = 1, n
+ !$acc loop vector ! { dg-message "optimized: assigned OpenACC vector loop parallelism" }
+ do l = 1, n
+ end do
+ end do
+ !$acc loop auto independent ! { dg-message "optimized: assigned OpenACC vector loop parallelism" }
+ do l = 1, n
+ end do
+ end do
+ !$acc loop worker ! { dg-message "optimized: assigned OpenACC worker loop parallelism" }
+ do j = 1, n
+ !$acc loop vector ! { dg-message "optimized: assigned OpenACC vector loop parallelism" }
+ do k = 1, n
+ end do
+ end do
+ end do
+
+ !$acc parallel loop ! { dg-message "optimized: assigned OpenACC gang loop parallelism" }
+ do i = 1, n
+ !$acc loop ! { dg-message "optimized: assigned OpenACC worker loop parallelism" }
+ do j = 1, n
+ !$acc loop ! { dg-message "optimized: assigned OpenACC seq loop parallelism" }
+ ! { dg-warning "insufficient partitioning available to parallelize loop" "" { target *-*-* } .-1 }
+ do k = 1, n
+ !$acc loop ! { dg-message "optimized: assigned OpenACC vector loop parallelism" }
+ do l = 1, n
+ end do
+ end do
+ end do
+ end do
+end program test
diff --git a/libgomp/testsuite/libgomp.oacc-c-c++-common/loop-auto-1.c b/libgomp/testsuite/libgomp.oacc-c-c++-common/loop-auto-1.c
index c13cab780cb..41827559297 100644
--- a/libgomp/testsuite/libgomp.oacc-c-c++-common/loop-auto-1.c
+++ b/libgomp/testsuite/libgomp.oacc-c-c++-common/loop-auto-1.c
@@ -107,7 +107,7 @@ int vector_1 (int *ary, int size)
{
#pragma acc loop gang
for (int jx = 0; jx < 1; jx++)
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < size; ix++)
ary[ix] = place ();
}
@@ -123,7 +123,7 @@ int vector_2 (int *ary, int size)
{
#pragma acc loop worker
for (int jx = 0; jx < size / 64; jx++)
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < 64; ix++)
ary[ix + jx * 64] = place ();
}
@@ -139,7 +139,7 @@ int worker_1 (int *ary, int size)
{
#pragma acc loop gang
for (int kx = 0; kx < 1; kx++)
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < size / 64; jx++)
#pragma acc loop vector
for (int ix = 0; ix < 64; ix++)
@@ -156,7 +156,7 @@ int gang_1 (int *ary, int size)
#pragma acc parallel num_gangs (32) num_workers (32) vector_length(32) copy(ary[0:size]) firstprivate (size)
/* { dg-warning "region is vector partitioned but does not contain vector partitioned code" "" { target *-*-* } .-1 } */
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < size / 64; jx++)
#pragma acc loop worker
for (int ix = 0; ix < 64; ix++)
@@ -172,11 +172,11 @@ int gang_2 (int *ary, int size)
#pragma acc parallel num_gangs (32) num_workers (32) vector_length(32) copy(ary[0:size]) firstprivate (size)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int kx = 0; kx < size / (32 * 32); kx++)
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < 32; jx++)
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < 32; ix++)
ary[ix + jx * 32 + kx * 32 * 32] = place ();
}
@@ -190,9 +190,9 @@ int gang_3 (int *ary, int size)
#pragma acc parallel num_workers (32) vector_length(32) copy(ary[0:size]) firstprivate (size)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < size / 64; jx++)
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int ix = 0; ix < 64; ix++)
ary[ix + jx * 64] = place ();
}
@@ -206,7 +206,7 @@ int gang_4 (int *ary, int size)
#pragma acc parallel vector_length(32) copy(ary[0:size]) firstprivate (size)
{
-#pragma acc loop auto
+#pragma acc loop auto independent
for (int jx = 0; jx < size; jx++)
ary[jx] = place ();
}
--
2.39.5
