On Fri, Jul 1, 2016 at 2:31 PM, H.J. Lu <hjl.to...@gmail.com> wrote:
> On Thu, Jun 30, 2016 at 7:51 AM, Yuri Rumyantsev <ysrum...@gmail.com> wrote:
>> Richard,
>>
>> Could you please review additional simple fix for 70729 - we need to
>> nullify safelen field of loops containing simduid intrinsics like
>> GOMP_SIMD_LANE (introduced e.g. for private variables). I checked
>> that this fix cures regression which was missed by me since AVX2
>> machine is required for libgomp.fortran/examples-4/simd-2.f90.
>>
>> Regression testing and bootstrapping did not show any new failures.
>> Is it OK for trunk?
>>
>> Patch:
>> diff --git a/gcc/tree-vectorizer.c b/gcc/tree-vectorizer.c
>> index 2669813..9fbd183 100644
>> --- a/gcc/tree-vectorizer.c
>> +++ b/gcc/tree-vectorizer.c
>> @@ -204,6 +204,10 @@ adjust_simduid_builtins (hash_table<simduid_to_vf>
>> *htab)
>> gcc_assert (TREE_CODE (arg) == SSA_NAME);
>> simduid_to_vf *p = NULL, data;
>> data.simduid = DECL_UID (SSA_NAME_VAR (arg));
>> + /* Need to nullify loop safelen field since it's value is not
>> + valid after transformation. */
>> + if (bb->loop_father && bb->loop_father->safelen > 0)
>> + bb->loop_father->safelen = 0;
>> if (htab)
>> {
>> p = htab->find (&data);
>>
>> ChangeLog:
>> 2016-06-30 Yuri Rumyantsev <ysrum...@gmail.com>
>>
>> PR tree-optimization/70729
>> * tree-vectorizer.c (adjust_simduid_builtins): Nullify safelen field
>> of loop since it can be not valid after transformation.
>>
>
> I still see
>
> FAIL: libgomp.fortran/simd3.f90 -O3 -fomit-frame-pointer
> -funroll-loops -fpeel-loops -ftracer -finline-functions execution
> test
> FAIL: libgomp.fortran/simd3.f90 -O3 -g execution test
> FAIL: libgomp.fortran/simd4.f90 -O3 -fomit-frame-pointer
> -funroll-loops -fpeel-loops -ftracer -finline-functions execution
> test
> FAIL: libgomp.fortran/simd4.f90 -O3 -g execution test
>
> https://gcc.gnu.org/bugzilla/show_bug.cgi?id=71734
Here are testcases:
FAIL: libgomp.fortran/pr71734-1.f90 -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions execution
test
FAIL: libgomp.fortran/pr71734-1.f90 -O3 -g execution test
FAIL: libgomp.fortran/pr71734-2.f90 -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions execution
test
FAIL: libgomp.fortran/pr71734-2.f90 -O3 -g execution test
FAIL: libgomp.fortran/pr71734-1.f90 -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions execution
test
FAIL: libgomp.fortran/pr71734-1.f90 -O3 -g execution test
FAIL: libgomp.fortran/pr71734-2.f90 -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions execution
test
FAIL: libgomp.fortran/pr71734-2.f90 -O3 -g execution test
--
H.J.
From d120301ab113db895164289433fcc068611d3fba Mon Sep 17 00:00:00 2001
From: "H.J. Lu" <hjl.to...@gmail.com>
Date: Sun, 3 Jul 2016 08:54:24 -0700
Subject: [PATCH] Add tests for PR 71734
---
libgomp/testsuite/libgomp.fortran/pr71734-1.f90 | 110 ++++++++++++++++++++++++
libgomp/testsuite/libgomp.fortran/pr71734-2.f90 | 104 ++++++++++++++++++++++
2 files changed, 214 insertions(+)
create mode 100644 libgomp/testsuite/libgomp.fortran/pr71734-1.f90
create mode 100644 libgomp/testsuite/libgomp.fortran/pr71734-2.f90
diff --git a/libgomp/testsuite/libgomp.fortran/pr71734-1.f90 b/libgomp/testsuite/libgomp.fortran/pr71734-1.f90
new file mode 100644
index 0000000..a5617d7
--- /dev/null
+++ b/libgomp/testsuite/libgomp.fortran/pr71734-1.f90
@@ -0,0 +1,110 @@
+! { dg-do run { target avx_runtime } }
+! { dg-additional-options "-msse2" }
+! The same as simd3.f90, but compiled with -msse2. we run it only on
+! AVX machine where simd4.f90 is compiled with -mavx.
+
+ integer :: a(1024), b(1024), k, m, i, s, t
+ k = 4
+ m = 2
+ t = 1
+ do i = 1, 1024
+ a(i) = i - 513
+ b(i) = modulo (i - 52, 39)
+ if (i.lt.52.and.b(i).ne.0) b(i) = b(i) - 39
+ end do
+ s = foo (b)
+ do i = 1, 1024
+ if (a(i).ne.((i - 513) * b(i))) call abort
+ if (i.lt.52.and.modulo (i - 52, 39).ne.0) then
+ if (b(i).ne.(modulo (i - 52, 39) - 39)) call abort
+ else
+ if (b(i).ne.(modulo (i - 52, 39))) call abort
+ end if
+ a(i) = i - 513
+ end do
+ if (k.ne.(4 + 3 * 1024).or.s.ne.1596127) call abort
+ k = 4
+ m = 2
+ t = 1
+ s = bar (b)
+ do i = 1, 1024
+ if (a(i).ne.((i - 513) * b(i))) call abort
+ if (i.lt.52.and.modulo (i - 52, 39).ne.0) then
+ if (b(i).ne.(modulo (i - 52, 39) - 39)) call abort
+ else
+ if (b(i).ne.(modulo (i - 52, 39))) call abort
+ end if
+ a(i) = i - 513
+ end do
+ if (k.ne.(4 + 3 * 1024).or.s.ne.1596127) call abort
+ k = 4
+ m = 2
+ t = 1
+ s = baz (b)
+ do i = 1, 1024
+ if (a(i).ne.((i - 513) * b(i))) call abort
+ if (i.lt.52.and.modulo (i - 52, 39).ne.0) then
+ if (b(i).ne.(modulo (i - 52, 39) - 39)) call abort
+ else
+ if (b(i).ne.(modulo (i - 52, 39))) call abort
+ end if
+ end do
+ if (k.ne.(4 + 3 * 1024).or.s.ne.1596127) call abort
+contains
+ function foo (p)
+ integer :: p(1024), u, v, i, s, foo
+ s = 0
+ !$omp parallel
+ !$omp do simd linear(k : m + 1) reduction(+: s) lastprivate(u, v) &
+ !$omp & schedule (static, 32)
+ do i = 1, 1024
+ a(i) = a(i) * p(i)
+ u = p(i) + k
+ k = k + m + 1
+ v = p(i) + k
+ s = s + p(i) + k
+ end do
+ !$omp end do simd
+ !$omp end parallel
+ if (i.ne.1025) call abort
+ if (u.ne.(36 + 4 + 3 * 1023).or.v.ne.(36 + 4 + 3 * 1024)) call abort
+ foo = s
+ end function foo
+ function bar (p)
+ integer :: p(1024), u, v, i, s, bar
+ s = 0
+ !$omp parallel
+ !$omp do simd linear(k : m + 1) reduction(+: s) lastprivate(u, v) &
+ !$omp & schedule (dynamic, 32)
+ do i = 1, 1024, t
+ a(i) = a(i) * p(i)
+ u = p(i) + k
+ k = k + m + 1
+ v = p(i) + k
+ s = s + p(i) + k
+ end do
+ !$omp end do simd
+ !$omp endparallel
+ if (i.ne.1025) call abort
+ if (u.ne.(36 + 4 + 3 * 1023).or.v.ne.(36 + 4 + 3 * 1024)) call abort
+ bar = s
+ end function bar
+ function baz (p)
+ integer :: p(1024), u, v, i, s, baz
+ s = 0
+ !$omp parallel
+ !$omp do simd linear(k : m + 1) reduction(+: s) lastprivate(u, v) &
+ !$omp & linear(i : t) schedule (static, 8)
+ do i = 1, 1024, t
+ a(i) = a(i) * p(i)
+ u = p(i) + k
+ k = k + m + 1
+ v = p(i) + k
+ s = s + p(i) + k
+ end do
+ !$omp end parallel
+ if (i.ne.1025) call abort
+ if (u.ne.(36 + 4 + 3 * 1023).or.v.ne.(36 + 4 + 3 * 1024)) call abort
+ baz = s
+ end function baz
+end
diff --git a/libgomp/testsuite/libgomp.fortran/pr71734-2.f90 b/libgomp/testsuite/libgomp.fortran/pr71734-2.f90
new file mode 100644
index 0000000..1cad4c1
--- /dev/null
+++ b/libgomp/testsuite/libgomp.fortran/pr71734-2.f90
@@ -0,0 +1,104 @@
+! { dg-do run { target avx_runtime } }
+! { dg-additional-options "-msse2" }
+! The same as simd4.f90, but compiled with -msse2. we run it only on
+! AVX machine where simd4.f90 is compiled with -mavx.
+
+ integer :: a(1024), b(1024), k, m, i, s, t
+ k = 4
+ m = 2
+ t = 1
+ do i = 1, 1024
+ a(i) = i - 513
+ b(i) = modulo (i - 52, 39)
+ if (i.lt.52.and.b(i).ne.0) b(i) = b(i) - 39
+ end do
+ s = foo (b)
+ do i = 1, 1024
+ if (a(i).ne.((i - 513) * b(i))) call abort
+ if (i.lt.52.and.modulo (i - 52, 39).ne.0) then
+ if (b(i).ne.(modulo (i - 52, 39) - 39)) call abort
+ else
+ if (b(i).ne.(modulo (i - 52, 39))) call abort
+ end if
+ a(i) = i - 513
+ end do
+ if (k.ne.(4 + 3 * 1024).or.s.ne.1596127) call abort
+ k = 4
+ m = 2
+ t = 1
+ s = bar (b)
+ do i = 1, 1024
+ if (a(i).ne.((i - 513) * b(i))) call abort
+ if (i.lt.52.and.modulo (i - 52, 39).ne.0) then
+ if (b(i).ne.(modulo (i - 52, 39) - 39)) call abort
+ else
+ if (b(i).ne.(modulo (i - 52, 39))) call abort
+ end if
+ a(i) = i - 513
+ end do
+ if (k.ne.(4 + 3 * 1024).or.s.ne.1596127) call abort
+ k = 4
+ m = 2
+ t = 1
+ s = baz (b)
+ do i = 1, 1024
+ if (a(i).ne.((i - 513) * b(i))) call abort
+ if (i.lt.52.and.modulo (i - 52, 39).ne.0) then
+ if (b(i).ne.(modulo (i - 52, 39) - 39)) call abort
+ else
+ if (b(i).ne.(modulo (i - 52, 39))) call abort
+ end if
+ end do
+ if (k.ne.(4 + 3 * 1024).or.s.ne.1596127) call abort
+contains
+ function foo (p)
+ integer :: p(1024), u, v, i, s, foo
+ s = 0
+ !$omp parallel do simd linear(k : m + 1) reduction(+: s) &
+ !$omp & lastprivate(u, v) schedule (static, 32)
+ do i = 1, 1024
+ a(i) = a(i) * p(i)
+ u = p(i) + k
+ k = k + m + 1
+ v = p(i) + k
+ s = s + p(i) + k
+ end do
+ !$omp end parallel do simd
+ if (i.ne.1025) call abort
+ if (u.ne.(36 + 4 + 3 * 1023).or.v.ne.(36 + 4 + 3 * 1024)) call abort
+ foo = s
+ end function foo
+ function bar (p)
+ integer :: p(1024), u, v, i, s, bar
+ s = 0
+ !$omp parallel do simd linear(k : m + 1) reduction(+: s) &
+ !$omp & lastprivate(u, v) schedule (dynamic, 32)
+ do i = 1, 1024, t
+ a(i) = a(i) * p(i)
+ u = p(i) + k
+ k = k + m + 1
+ v = p(i) + k
+ s = s + p(i) + k
+ end do
+ !$omp endparalleldosimd
+ if (i.ne.1025) call abort
+ if (u.ne.(36 + 4 + 3 * 1023).or.v.ne.(36 + 4 + 3 * 1024)) call abort
+ bar = s
+ end function bar
+ function baz (p)
+ integer :: p(1024), u, v, i, s, baz
+ s = 0
+ !$omp parallel do simd linear(k : m + 1) reduction(+: s) &
+ !$omp & lastprivate(u, v) linear(i : t) schedule (static, 8)
+ do i = 1, 1024, t
+ a(i) = a(i) * p(i)
+ u = p(i) + k
+ k = k + m + 1
+ v = p(i) + k
+ s = s + p(i) + k
+ end do
+ if (i.ne.1025) call abort
+ if (u.ne.(36 + 4 + 3 * 1023).or.v.ne.(36 + 4 + 3 * 1024)) call abort
+ baz = s
+ end function baz
+end
--
2.7.4
