Update OpenACC test cases

	gcc/testsuite/
	* c-c++-common/goacc/combined-directives.c: Clean up dg-*
	directives.
	* c-c++-common/goacc/loop-clauses.c: Likewise.
	* g++.dg/goacc/template.C: Likewise.
	* gfortran.dg/goacc/combined-directives.f90: Likewise.
	* gfortran.dg/goacc/loop-1.f95: Likewise.
	* gfortran.dg/goacc/loop-5.f95: Likewise.
	* gfortran.dg/goacc/loop-6.f95: Likewise.
	* gfortran.dg/goacc/loop-tree-1.f90: Likewise.
	* c-c++-common/goacc-gomp/nesting-1.c: Update.
	* c-c++-common/goacc-gomp/nesting-fail-1.c: Likewise.
	* c-c++-common/goacc/clauses-fail.c: Likewise.
	* c-c++-common/goacc/parallel-1.c: Likewise.
	* c-c++-common/goacc/reduction-1.c: Likewise.
	* c-c++-common/goacc/reduction-2.c: Likewise.
	* c-c++-common/goacc/reduction-3.c: Likewise.
	* c-c++-common/goacc/reduction-4.c: Likewise.
	* c-c++-common/goacc/routine-3.c: Likewise.
	* c-c++-common/goacc/routine-4.c: Likewise.
	* c-c++-common/goacc/routine-5.c: Likewise.
	* c-c++-common/goacc/tile.c: Likewise.
	* g++.dg/goacc/template.C: Likewise.
	* gfortran.dg/goacc/combined-directives.f90: Likewise.
	* c-c++-common/goacc/nesting-1.c: Move dg-error test cases into...
	* c-c++-common/goacc/nesting-fail-1.c: ... this file.  Update.
	* c-c++-common/goacc/kernels-1.c: Update.  Incorporate...
	* c-c++-common/goacc/kernels-empty.c: ... this file, and...
	* c-c++-common/goacc/kernels-eternal.c: ... this file, and...
	* c-c++-common/goacc/kernels-noreturn.c: ... this file.
	* c-c++-common/goacc/host_data-1.c: New file.  Incorporate...
	* c-c++-common/goacc/use_device-1.c: ... this file.
	* c-c++-common/goacc/host_data-2.c: New file.  Incorporate...
	* c-c++-common/goacc/host_data-5.c: ... this file, and...
	* c-c++-common/goacc/host_data-6.c: ... this file.
	* c-c++-common/goacc/loop-2-kernels.c: New file.
	* c-c++-common/goacc/loop-2-parallel.c: Likewise.
	* c-c++-common/goacc/loop-3.c: Likewise.
	* g++.dg/goacc/reference.C: Likewise.
	* g++.dg/goacc/routine-1.C: Likewise.
	* g++.dg/goacc/routine-2.C: Likewise.
	libgomp/
	* testsuite/libgomp.oacc-c-c++-common/clauses-1.c: Update.
	* testsuite/libgomp.oacc-c-c++-common/deviceptr-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/if-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/vector-loop.c: Likewise.
	* testsuite/libgomp.oacc-fortran/asyncwait-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/asyncwait-2.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/asyncwait-3.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/declare-1.f90: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/asyncwait-1.c: Likewise.
	XFAIL.
	* testsuite/libgomp.oacc-c-c++-common/firstprivate-1.c: Update.
	Incorporate...
	* testsuite/libgomp.oacc-c-c++-common/firstprivate-2.c: ... this
	file.
	* testsuite/libgomp.oacc-c++/template-reduction.C: New file.
	* testsuite/libgomp.oacc-c-c++-common/gang-static-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/gang-static-2.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/kernels-loop-clauses.c:
	Likewise.
	* testsuite/libgomp.oacc-c-c++-common/private-variables.c:
	Likewise.
	* testsuite/libgomp.oacc-c-c++-common/reduction-7.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/routine-1.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/routine-4.c: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/routine-wv-2.c: Likewise.
	* testsuite/libgomp.oacc-fortran/clauses-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/default-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/firstprivate-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/gang-static-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/if-1.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/implicit-firstprivate-ref.f90:
	Likewise.
	* testsuite/libgomp.oacc-fortran/pr68813.f90: Likewise.
	* testsuite/libgomp.oacc-fortran/private-variables.f90: Likewise.
	* testsuite/libgomp.oacc-c-c++-common/kernels-1.c: Merge this
	file...
	* testsuite/libgomp.oacc-c-c++-common/parallel-1.c: ..., and this
	file into...
	* testsuite/libgomp.oacc-c-c++-common/data-clauses.h: ... this new
	file.  Update.
	* testsuite/libgomp.oacc-c-c++-common/data-clauses-kernels.c: New
	file.
	* testsuite/libgomp.oacc-c-c++-common/data-clauses-parallel.c:
	Likewise.
	* testsuite/libgomp.oacc-c-c++-common/kernels-2.c: Rename to...
	* testsuite/libgomp.oacc-c-c++-common/data-clauses-kernels-ipa-pta.c:
	... this new file.  Update.
	* testsuite/libgomp.oacc-c-c++-common/parallel-2.c: Rename to...
	* testsuite/libgomp.oacc-c-c++-common/data-clauses-parallel-ipa-pta.c:
	... this new file.  Update.
	* testsuite/libgomp.oacc-c-c++-common/mode-transitions.c: New
	file.  Incorporate...
	* testsuite/libgomp.oacc-c-c++-common/worker-single-1a.c: ... this
	file, and...
	* testsuite/libgomp.oacc-c-c++-common/worker-single-4.c: ... this
	file, and...
	* testsuite/libgomp.oacc-c-c++-common/worker-single-6.c: ... this
	file.
	* testsuite/libgomp.oacc-c-c++-common/update-1-2.c: Remove file.

Co-Authored-By: Cesar Philippidis <cesar@codesourcery.com>
Co-Authored-By: Chung-Lin Tang <cltang@codesourcery.com>
Co-Authored-By: James Norris <jnorris@codesourcery.com>
Co-Authored-By: Julian Brown <julian@codesourcery.com>
Co-Authored-By: Nathan Sidwell <nathan@codesourcery.com>
Co-Authored-By: Tom de Vries <tom@codesourcery.com>

From-SVN: r234575

12 files changed, 2146 insertions, 33 deletions

diff --git a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-1.f90
index b6e637b..01728bd 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-1.f90
+++ b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-1.f90
@@ -132,4 +132,126 @@ program asyncwait
      if (d(i) .ne. 1.0) call abort
      if (e(i) .ne. 11.0) call abort
   end do
+
+  a(:) = 3.0
+  b(:) = 0.0
+
+  !$acc data copy (a(1:N)) copy (b(1:N))
+
+  !$acc kernels async
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end kernels
+
+  !$acc wait
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 3.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+
+  !$acc data copy (a(1:N)) copy (b(1:N))
+
+  !$acc kernels async (1)
+  !$acc loop
+  do i = 1, N
+     b(i) = a(i)
+  end do
+  !$acc end kernels
+
+  !$acc wait (1)
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 2.0) call abort
+  end do
+
+  a(:) = 3.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+
+  !$acc data copy (a(1:N)) copy (b(1:N)) copy (c(1:N)) copy (d(1:N))
+
+  !$acc kernels async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end kernels
+
+  !$acc kernels async (1)
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end kernels
+
+  !$acc kernels async (1)
+  !$acc loop
+  do i = 1, N
+     d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end kernels
+
+  !$acc wait (1)
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 3.0) call abort
+     if (b(i) .ne. 9.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+  end do
+
+  a(:) = 2.0
+  b(:) = 0.0
+  c(:) = 0.0
+  d(:) = 0.0
+  e(:) = 0.0
+
+  !$acc data copy (a(1:N), b(1:N), c(1:N), d(1:N), e(1:N))
+
+  !$acc kernels async (1)
+  do i = 1, N
+     b(i) = (a(i) * a(i) * a(i)) / a(i)
+  end do
+  !$acc end kernels
+
+  !$acc kernels async (1)
+  !$acc loop
+  do i = 1, N
+     c(i) = (a(i) * 4) / a(i)
+  end do
+  !$acc end kernels
+
+  !$acc kernels async (1)
+  !$acc loop
+  do i = 1, N
+     d(i) = ((a(i) * a(i) + a(i)) / a(i)) - a(i)
+  end do
+  !$acc end kernels
+
+  !$acc kernels wait (1) async (1)
+  !$acc loop
+  do i = 1, N
+     e(i) = a(i) + b(i) + c(i) + d(i)
+  end do
+  !$acc end kernels
+
+  !$acc wait (1)
+  !$acc end data
+
+  do i = 1, N
+     if (a(i) .ne. 2.0) call abort
+     if (b(i) .ne. 4.0) call abort
+     if (c(i) .ne. 4.0) call abort
+     if (d(i) .ne. 1.0) call abort
+     if (e(i) .ne. 11.0) call abort
+  end do
 end program asyncwait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-2.f90 b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-2.f90
index bade52b..fe131b6 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-2.f90
+++ b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-2.f90
@@ -1,6 +1,6 @@
 ! { dg-do run }
 
-program parallel_wait
+program asyncwait
   integer, parameter :: N = 64
   real, allocatable :: a(:), b(:), c(:)
   integer i
@@ -33,8 +33,33 @@ program parallel_wait
   do i = 1, N
     if (c(i) .ne. 2.0) call abort
   end do
+
+  !$acc kernels async (0)
+  !$acc loop
+  do i = 1, N
+    a(i) = 1
+  end do
+  !$acc end kernels
+
+  !$acc kernels async (1)
+  !$acc loop
+  do i = 1, N
+    b(i) = 1
+  end do
+  !$acc end kernels
+
+  !$acc kernels wait (0, 1)
+  !$acc loop
+  do i = 1, N
+    c(i) = a(i) + b(i)
+  end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (c(i) .ne. 2.0) call abort
+  end do
   
   deallocate (a)
   deallocate (b)
   deallocate (c)
-end program parallel_wait
+end program asyncwait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-3.f90 b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-3.f90
index d48dc11..fa96a01 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-3.f90
+++ b/libgomp/testsuite/libgomp.oacc-fortran/asyncwait-3.f90
@@ -1,6 +1,6 @@
 ! { dg-do run }
 
-program parallel_wait
+program asyncwait
   integer, parameter :: N = 64
   real, allocatable :: a(:), b(:), c(:)
   integer i
@@ -35,8 +35,35 @@ program parallel_wait
   do i = 1, N
     if (c(i) .ne. 2.0) call abort
   end do
+
+  !$acc kernels async (0)
+  !$acc loop
+  do i = 1, N
+    a(i) = 1
+  end do
+  !$acc end kernels
+
+  !$acc kernels async (1)
+  !$acc loop
+  do i = 1, N
+    b(i) = 1
+  end do
+  !$acc end kernels
+
+  !$acc wait (0, 1)
+
+  !$acc kernels
+  !$acc loop
+  do i = 1, N
+    c(i) = a(i) + b(i)
+  end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (c(i) .ne. 2.0) call abort
+  end do
   
   deallocate (a)
   deallocate (b)
   deallocate (c)
-end program parallel_wait
+end program asyncwait
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/clauses-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/clauses-1.f90
new file mode 100644
index 0000000..e6ab78d
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/clauses-1.f90
@@ -0,0 +1,290 @@
+! { dg-do run }
+! { dg-skip-if "" { *-*-* } { "*" } { "-DACC_MEM_SHARED=0" } }
+
+program main
+  use openacc
+  implicit none
+
+  integer, parameter :: N = 32
+  real, allocatable :: a(:), b(:), c(:)
+  integer i
+
+  i = 0
+
+  allocate (a(N))
+  allocate (b(N))
+  allocate (c(N))
+
+  a(:) = 3.0
+  b(:) = 0.0
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N))
+    do i = 1, N
+      b(i) = a(i)
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 3.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 5.0
+  b(:) = 1.0
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N))
+    do i = 1, N
+      b(i) = a(i)
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 5.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 6.0
+  b(:) = 0.0
+
+  call acc_copyin (a, sizeof (a))
+
+  a(:) = 9.0
+
+  !$acc parallel present_or_copyin (a(1:N)) copyout (b(1:N))
+    do i = 1, N
+      b(i) = a(i)
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 6.0) call abort
+  end do
+
+  call acc_copyout (a, sizeof (a))
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 6.0
+  b(:) = 0.0
+
+  !$acc parallel copyin (a(1:N)) present_or_copyout (b(1:N))
+     do i = 1, N
+       b(i) = a(i)
+     end do
+  !$acc end parallel
+
+  do i = 1, N
+     if (b(i) .ne. 6.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 5.0
+  b(:) = 2.0
+
+  call acc_copyin (b, sizeof (b))
+
+  !$acc parallel copyin (a(1:N)) present_or_copyout (b(1:N))
+     do i = 1, N
+       b(i) = a(i)
+     end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (a(i) .ne. 5.0) call abort
+    if (b(i) .ne. 2.0) call abort
+  end do
+
+  call acc_copyout (b, sizeof (b))
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 3.0;
+  b(:) = 4.0;
+
+  !$acc parallel copy (a(1:N)) copyout (b(1:N))
+    do i = 1, N
+      a(i) = a(i) + 1
+      b(i) = a(i) + 2
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (a(i) .ne. 4.0) call abort
+    if (b(i) .ne. 6.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 4.0
+  b(:) = 7.0
+
+  !$acc parallel present_or_copy (a(1:N)) present_or_copy (b(1:N))
+    do i = 1, N
+      a(i) = a(i) + 1
+      b(i) = b(i) + 2
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (a(i) .ne. 5.0) call abort
+    if (b(i) .ne. 9.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 3.0
+  b(:) = 7.0
+
+  call acc_copyin (a, sizeof (a))
+  call acc_copyin (b, sizeof (b))
+
+  !$acc parallel present_or_copy (a(1:N)) present_or_copy (b(1:N))
+    do i = 1, N
+      a(i) = a(i) + 1
+      b(i) = b(i) + 2
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (a(i) .ne. 3.0) call abort
+    if (b(i) .ne. 7.0) call abort
+  end do
+
+  call acc_copyout (a, sizeof (a))
+  call acc_copyout (b, sizeof (b))
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 3.0
+  b(:) = 7.0
+
+  !$acc parallel copyin (a(1:N)) create (c(1:N)) copyout (b(1:N))
+    do i = 1, N
+      c(i) = a(i)
+      b(i) = c(i)
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (a(i) .ne. 3.0) call abort
+    if (b(i) .ne. 3.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+  if (acc_is_present (c) .eqv. .TRUE.) call abort
+
+  a(:) = 4.0
+  b(:) = 8.0
+
+  !$acc parallel copyin (a(1:N)) present_or_create (c(1:N)) copyout (b(1:N))
+    do i = 1, N
+      c(i) = a(i)
+      b(i) = c(i)
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (a(i) .ne. 4.0) call abort
+    if (b(i) .ne. 4.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+  if (acc_is_present (c) .eqv. .TRUE.) call abort
+
+  a(:) = 4.0
+
+  call acc_copyin (a, sizeof (a))
+  call acc_copyin (b, sizeof (b))
+  call acc_copyin (c, sizeof (c))
+
+  !$acc parallel present (a(1:N)) present (c(1:N)) present (b(1:N))
+    do i = 1, N
+      c(i) = a(i)
+      b(i) = c(i)
+    end do
+  !$acc end parallel
+
+  call acc_copyout (a, sizeof (a))
+  call acc_copyout (b, sizeof (b))
+  call acc_copyout (c, sizeof (c))
+  
+  do i = 1, N
+    if (a(i) .ne. 4.0) call abort
+    if (b(i) .ne. 4.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+  if (acc_is_present (c) .eqv. .TRUE.) call abort
+
+  a(:) = 6.0
+  b(:) = 0.0
+
+  call acc_copyin (a, sizeof (a))
+
+  a(:) = 9.0
+
+  !$acc parallel pcopyin (a(1:N)) copyout (b(1:N))
+    do i = 1, N
+      b(i) = a(i)
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 6.0) call abort
+  end do
+  
+  call acc_copyout (a, sizeof (a))
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 6.0
+  b(:) = 0.0
+
+  !$acc parallel copyin (a(1:N)) pcopyout (b(1:N))
+   do i = 1, N
+     b(i) = a(i)
+   end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 6.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+
+  a(:) = 5.0
+  b(:) = 7.0
+
+  !$acc parallel copyin (a(1:N)) pcreate (c(1:N)) copyout (b(1:N))
+    do i = 1, N
+      c(i) = a(i)
+      b(i) = c(i)
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (a(i) .ne. 5.0) call abort
+    if (b(i) .ne. 5.0) call abort
+  end do
+
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+  if (acc_is_present (c) .eqv. .TRUE.) call abort
+
+end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/declare-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/declare-1.f90
index f717d1b..2d4b707 100644
--- a/libgomp/testsuite/libgomp.oacc-fortran/declare-1.f90
+++ b/libgomp/testsuite/libgomp.oacc-fortran/declare-1.f90
@@ -1,29 +1,22 @@
 ! { dg-do run  { target openacc_nvidia_accel_selected } }
 
+! Tests to exercise the declare directive along with
+! the clauses: copy
+!              copyin
+!              copyout
+!              create
+!              present
+!              present_or_copy
+!              present_or_copyin
+!              present_or_copyout
+!              present_or_create
+
 module vars
   implicit none
   integer z
   !$acc declare create (z)
 end module vars
 
-subroutine subr6 (a, d)
-  implicit none
-  integer, parameter :: N = 8
-  integer :: i
-  integer :: a(N)
-  !$acc declare deviceptr (a)
-  integer :: d(N)
-
-  i = 0
-
-  !$acc parallel copy (d)
-    do i = 1, N
-      d(i) = a(i) + a(i)
-    end do
-  !$acc end parallel
-
-end subroutine
-
 subroutine subr5 (a, b, c, d)
   implicit none
   integer, parameter :: N = 8
@@ -201,15 +194,6 @@ subroutine subr0 (a, b, c, d)
     if (d(i) .ne. 13) call abort
   end do
 
-  call subr6 (a, d)
-
-  call test (a, .true.)
-  call test (d, .false.)
-
-  do i = 1, N
-    if (d(i) .ne. 16) call abort
-  end do
-
 end subroutine
 
 program main
@@ -241,8 +225,7 @@ program main
     if (a(i) .ne. 8) call abort
     if (b(i) .ne. 8) call abort
     if (c(i) .ne. 8) call abort
-    if (d(i) .ne. 16) call abort
+    if (d(i) .ne. 13) call abort
   end do
 
-
 end program
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/default-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/default-1.f90
new file mode 100644
index 0000000..1059089
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/default-1.f90
@@ -0,0 +1,54 @@
+! { dg-do run }
+
+program main
+  implicit none
+  real a, b
+  real c
+  !$acc declare create (c)
+
+  a = 2.0
+  b = 0.0
+
+  !$acc parallel copy (a) create (b) default (none)
+    b = a
+    a = 1.0
+    a = a + b
+  !$acc end parallel
+
+  if (a .ne. 3.0) call abort
+
+  !$acc kernels copy (a) create (b) default (none)
+    b = a
+    a = 1.0
+    a = a + b
+  !$acc end kernels
+
+  if (a .ne. 4.0) call abort
+
+  !$acc parallel default (none) copy (a) create (b)
+    b = a
+    a = 1.0
+    a = a + b
+  !$acc end parallel
+
+  if (a .ne. 5.0) call abort
+
+  !$acc parallel default (none) copy (a)
+    c = a
+    a = 1.0
+    a = a + c
+  !$acc end parallel
+
+  if (a .ne. 6.0) call abort
+
+  !$acc data copy (a)
+  !$acc parallel default (none)
+    c = a
+    a = 1.0
+    a = a + c
+  !$acc end parallel
+  !$acc end data
+
+  if (a .ne. 7.0) call abort
+
+end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/firstprivate-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/firstprivate-1.f90
new file mode 100644
index 0000000..d3f9093
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/firstprivate-1.f90
@@ -0,0 +1,42 @@
+! { dg-do run }
+
+program firstprivate
+  integer, parameter :: Nupper=100
+  integer :: a, b(Nupper), c, d, n
+  include "openacc_lib.h"
+
+  if (acc_get_device_type () .eq. acc_device_nvidia) then
+     n = Nupper
+  else
+     n = 1
+  end if
+
+  b(:) = -1
+  a = 5
+
+  !$acc parallel firstprivate (a) num_gangs (n)
+  !$acc loop gang
+  do i = 1, n
+     a = a + i
+     b(i) = a
+  end do
+  !$acc end parallel
+
+  do i = 1, n
+     if (b(i) .ne. i + a) call abort ()
+  end do
+
+  !$acc data copy (a)
+  !$acc parallel firstprivate (a) copyout (c)
+  a = 10
+  c = a
+  !$acc end parallel
+
+  !$acc parallel copyout (d) present (a)
+  d = a
+  !$acc end parallel
+  !$acc end data
+
+  if (c .ne. 10) call abort ()
+  if (d .ne. 5) call abort ()
+end program firstprivate
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/gang-static-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/gang-static-1.f90
new file mode 100644
index 0000000..7d56060
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/gang-static-1.f90
@@ -0,0 +1,79 @@
+! { dg-do run }
+
+program main
+  integer, parameter :: n = 100
+  integer i, a(n), b(n)
+  integer x
+
+  do i = 1, n
+     b(i) = i
+  end do
+
+  !$acc parallel loop gang (static:*) num_gangs (10)
+  do i = 1, n
+     a(i) = b(i) + 0
+  end do
+  !$acc end parallel loop
+
+  call test (a, b, 0, n)
+
+  !$acc parallel loop gang (static:1) num_gangs (10)
+  do i = 1, n
+     a(i) = b(i) + 1
+  end do
+  !$acc end parallel loop
+
+  call test (a, b, 1, n)
+
+  !$acc parallel loop gang (static:2) num_gangs (10)
+  do i = 1, n
+     a(i) = b(i) + 2
+  end do
+  !$acc end parallel loop
+
+  call test (a, b, 2, n)
+
+  !$acc parallel loop gang (static:5) num_gangs (10)
+  do i = 1, n
+     a(i) = b(i) + 5
+  end do
+  !$acc end parallel loop
+
+  call test (a, b, 5, n)
+
+  !$acc parallel loop gang (static:20) num_gangs (10)
+  do i = 1, n
+     a(i) = b(i) + 20
+  end do
+  !$acc end parallel loop
+
+  call test (a, b, 20, n)
+
+  x = 5
+  !$acc parallel loop gang (static:0+x) num_gangs (10)
+  do i = 1, n
+     a(i) = b(i) + 5
+  end do
+  !$acc end parallel loop
+
+  call test (a, b, 5, n)
+
+  x = 10
+  !$acc parallel loop gang (static:x) num_gangs (10)
+  do i = 1, n
+     a(i) = b(i) + 10
+  end do
+  !$acc end parallel loop
+
+  call test (a, b, 10, n)
+end program main
+
+subroutine test (a, b, sarg, n)
+  integer n
+  integer a (n), b(n), sarg
+  integer i
+
+  do i = 1, n
+     if (a(i) .ne. b(i) + sarg) call abort ()
+  end do
+end subroutine test
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/if-1.f90 b/libgomp/testsuite/libgomp.oacc-fortran/if-1.f90
new file mode 100644
index 0000000..44055e1
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/if-1.f90
@@ -0,0 +1,886 @@
+! { dg-do run }
+! { dg-additional-options "-cpp" }
+
+program main
+  use openacc
+  implicit none
+
+  integer, parameter :: N = 8
+  integer, parameter :: one = 1
+  integer, parameter :: zero = 0
+  integer i, nn
+  real, allocatable :: a(:), b(:)
+  real exp, exp2
+
+  i = 0
+
+  allocate (a(N))
+  allocate (b(N))
+
+  a(:) = 4.0
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if (1 == 1)
+     do i = 1, N
+        if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+          b(i) = a(i) + 1
+        else
+          b(i) = a(i)
+        end if
+     end do
+  !$acc end parallel
+
+#if ACC_MEM_SHARED
+  exp = 5.0
+#else
+  exp = 4.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 16.0
+
+  !$acc parallel if (0 == 1)
+     do i = 1, N
+       if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+         b(i) = a(i) + 1
+       else
+         b(i) = a(i)
+       end if
+     end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 17.0) call abort
+  end do
+
+  a(:) = 8.0
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if (one == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+#if ACC_MEM_SHARED
+  exp = 9.0
+#else
+  exp = 8.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 22.0
+
+  !$acc parallel if (zero == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 23.0) call abort
+  end do
+
+  a(:) = 16.0
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if (.TRUE.)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+#if ACC_MEM_SHARED
+  exp = 17.0;
+#else
+  exp = 16.0;
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 76.0
+
+  !$acc parallel if (.FALSE.)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 77.0) call abort
+  end do
+
+  a(:) = 22.0
+
+  nn = 1
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if (nn == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+#if ACC_MEM_SHARED
+  exp = 23.0;
+#else
+  exp = 22.0;
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 18.0
+
+  nn = 0
+
+  !$acc parallel if (nn == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 19.0) call abort
+  end do
+
+  a(:) = 49.0
+
+  nn = 1
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if ((nn + nn) > 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+#if ACC_MEM_SHARED
+  exp = 50.0
+#else
+  exp = 49.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 38.0
+
+  nn = 0;
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if ((nn + nn) > 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 39.0) call abort
+  end do
+
+  a(:) = 91.0
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if (-2 > 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 92.0) call abort
+  end do
+
+  a(:) = 43.0
+
+  !$acc parallel copyin (a(1:N)) copyout (b(1:N)) if (one == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+#if ACC_MEM_SHARED
+  exp = 44.0
+#else
+  exp = 43.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 87.0
+
+  !$acc parallel if (one == 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end parallel
+
+  do i = 1, N
+    if (b(i) .ne. 88.0) call abort
+  end do
+
+  a(:) = 3.0
+  b(:) = 9.0
+
+#if ACC_MEM_SHARED
+  exp = 0.0
+  exp2 = 0.0
+#else
+  call acc_copyin (a, sizeof (a))
+  call acc_copyin (b, sizeof (b))
+  exp = 3.0;
+  exp2 = 9.0;
+#endif
+
+  !$acc update device (a(1:N), b(1:N)) if (1 == 1)
+
+  a(:) = 0.0
+  b(:) = 0.0
+
+  !$acc update host (a(1:N), b(1:N)) if (1 == 1)
+
+  do i = 1, N
+    if (a(i) .ne. exp) call abort
+    if (b(i) .ne. exp2) call abort
+  end do
+
+  a(:) = 6.0
+  b(:) = 12.0
+
+  !$acc update device (a(1:N), b(1:N)) if (0 == 1)
+
+  a(:) = 0.0
+  b(:) = 0.0
+
+  !$acc update host (a(1:N), b(1:N)) if (1 == 1)
+
+  do i = 1, N
+    if (a(i) .ne. exp) call abort
+    if (b(i) .ne. exp2) call abort
+  end do
+
+  a(:) = 26.0
+  b(:) = 21.0
+
+  !$acc update device (a(1:N), b(1:N)) if (1 == 1)
+
+  a(:) = 0.0
+  b(:) = 0.0
+
+  !$acc update host (a(1:N), b(1:N)) if (0 == 1)
+
+  do i = 1, N
+    if (a(i) .ne. 0.0) call abort
+    if (b(i) .ne. 0.0) call abort
+  end do
+
+#if !ACC_MEM_SHARED
+  call acc_copyout (a, sizeof (a))
+  call acc_copyout (b, sizeof (b))
+#endif
+
+  a(:) = 4.0
+  b(:) = 0.0
+
+  !$acc data copyin (a(1:N)) copyout (b(1:N)) if (1 == 1)
+
+    !$acc parallel present (a(1:N))
+       do i = 1, N
+           b(i) = a(i)
+       end do
+    !$acc end parallel
+  !$acc end data
+
+  do i = 1, N
+    if (b(i) .ne. 4.0) call abort
+  end do
+
+  a(:) = 8.0
+  b(:) = 1.0
+
+  !$acc data copyin (a(1:N)) copyout (b(1:N)) if (0 == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc end data
+
+  a(:) = 18.0
+  b(:) = 21.0
+
+  !$acc data copyin (a(1:N)) if (1 == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (a) .eqv. .FALSE.) call abort
+#endif
+
+    !$acc data copyout (b(1:N)) if (0 == 1)
+#if !ACC_MEM_SHARED
+      if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+        !$acc data copyout (b(1:N)) if (1 == 1)
+
+        !$acc parallel present (a(1:N)) present (b(1:N))
+          do i = 1, N
+            b(i) = a(i)
+          end do
+      !$acc end parallel
+
+    !$acc end data
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+    !$acc end data
+  !$acc end data
+
+  do i = 1, N
+   if (b(1) .ne. 18.0) call abort
+  end do
+
+  !$acc enter data copyin (b(1:N)) if (0 == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (0 == 1)
+
+  !$acc enter data copyin (b(1:N)) if (1 == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .FALSE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (1 == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc enter data copyin (b(1:N)) if (zero == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (zero == 1)
+
+  !$acc enter data copyin (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .FALSE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc enter data copyin (b(1:N)) if (one == 0)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (one == 0)
+
+  !$acc enter data copyin (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .FALSE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  a(:) = 4.0
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if (1 == 1)
+     do i = 1, N
+        if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+          b(i) = a(i) + 1
+        else
+          b(i) = a(i)
+        end if
+     end do
+  !$acc end kernels
+
+#if ACC_MEM_SHARED
+  exp = 5.0
+#else
+  exp = 4.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 16.0
+
+  !$acc kernels if (0 == 1)
+     do i = 1, N
+       if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+         b(i) = a(i) + 1
+       else
+         b(i) = a(i)
+       end if
+     end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (b(i) .ne. 17.0) call abort
+  end do
+
+  a(:) = 8.0
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if (one == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+#if ACC_MEM_SHARED
+  exp = 9.0
+#else
+  exp = 8.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 22.0
+
+  !$acc kernels if (zero == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (b(i) .ne. 23.0) call abort
+  end do
+
+  a(:) = 16.0
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if (.TRUE.)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+#if ACC_MEM_SHARED
+  exp = 17.0;
+#else
+  exp = 16.0;
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 76.0
+
+  !$acc kernels if (.FALSE.)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (b(i) .ne. 77.0) call abort
+  end do
+
+  a(:) = 22.0
+
+  nn = 1
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if (nn == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+#if ACC_MEM_SHARED
+  exp = 23.0;
+#else
+  exp = 22.0;
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 18.0
+
+  nn = 0
+
+  !$acc kernels if (nn == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (b(i) .ne. 19.0) call abort
+  end do
+
+  a(:) = 49.0
+
+  nn = 1
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if ((nn + nn) > 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+#if ACC_MEM_SHARED
+  exp = 50.0
+#else
+  exp = 49.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 38.0
+
+  nn = 0;
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if ((nn + nn) > 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (b(i) .ne. 39.0) call abort
+  end do
+
+  a(:) = 91.0
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if (-2 > 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (b(i) .ne. 92.0) call abort
+  end do
+
+  a(:) = 43.0
+
+  !$acc kernels copyin (a(1:N)) copyout (b(1:N)) if (one == 1)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+#if ACC_MEM_SHARED
+  exp = 44.0
+#else
+  exp = 43.0
+#endif
+
+  do i = 1, N
+    if (b(i) .ne. exp) call abort
+  end do
+
+  a(:) = 87.0
+
+  !$acc kernels if (one == 0)
+    do i = 1, N
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) then
+        b(i) = a(i) + 1
+      else
+        b(i) = a(i)
+      end if
+    end do
+  !$acc end kernels
+
+  do i = 1, N
+    if (b(i) .ne. 88.0) call abort
+  end do
+
+  a(:) = 3.0
+  b(:) = 9.0
+
+#if ACC_MEM_SHARED
+  exp = 0.0
+  exp2 = 0.0
+#else
+  call acc_copyin (a, sizeof (a))
+  call acc_copyin (b, sizeof (b))
+  exp = 3.0;
+  exp2 = 9.0;
+#endif
+
+  !$acc update device (a(1:N), b(1:N)) if (1 == 1)
+
+  a(:) = 0.0
+  b(:) = 0.0
+
+  !$acc update host (a(1:N), b(1:N)) if (1 == 1)
+
+  do i = 1, N
+    if (a(i) .ne. exp) call abort
+    if (b(i) .ne. exp2) call abort
+  end do
+
+  a(:) = 6.0
+  b(:) = 12.0
+
+  !$acc update device (a(1:N), b(1:N)) if (0 == 1)
+
+  a(:) = 0.0
+  b(:) = 0.0
+
+  !$acc update host (a(1:N), b(1:N)) if (1 == 1)
+
+  do i = 1, N
+    if (a(i) .ne. exp) call abort
+    if (b(i) .ne. exp2) call abort
+  end do
+
+  a(:) = 26.0
+  b(:) = 21.0
+
+  !$acc update device (a(1:N), b(1:N)) if (1 == 1)
+
+  a(:) = 0.0
+  b(:) = 0.0
+
+  !$acc update host (a(1:N), b(1:N)) if (0 == 1)
+
+  do i = 1, N
+    if (a(i) .ne. 0.0) call abort
+    if (b(i) .ne. 0.0) call abort
+  end do
+
+#if !ACC_MEM_SHARED
+  call acc_copyout (a, sizeof (a))
+  call acc_copyout (b, sizeof (b))
+#endif
+
+  a(:) = 4.0
+  b(:) = 0.0
+
+  !$acc data copyin (a(1:N)) copyout (b(1:N)) if (1 == 1)
+
+    !$acc kernels present (a(1:N))
+       do i = 1, N
+           b(i) = a(i)
+       end do
+    !$acc end kernels
+  !$acc end data
+
+  do i = 1, N
+    if (b(i) .ne. 4.0) call abort
+  end do
+
+  a(:) = 8.0
+  b(:) = 1.0
+
+  !$acc data copyin (a(1:N)) copyout (b(1:N)) if (0 == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (a) .eqv. .TRUE.) call abort
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc end data
+
+  a(:) = 18.0
+  b(:) = 21.0
+
+  !$acc data copyin (a(1:N)) if (1 == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (a) .eqv. .FALSE.) call abort
+#endif
+
+    !$acc data copyout (b(1:N)) if (0 == 1)
+#if !ACC_MEM_SHARED
+      if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+        !$acc data copyout (b(1:N)) if (1 == 1)
+
+        !$acc kernels present (a(1:N)) present (b(1:N))
+          do i = 1, N
+            b(i) = a(i)
+          end do
+      !$acc end kernels
+
+    !$acc end data
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+    !$acc end data
+  !$acc end data
+
+  do i = 1, N
+   if (b(1) .ne. 18.0) call abort
+  end do
+
+  !$acc enter data copyin (b(1:N)) if (0 == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (0 == 1)
+
+  !$acc enter data copyin (b(1:N)) if (1 == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .FALSE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (1 == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc enter data copyin (b(1:N)) if (zero == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (zero == 1)
+
+  !$acc enter data copyin (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .FALSE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc enter data copyin (b(1:N)) if (one == 0)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (one == 0)
+
+  !$acc enter data copyin (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+    if (acc_is_present (b) .eqv. .FALSE.) call abort
+#endif
+
+  !$acc exit data delete (b(1:N)) if (one == 1)
+
+#if !ACC_MEM_SHARED
+  if (acc_is_present (b) .eqv. .TRUE.) call abort
+#endif
+
+end program main
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/implicit-firstprivate-ref.f90 b/libgomp/testsuite/libgomp.oacc-fortran/implicit-firstprivate-ref.f90
new file mode 100644
index 0000000..a5f3840
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/implicit-firstprivate-ref.f90
@@ -0,0 +1,42 @@
+! This test checks if the runtime can properly handle implicit
+! firstprivate varaibles inside subroutines in modules.
+
+! { dg-do run }
+
+module test_mod
+  contains
+    subroutine test(x)
+
+      IMPLICIT NONE
+
+      INTEGER      :: x, y, j
+
+      x = 5
+
+      !$ACC PARALLEL LOOP copyout (y)
+      DO j=1,10
+         y=x
+      ENDDO
+      !$ACC END PARALLEL LOOP
+
+      y = -1;
+
+      !$ACC PARALLEL LOOP firstprivate (y) copyout (x)
+      DO j=1,10
+         x=y
+      ENDDO
+      !$ACC END PARALLEL LOOP
+    end subroutine test
+end module test_mod
+
+program t
+  use test_mod
+
+  INTEGER      :: x_min
+
+  x_min = 8
+
+  CALL test(x_min)
+
+  if (x_min .ne. -1) call abort
+end program t
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/pr68813.f90 b/libgomp/testsuite/libgomp.oacc-fortran/pr68813.f90
new file mode 100644
index 0000000..735350f
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/pr68813.f90
@@ -0,0 +1,19 @@
+program foo
+  implicit none
+  integer, parameter :: n = 100
+  integer, dimension(n,n) :: a
+  integer :: i, j, sum = 0
+
+  a = 1
+
+  !$acc parallel copyin(a(1:n,1:n)) firstprivate (sum)
+  !$acc loop gang reduction(+:sum)
+  do i=1, n
+     !$acc loop vector reduction(+:sum)
+     do j=1, n
+        sum = sum + a(i, j)
+     enddo
+  enddo
+  !$acc end parallel
+
+end program foo
diff --git a/libgomp/testsuite/libgomp.oacc-fortran/private-variables.f90 b/libgomp/testsuite/libgomp.oacc-fortran/private-variables.f90
new file mode 100644
index 0000000..3c1940b
--- /dev/null
+++ b/libgomp/testsuite/libgomp.oacc-fortran/private-variables.f90
@@ -0,0 +1,544 @@
+! Miscellaneous tests for private variables.
+
+! { dg-do run }
+
+
+! Test of gang-private variables declared on loop directive.
+
+subroutine t1()
+  integer :: x, i, arr(32)
+
+  do i = 1, 32
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang private(x)
+  do i = 1, 32
+     x = i * 2;
+     arr(i) = arr(i) + x
+  end do
+  !$acc end parallel
+
+  do i = 1, 32
+     if (arr(i) .ne. i * 3) call abort
+  end do
+end subroutine t1
+
+
+! Test of gang-private variables declared on loop directive, with broadcasting
+! to partitioned workers.
+
+subroutine t2()
+  integer :: x, i, j, arr(0:32*32)
+
+  do i = 0, 32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang private(x)
+  do i = 0, 31
+     x = i * 2;
+
+     !$acc loop worker
+     do j = 0, 31
+        arr(i * 32 + j) = arr(i * 32 + j) + x
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 * 32 - 1
+     if (arr(i) .ne. i + (i / 32) * 2) call abort
+  end do
+end subroutine t2
+
+
+! Test of gang-private variables declared on loop directive, with broadcasting
+! to partitioned vectors.
+
+subroutine t3()
+  integer :: x, i, j, arr(0:32*32)
+
+  do i = 0, 32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang private(x)
+  do i = 0, 31
+     x = i * 2;
+
+     !$acc loop vector
+     do j = 0, 31
+        arr(i * 32 + j) = arr(i * 32 + j) + x
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 * 32 - 1
+     if (arr(i) .ne. i + (i / 32) * 2) call abort
+  end do
+end subroutine t3
+
+
+! Test of gang-private addressable variable declared on loop directive, with
+! broadcasting to partitioned workers.
+
+subroutine t4()
+  type vec3
+     integer x, y, z, attr(13)
+  end type vec3
+
+  integer i, j, arr(0:32*32)
+  type(vec3) pt
+  
+  do i = 0, 32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang private(pt)
+  do i = 0, 31
+     pt%x = i
+     pt%y = i * 2
+     pt%z = i * 4
+     pt%attr(5) = i * 6
+
+     !$acc loop vector
+     do j = 0, 31
+        arr(i * 32 + j) = arr(i * 32 + j) + pt%x + pt%y + pt%z + pt%attr(5);
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 * 32 - 1
+     if (arr(i) .ne. i + (i / 32) * 13) call abort
+  end do
+end subroutine t4
+
+
+! Test of vector-private variables declared on loop directive.
+
+subroutine t5()
+  integer :: x, i, j, k, idx, arr(0:32*32*32)
+
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker
+     do j = 0, 31
+        !$acc loop vector private(x)
+        do k = 0, 31
+           x = ieor(i, j * 3)
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+        !$acc loop vector private(x)
+        do k = 0, 31
+           x = ior(i, j * 5)
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k + ior(i, j * 5) * k) then
+              call abort
+           end if
+        end do
+     end do
+  end do
+end subroutine t5
+
+
+! Test of vector-private variables declared on loop directive. Array type.
+
+subroutine t6()
+  integer :: i, j, k, idx, arr(0:32*32*32), pt(2)
+
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker
+     do j = 0, 31
+        !$acc loop vector private(x, pt)
+        do k = 0, 31
+           pt(1) = ieor(i, j * 3)
+           pt(2) = ior(i, j * 5)
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + pt(1) * k
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + pt(2) * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k + ior(i, j * 5) * k) then
+              call abort
+           end if
+        end do
+     end do
+  end do
+end subroutine t6
+
+
+! Test of worker-private variables declared on a loop directive.
+
+subroutine t7()
+  integer :: x, i, j, arr(0:32*32)
+  common x
+
+  do i = 0, 32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang private(x)
+  do i = 0, 31
+     !$acc loop worker private(x)
+     do j = 0, 31
+        x = ieor(i, j * 3)
+        arr(i * 32 + j) = arr(i * 32 + j) + x
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 * 32 - 1
+     if (arr(i) .ne. i + ieor(i / 32, mod(i, 32) * 3)) call abort
+  end do
+end subroutine t7
+
+
+! Test of worker-private variables declared on a loop directive, broadcasting
+! to vector-partitioned mode.
+
+subroutine t8()
+  integer :: x, i, j, k, idx, arr(0:32*32*32)
+
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker private(x)
+     do j = 0, 31
+        x = ieor(i, j * 3)
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k) call abort
+        end do
+     end do
+  end do
+end subroutine t8
+
+
+! Test of worker-private variables declared on a loop directive, broadcasting
+! to vector-partitioned mode.  Back-to-back worker loops.
+
+subroutine t9()
+  integer :: x, i, j, k, idx, arr(0:32*32*32)
+
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker private(x)
+     do j = 0, 31
+        x = ieor(i, j * 3)
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+     end do
+
+     !$acc loop worker private(x)
+     do j = 0, 31
+        x = ior(i, j * 5)
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k + ior(i, j * 5) * k) then
+              call abort
+           end if
+        end do
+     end do
+  end do
+end subroutine t9
+
+
+! Test of worker-private variables declared on a loop directive, broadcasting
+! to vector-partitioned mode.  Successive vector loops.  */
+
+subroutine t10()
+  integer :: x, i, j, k, idx, arr(0:32*32*32)
+
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker private(x)
+     do j = 0, 31
+        x = ieor(i, j * 3)
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+
+        x = ior(i, j * 5)
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k + ior(i, j * 5) * k) then
+              call abort
+           end if
+        end do
+     end do
+  end do
+end subroutine t10
+
+
+! Test of worker-private variables declared on a loop directive, broadcasting
+! to vector-partitioned mode.  Addressable worker variable.
+
+subroutine t11()
+  integer :: i, j, k, idx, arr(0:32*32*32)
+  integer, target :: x
+  integer, pointer :: p
+
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker private(x, p)
+     do j = 0, 31
+        p => x
+        x = ieor(i, j * 3)
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+
+        p = ior(i, j * 5)
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + x * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k + ior(i, j * 5) * k) then
+              call abort
+           end if
+        end do
+     end do
+  end do
+end subroutine t11
+
+
+! Test of worker-private variables declared on a loop directive, broadcasting
+! to vector-partitioned mode.  Aggregate worker variable.
+
+subroutine t12()
+  type vec2
+     integer x, y
+  end type vec2
+  
+  integer :: i, j, k, idx, arr(0:32*32*32)
+  type(vec2) :: pt
+  
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker private(pt)
+     do j = 0, 31
+        pt%x = ieor(i, j * 3)
+        pt%y = ior(i, j * 5)
+        
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + pt%x * k
+        end do
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + pt%y * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k + ior(i, j * 5) * k) then
+              call abort
+           end if
+        end do
+     end do
+  end do
+end subroutine t12
+
+
+! Test of worker-private variables declared on loop directive, broadcasting
+! to vector-partitioned mode.  Array worker variable.
+
+subroutine t13()
+  integer :: i, j, k, idx, arr(0:32*32*32), pt(2)
+
+  do i = 0, 32*32*32-1
+     arr(i) = i
+  end do
+
+  !$acc parallel copy(arr) num_gangs(32) num_workers(8) vector_length(32)
+  !$acc loop gang
+  do i = 0, 31
+     !$acc loop worker private(pt)
+     do j = 0, 31
+        pt(1) = ieor(i, j * 3)
+        pt(2) = ior(i, j * 5)
+        
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + pt(1) * k
+        end do
+
+        !$acc loop vector
+        do k = 0, 31
+           arr(i * 1024 + j * 32 + k) = arr(i * 1024 + j * 32 + k) + pt(2) * k
+        end do
+     end do
+  end do
+  !$acc end parallel
+
+  do i = 0, 32 - 1
+     do j = 0, 32 -1
+        do k = 0, 32 - 1
+           idx = i * 1024 + j * 32 + k
+           if (arr(idx) .ne. idx + ieor(i, j * 3) * k + ior(i, j * 5) * k) then
+              call abort
+           end if
+        end do
+     end do
+  end do
+end subroutine t13
+
+
+! Test of gang-private variables declared on the parallel directive.
+
+subroutine t14()
+  use openacc
+  integer :: x = 5
+  integer, parameter :: n = 32
+  integer :: arr(n)
+
+  do i = 1, n
+    arr(i) = 3
+  end do
+
+  !$acc parallel private(x) copy(arr) num_gangs(n) num_workers(8) vector_length(32)
+    !$acc loop gang(static:1)
+    do i = 1, n
+      x = i * 2;
+    end do
+
+   !$acc loop gang(static:1)
+    do i = 1, n
+      if (acc_on_device (acc_device_host) .eqv. .TRUE.) x = i * 2
+      arr(i) = arr(i) + x
+    end do
+  !$acc end parallel
+
+  do i = 1, n
+    if (arr(i) .ne. (3 + i * 2)) call abort
+  end do
+
+end subroutine t14
+
+
+program main
+  call t1()
+  call t2()
+  call t3()
+  call t4()
+  call t5()
+  call t6()
+  call t7()
+  call t8()
+  call t9()
+  call t10()
+  call t11()
+  call t12()
+  call t13()
+  call t14()
+end program main