[MLIR][OpenMP] Refactor bounds offsetting and fix to apply to all directives (#84349)
This PR refactors bounds offsetting by combining the two differing
implementations (one applying to initial derived type member map
implementation for descriptors and the other for regular arrays,
effectively allocatable array vs regular array in fortran) now that it's
a little simpler to do.
The PR also moves the utilization of createAlteredByCaptureMap into
genMapInfoOp, where it will be correctly applied to all MapInfoData,
appropriately offsetting and altering Pointer data set in the kernel
argument structure on the host. This primarily means bounds offsets will
now correctly apply to enter/exit/update map clauses as opposed to just
the Target directive that is currently the case. A few fortran runtime
tests have been added to verify this new behavior.
This PR depends on: https://github.com/llvm/llvm-project/pull/84328 and
is an extraction of the larger derived type member map PR stack (so a
requirement for it to land).
GitOrigin-RevId: 8612fa0d84c730a753d04de012a8372ba5a10677
diff --git a/libomptarget/test/offloading/fortran/target-map-enter-exit-array-2.f90 b/libomptarget/test/offloading/fortran/target-map-enter-exit-array-2.f90
new file mode 100644
index 0000000..489c253
--- /dev/null
+++ b/libomptarget/test/offloading/fortran/target-map-enter-exit-array-2.f90
@@ -0,0 +1,39 @@
+! Offloading test checking interaction of an
+! enter and exit map of an array of scalars
+! REQUIRES: flang, amdgcn-amd-amdhsa
+! UNSUPPORTED: nvptx64-nvidia-cuda
+! UNSUPPORTED: nvptx64-nvidia-cuda-LTO
+! UNSUPPORTED: aarch64-unknown-linux-gnu
+! UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
+! UNSUPPORTED: x86_64-pc-linux-gnu
+! UNSUPPORTED: x86_64-pc-linux-gnu-LTO
+
+! RUN: %libomptarget-compile-fortran-run-and-check-generic
+program main
+ integer :: array(10)
+
+ do I = 1, 10
+ array(I) = I + I
+ end do
+
+ !$omp target enter data map(to: array)
+
+ ! Shouldn't overwrite data already locked in
+ ! on target via enter, this will then be
+ ! overwritten by our exit
+ do I = 1, 10
+ array(I) = 10
+ end do
+
+ !$omp target
+ do i=1,10
+ array(i) = array(i) + i
+ end do
+ !$omp end target
+
+ !$omp target exit data map(from: array)
+
+ print*, array
+end program
+
+!CHECK: 3 6 9 12 15 18 21 24 27 30
diff --git a/libomptarget/test/offloading/fortran/target-map-enter-exit-array-bounds.f90 b/libomptarget/test/offloading/fortran/target-map-enter-exit-array-bounds.f90
new file mode 100644
index 0000000..3c8c350
--- /dev/null
+++ b/libomptarget/test/offloading/fortran/target-map-enter-exit-array-bounds.f90
@@ -0,0 +1,44 @@
+! Offloading test checking interaction of an
+! enter and exit map of an array of scalars
+! with specified bounds
+! REQUIRES: flang, amdgcn-amd-amdhsa
+! UNSUPPORTED: nvptx64-nvidia-cuda
+! UNSUPPORTED: nvptx64-nvidia-cuda-LTO
+! UNSUPPORTED: aarch64-unknown-linux-gnu
+! UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
+! UNSUPPORTED: x86_64-pc-linux-gnu
+! UNSUPPORTED: x86_64-pc-linux-gnu-LTO
+
+! RUN: %libomptarget-compile-fortran-run-and-check-generic
+
+program main
+ integer :: array(10)
+
+ do I = 1, 10
+ array(I) = I + I
+ end do
+
+ !$omp target enter data map(to: array(3:6))
+
+ ! Shouldn't overwrite data already locked in
+ ! on target via enter, which will then be
+ ! overwritten by our exit
+ do I = 1, 10
+ array(I) = 10
+ end do
+
+ ! The compiler/runtime is less lenient about read/write out of
+ ! bounds when using enter and exit, we have to specifically loop
+ ! over the correctly mapped range
+ !$omp target
+ do i=3,6
+ array(i) = array(i) + i
+ end do
+ !$omp end target
+
+ !$omp target exit data map(from: array(3:6))
+
+ print *, array
+end program
+
+!CHECK: 10 10 9 12 15 18 10 10 10 10
diff --git a/libomptarget/test/offloading/fortran/target-map-enter-exit-scalar.f90 b/libomptarget/test/offloading/fortran/target-map-enter-exit-scalar.f90
new file mode 100644
index 0000000..29a0b5e
--- /dev/null
+++ b/libomptarget/test/offloading/fortran/target-map-enter-exit-scalar.f90
@@ -0,0 +1,33 @@
+! Offloading test checking interaction of an
+! enter and exit map of an scalar
+! REQUIRES: flang, amdgcn-amd-amdhsa
+! UNSUPPORTED: nvptx64-nvidia-cuda
+! UNSUPPORTED: nvptx64-nvidia-cuda-LTO
+! UNSUPPORTED: aarch64-unknown-linux-gnu
+! UNSUPPORTED: aarch64-unknown-linux-gnu-LTO
+! UNSUPPORTED: x86_64-pc-linux-gnu
+! UNSUPPORTED: x86_64-pc-linux-gnu-LTO
+
+! RUN: %libomptarget-compile-fortran-run-and-check-generic
+program main
+ integer :: scalar
+ scalar = 10
+
+ !$omp target enter data map(to: scalar)
+
+ !ignored, as we've already attached
+ scalar = 20
+
+ !$omp target
+ scalar = scalar + 50
+ !$omp end target
+
+ !$omp target exit data map(from: scalar)
+
+ ! not the answer one may expect, but it is the same
+ ! answer Clang gives so we are correctly on par with
+ ! Clang for the moment.
+ print *, scalar
+end program
+
+!CHECK: 10
