mirrors/llvm-project
0
0
Fork 0
mirror of https://github.com/llvm/llvm-project.git synced 2025-04-16 11:26:42 +00:00
Code Issues Projects Releases Wiki Activity
llvm-project/flang/test/Lower/OpenACC/acc-host-data.f90

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

54 lines
2.2 KiB
Fortran
Raw Normal View History

[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
! This test checks lowering of OpenACC host_data directive.
[flang][openacc][NFC] Check only HLFIR lowering for compute construct tests (#73051)
2023-11-21 15:53:30 -08:00
! RUN: bbc -fopenacc -emit-hlfir %s -o - | FileCheck %s
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
subroutine acc_host_data()
real, dimension(10) :: a
logical :: ifCondition = .TRUE.
! CHECK: %[[A:.*]] = fir.alloca !fir.array<10xf32> {bindc_name = "a", uniq_name = "_QFacc_host_dataEa"}
[flang][openacc][NFC] Check only HLFIR lowering for compute construct tests (#73051)
2023-11-21 15:53:30 -08:00
! CHECK: %[[DECLA:.*]]:2 = hlfir.declare %[[A]]
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
! CHECK: %[[IFCOND:.*]] = fir.address_of(@_QFacc_host_dataEifcondition) : !fir.ref<!fir.logical<4>>
[flang][openacc][NFC] Check only HLFIR lowering for compute construct tests (#73051)
2023-11-21 15:53:30 -08:00
! CHECK: %[[DECLIFCOND:.*]]:2 = hlfir.declare %[[IFCOND]]
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
!$acc host_data use_device(a)
!$acc end host_data
[flang][acc] Improve acc lowering around fir.box and arrays (#125600) The current implementation of OpenACC lowering includes explicit expansion of following cases: - Creation of `acc.bounds` operations for all arrays, including those whose dimensions are captured in the type (eg `!fir.array<100xf32>`) - Expansion of box types by only putting the box's address in the data clause. The address was extracted with a `fir.box_addr` operation and the bounds were filled with `fir.box_dims` operation. However, with the creation of the new type interface `MappableType`, the idea is that specific type-based semantics can now be used. This also really simplifies representation in the IR. Consider the following example: ``` subroutine sub(arr) real :: arr(:) !$acc enter data copyin(arr) end subroutine ``` Before the current PR, the relevant acc dialect IR looked like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %c1 = arith.constant 1 : index %c0 = arith.constant 0 : index %2:3 = fir.box_dims %1#0, %c0 : (!fir.box<!fir.array<?xf32>>, index) -> (index, index, index) %c0_0 = arith.constant 0 : index %3 = arith.subi %2#1, %c1 : index %4 = acc.bounds lowerbound(%c0_0 : index) upperbound(%3 : index) extent(%2#1 : index) stride(%2#2 : index) startIdx(%c1 : index) {strideInBytes = true} %5 = fir.box_addr %1#0 : (!fir.box<!fir.array<?xf32>>) -> !fir.ref<!fir.array<?xf32>> %6 = acc.copyin varPtr(%5 : !fir.ref<!fir.array<?xf32>>) bounds(%4) -> !fir.ref<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%6 : !fir.ref<!fir.array<?xf32>>) ``` After the current change, it looks like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %2 = acc.copyin var(%1#0 : !fir.box<!fir.array<?xf32>>) -> !fir.box<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%2 : !fir.box<!fir.array<?xf32>>) ``` Restoring the old behavior can be done with following command line options: `--openacc-unwrap-fir-box=true --openacc-generate-default-bounds=true`
2025-02-04 08:08:16 -08:00
! CHECK: %[[DA:.*]] = acc.use_device varPtr(%[[DECLA]]#0 : !fir.ref<!fir.array<10xf32>>) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
! CHECK: acc.host_data dataOperands(%[[DA]] : !fir.ref<!fir.array<10xf32>>)
!$acc host_data use_device(a) if_present
!$acc end host_data
[flang][acc] Improve acc lowering around fir.box and arrays (#125600) The current implementation of OpenACC lowering includes explicit expansion of following cases: - Creation of `acc.bounds` operations for all arrays, including those whose dimensions are captured in the type (eg `!fir.array<100xf32>`) - Expansion of box types by only putting the box's address in the data clause. The address was extracted with a `fir.box_addr` operation and the bounds were filled with `fir.box_dims` operation. However, with the creation of the new type interface `MappableType`, the idea is that specific type-based semantics can now be used. This also really simplifies representation in the IR. Consider the following example: ``` subroutine sub(arr) real :: arr(:) !$acc enter data copyin(arr) end subroutine ``` Before the current PR, the relevant acc dialect IR looked like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %c1 = arith.constant 1 : index %c0 = arith.constant 0 : index %2:3 = fir.box_dims %1#0, %c0 : (!fir.box<!fir.array<?xf32>>, index) -> (index, index, index) %c0_0 = arith.constant 0 : index %3 = arith.subi %2#1, %c1 : index %4 = acc.bounds lowerbound(%c0_0 : index) upperbound(%3 : index) extent(%2#1 : index) stride(%2#2 : index) startIdx(%c1 : index) {strideInBytes = true} %5 = fir.box_addr %1#0 : (!fir.box<!fir.array<?xf32>>) -> !fir.ref<!fir.array<?xf32>> %6 = acc.copyin varPtr(%5 : !fir.ref<!fir.array<?xf32>>) bounds(%4) -> !fir.ref<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%6 : !fir.ref<!fir.array<?xf32>>) ``` After the current change, it looks like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %2 = acc.copyin var(%1#0 : !fir.box<!fir.array<?xf32>>) -> !fir.box<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%2 : !fir.box<!fir.array<?xf32>>) ``` Restoring the old behavior can be done with following command line options: `--openacc-unwrap-fir-box=true --openacc-generate-default-bounds=true`
2025-02-04 08:08:16 -08:00
! CHECK: %[[DA:.*]] = acc.use_device varPtr(%[[DECLA]]#0 : !fir.ref<!fir.array<10xf32>>) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
! CHECK: acc.host_data dataOperands(%[[DA]] : !fir.ref<!fir.array<10xf32>>) {
[flang][openacc] Allow if_present multiple times on host_data and update (#135422) Similar to #135415. The spec has not strict restriction to allow a single `if_present` clause on the host_data and update directives. Allowing this clause multiple times does not change the semantic of it. This patch relax the rules in ACC.td since there is no restriction in the standard. The OpenACC dialect represents the `if_present` clause with a `UnitAttr` so the attribute will be set if the is one or more `if_present` clause.
2025-04-11 14:01:03 -07:00
! CHECK: } attributes {ifPresent}
!$acc host_data use_device(a) if_present if_present
!$acc end host_data
! CHECK: acc.host_data dataOperands(%{{.*}} : !fir.ref<!fir.array<10xf32>>) {
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
! CHECK: } attributes {ifPresent}
!$acc host_data use_device(a) if(ifCondition)
!$acc end host_data
[flang][acc] Improve acc lowering around fir.box and arrays (#125600) The current implementation of OpenACC lowering includes explicit expansion of following cases: - Creation of `acc.bounds` operations for all arrays, including those whose dimensions are captured in the type (eg `!fir.array<100xf32>`) - Expansion of box types by only putting the box's address in the data clause. The address was extracted with a `fir.box_addr` operation and the bounds were filled with `fir.box_dims` operation. However, with the creation of the new type interface `MappableType`, the idea is that specific type-based semantics can now be used. This also really simplifies representation in the IR. Consider the following example: ``` subroutine sub(arr) real :: arr(:) !$acc enter data copyin(arr) end subroutine ``` Before the current PR, the relevant acc dialect IR looked like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %c1 = arith.constant 1 : index %c0 = arith.constant 0 : index %2:3 = fir.box_dims %1#0, %c0 : (!fir.box<!fir.array<?xf32>>, index) -> (index, index, index) %c0_0 = arith.constant 0 : index %3 = arith.subi %2#1, %c1 : index %4 = acc.bounds lowerbound(%c0_0 : index) upperbound(%3 : index) extent(%2#1 : index) stride(%2#2 : index) startIdx(%c1 : index) {strideInBytes = true} %5 = fir.box_addr %1#0 : (!fir.box<!fir.array<?xf32>>) -> !fir.ref<!fir.array<?xf32>> %6 = acc.copyin varPtr(%5 : !fir.ref<!fir.array<?xf32>>) bounds(%4) -> !fir.ref<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%6 : !fir.ref<!fir.array<?xf32>>) ``` After the current change, it looks like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %2 = acc.copyin var(%1#0 : !fir.box<!fir.array<?xf32>>) -> !fir.box<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%2 : !fir.box<!fir.array<?xf32>>) ``` Restoring the old behavior can be done with following command line options: `--openacc-unwrap-fir-box=true --openacc-generate-default-bounds=true`
2025-02-04 08:08:16 -08:00
! CHECK: %[[DA:.*]] = acc.use_device varPtr(%[[DECLA]]#0 : !fir.ref<!fir.array<10xf32>>) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
[flang][openacc][NFC] Check only HLFIR lowering for compute construct tests (#73051)
2023-11-21 15:53:30 -08:00
! CHECK: %[[LOAD_IFCOND:.*]] = fir.load %[[DECLIFCOND]]#0 : !fir.ref<!fir.logical<4>>
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
! CHECK: %[[IFCOND_I1:.*]] = fir.convert %[[LOAD_IFCOND]] : (!fir.logical<4>) -> i1
! CHECK: acc.host_data if(%[[IFCOND_I1]]) dataOperands(%[[DA]] : !fir.ref<!fir.array<10xf32>>)
!$acc host_data use_device(a) if(.true.)
!$acc end host_data
[flang][acc] Improve acc lowering around fir.box and arrays (#125600) The current implementation of OpenACC lowering includes explicit expansion of following cases: - Creation of `acc.bounds` operations for all arrays, including those whose dimensions are captured in the type (eg `!fir.array<100xf32>`) - Expansion of box types by only putting the box's address in the data clause. The address was extracted with a `fir.box_addr` operation and the bounds were filled with `fir.box_dims` operation. However, with the creation of the new type interface `MappableType`, the idea is that specific type-based semantics can now be used. This also really simplifies representation in the IR. Consider the following example: ``` subroutine sub(arr) real :: arr(:) !$acc enter data copyin(arr) end subroutine ``` Before the current PR, the relevant acc dialect IR looked like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %c1 = arith.constant 1 : index %c0 = arith.constant 0 : index %2:3 = fir.box_dims %1#0, %c0 : (!fir.box<!fir.array<?xf32>>, index) -> (index, index, index) %c0_0 = arith.constant 0 : index %3 = arith.subi %2#1, %c1 : index %4 = acc.bounds lowerbound(%c0_0 : index) upperbound(%3 : index) extent(%2#1 : index) stride(%2#2 : index) startIdx(%c1 : index) {strideInBytes = true} %5 = fir.box_addr %1#0 : (!fir.box<!fir.array<?xf32>>) -> !fir.ref<!fir.array<?xf32>> %6 = acc.copyin varPtr(%5 : !fir.ref<!fir.array<?xf32>>) bounds(%4) -> !fir.ref<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%6 : !fir.ref<!fir.array<?xf32>>) ``` After the current change, it looks like: ``` func.func @_QPsub(%arg0: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "arr"}) { ... %1:2 = hlfir.declare %arg0 dummy_scope %0 {uniq_name = "_QFsubEarr"} : (!fir.box<!fir.array<?xf32>>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.box<!fir.array<?xf32>>) %2 = acc.copyin var(%1#0 : !fir.box<!fir.array<?xf32>>) -> !fir.box<!fir.array<?xf32>> {name = "arr", structured = false} acc.enter_data dataOperands(%2 : !fir.box<!fir.array<?xf32>>) ``` Restoring the old behavior can be done with following command line options: `--openacc-unwrap-fir-box=true --openacc-generate-default-bounds=true`
2025-02-04 08:08:16 -08:00
! CHECK: %[[DA:.*]] = acc.use_device varPtr(%[[DECLA]]#0 : !fir.ref<!fir.array<10xf32>>) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
! CHECK: acc.host_data dataOperands(%[[DA]] : !fir.ref<!fir.array<10xf32>>)
!$acc host_data use_device(a) if(.false.)
a = 1.0
!$acc end host_data
! CHECK-NOT: acc.host_data
[flang][openacc][NFC] Check only HLFIR lowering for compute construct tests (#73051)
2023-11-21 15:53:30 -08:00
! CHECK: hlfir.assign %{{.*}} to %[[DECLA]]#0
[flang][openacc] Lower host_data construct Lower host_data construct to the acc.host_data operation. Depends on D150289 Reviewed By: razvanlupusoru, jeanPerier Differential Revision: https://reviews.llvm.org/D150290
2023-05-15 11:22:12 -07:00
end subroutine
Reference in New Issue Copy Permalink