diff options
Diffstat (limited to 'flang/test')
| -rw-r--r-- | flang/test/Lower/OpenACC/Todo/do-loops-to-acc-loops-todo.f90 | 91 | ||||
| -rw-r--r-- | flang/test/Lower/OpenACC/do-loops-to-acc-loops.f90 | 267 | ||||
| -rw-r--r-- | flang/test/Lower/OpenMP/unroll-heuristic01.f90 | 63 | ||||
| -rw-r--r-- | flang/test/Lower/OpenMP/unroll-heuristic02.f90 | 98 | ||||
| -rw-r--r-- | flang/test/Lower/OpenMP/unroll-heuristic03.f90 | 61 |
5 files changed, 504 insertions, 76 deletions
diff --git a/flang/test/Lower/OpenACC/Todo/do-loops-to-acc-loops-todo.f90 b/flang/test/Lower/OpenACC/Todo/do-loops-to-acc-loops-todo.f90 new file mode 100644 index 0000000..aa1d443 --- /dev/null +++ b/flang/test/Lower/OpenACC/Todo/do-loops-to-acc-loops-todo.f90 @@ -0,0 +1,91 @@ +! RUN: split-file %s %t +! RUN: %not_todo_cmd bbc -fopenacc -emit-hlfir %t/do_loop_with_stop.f90 -o - 2>&1 | FileCheck %s --check-prefix=CHECK1 +! RUN: %not_todo_cmd bbc -fopenacc -emit-hlfir %t/do_loop_with_cycle_goto.f90 -o - 2>&1 | FileCheck %s --check-prefix=CHECK2 +! RUN: %not_todo_cmd bbc -fopenacc -emit-hlfir %t/nested_goto_loop.f90 -o - 2>&1 | FileCheck %s --check-prefix=CHECK3 +! RUN: %not_todo_cmd bbc -fopenacc -emit-hlfir %t/nested_loop_with_inner_goto.f90 -o - 2>&1 | FileCheck %s --check-prefix=CHECK4 + +//--- do_loop_with_stop.f90 + +subroutine do_loop_with_stop() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + !$acc kernels + do i = 1, n + a(i) = b(i) + 1.0 + if (i == 5) stop + end do + !$acc end kernels + +! CHECK1: not yet implemented: unstructured do loop in acc kernels + +end subroutine + +//--- do_loop_with_cycle_goto.f90 + +subroutine do_loop_with_cycle_goto() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + ! Do loop with cycle and goto - unstructured control flow is not converted. + !$acc kernels + do i = 1, n + if (i == 3) cycle + a(i) = b(i) + 1.0 + if (i == 7) goto 200 + a(i) = a(i) * 2.0 + end do +200 continue + !$acc end kernels + +! CHECK2: not yet implemented: unstructured do loop in acc kernels + +end subroutine + +//--- nested_goto_loop.f90 + +subroutine nested_goto_loop() + integer :: i, j + integer, parameter :: n = 10, m = 5 + real, dimension(n,m) :: a, b + + ! Nested loop with goto from inner to outer - should NOT convert to acc.loop + !$acc kernels + do i = 1, n + do j = 1, m + a(i,j) = b(i,j) + 1.0 + if (i * j > 20) goto 300 ! Exit both loops + end do + end do +300 continue + !$acc end kernels + +! CHECK3: not yet implemented: unstructured do loop in acc kernels + +end subroutine + +//--- nested_loop_with_inner_goto.f90 + +subroutine nested_loop_with_inner_goto() + integer :: ii = 0, jj = 0 + integer, parameter :: nn = 3 + real, dimension(nn, nn) :: aa + + aa = -1 + + ! Nested loop with goto from inner loop - unstructured control flow is not converted. + !$acc kernels + do ii = 1, nn + do jj = 1, nn + if (jj > 1) goto 300 + aa(jj, ii) = 1337 + end do + 300 continue + end do + !$acc end kernels + +! CHECK4: not yet implemented: unstructured do loop in acc kernels + +end subroutine
\ No newline at end of file diff --git a/flang/test/Lower/OpenACC/do-loops-to-acc-loops.f90 b/flang/test/Lower/OpenACC/do-loops-to-acc-loops.f90 new file mode 100644 index 0000000..5f8ea03 --- /dev/null +++ b/flang/test/Lower/OpenACC/do-loops-to-acc-loops.f90 @@ -0,0 +1,267 @@ +! This test checks lowering of Fortran do loops and do concurrent loops to OpenACC loop constructs. +! Tests the new functionality that converts Fortran iteration constructs to acc.loop with proper IV handling. + +! RUN: bbc -fopenacc -emit-hlfir %s -o - | FileCheck %s + +! CHECK-LABEL: func.func @_QPbasic_do_loop +subroutine basic_do_loop() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + ! Basic do loop that should be converted to acc.loop + !$acc kernels + do i = 1, n + a(i) = b(i) + 1.0 + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {auto_ = [#acc.device_type<none>], inclusiveUpperbound = array<i1: true>} + +end subroutine + +! CHECK-LABEL: func.func @_QPbasic_do_concurrent +subroutine basic_do_concurrent() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + ! Basic do concurrent loop + !$acc kernels + do concurrent (i = 1:n) + a(i) = b(i) + 1.0 + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {auto_ = [#acc.device_type<none>], inclusiveUpperbound = array<i1: true>} + +end subroutine + +! CHECK-LABEL: func.func @_QPbasic_do_loop_parallel +subroutine basic_do_loop_parallel() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + ! Basic do loop with acc parallel that should be converted to acc.loop + !$acc parallel + do i = 1, n + a(i) = b(i) + 1.0 + end do + !$acc end parallel + +! CHECK: acc.parallel { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {auto_ = [#acc.device_type<none>], inclusiveUpperbound = array<i1: true>} + +end subroutine + +! CHECK-LABEL: func.func @_QPbasic_do_loop_serial +subroutine basic_do_loop_serial() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + ! Basic do loop with acc serial that should be converted to acc.loop + !$acc serial + do i = 1, n + a(i) = b(i) + 1.0 + end do + !$acc end serial + +! CHECK: acc.serial { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {inclusiveUpperbound = array<i1: true>, seq = [#acc.device_type<none>]} + +end subroutine + +! CHECK-LABEL: func.func @_QPbasic_do_concurrent_parallel +subroutine basic_do_concurrent_parallel() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + ! Basic do concurrent loop with acc parallel + !$acc parallel + do concurrent (i = 1:n) + a(i) = b(i) + 1.0 + end do + !$acc end parallel + +! CHECK: acc.parallel { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {inclusiveUpperbound = array<i1: true>, independent = [#acc.device_type<none>]} + +end subroutine + +! CHECK-LABEL: func.func @_QPbasic_do_concurrent_serial +subroutine basic_do_concurrent_serial() + integer :: i + integer, parameter :: n = 10 + real, dimension(n) :: a, b + + ! Basic do concurrent loop with acc serial + !$acc serial + do concurrent (i = 1:n) + a(i) = b(i) + 1.0 + end do + !$acc end serial + +! CHECK: acc.serial { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {inclusiveUpperbound = array<i1: true>, seq = [#acc.device_type<none>]} + +end subroutine + +! CHECK-LABEL: func.func @_QPmulti_dimension_do_concurrent +subroutine multi_dimension_do_concurrent() + integer :: i, j, k + integer, parameter :: n = 10, m = 20, l = 5 + real, dimension(n,m,l) :: a, b + + ! Multi-dimensional do concurrent with multiple iteration variables + !$acc kernels + do concurrent (i = 1:n, j = 1:m, k = 1:l) + a(i,j,k) = b(i,j,k) * 2.0 + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32, %{{.*}} : i32, %{{.*}} : i32) = (%c1{{.*}}, %c1{{.*}}, %c1{{.*}} : i32, i32, i32) to (%{{.*}}, %{{.*}}, %{{.*}} : i32, i32, i32) step (%c1{{.*}}, %c1{{.*}}, %c1{{.*}} : i32, i32, i32) +! CHECK: acc.yield +! CHECK: attributes {auto_ = [#acc.device_type<none>], inclusiveUpperbound = array<i1: true, true, true>} +end subroutine + + +! CHECK-LABEL: func.func @_QPnested_do_loops +subroutine nested_do_loops() + integer :: i, j + integer, parameter :: n = 10, m = 20 + real, dimension(n,m) :: a, b + + ! Nested do loops + !$acc kernels + do i = 1, n + do j = 1, m + a(i,j) = b(i,j) + i + j + end do + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {auto_ = [#acc.device_type<none>], inclusiveUpperbound = array<i1: true>} +! CHECK: acc.yield +! CHECK: attributes {auto_ = [#acc.device_type<none>], inclusiveUpperbound = array<i1: true>} + +end subroutine + +! CHECK-LABEL: func.func @_QPvariable_bounds_and_step +subroutine variable_bounds_and_step(n, start_val, step_val) + integer, intent(in) :: n, start_val, step_val + integer :: i + real, dimension(n) :: a, b + + ! Do loop with variable bounds and step + !$acc kernels + do i = start_val, n, step_val + a(i) = b(i) * 2.0 + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.yield +! CHECK: attributes {auto_ = [#acc.device_type<none>], inclusiveUpperbound = array<i1: true>} + +end subroutine + +! CHECK-LABEL: func.func @_QPdifferent_iv_types +subroutine different_iv_types() + integer(kind=8) :: i8 + integer(kind=4) :: i4 + integer(kind=2) :: i2 + integer, parameter :: n = 10 + real, dimension(n) :: a, b, c, d + + ! Test different iteration variable types + !$acc kernels + do i8 = 1_8, int(n,8) + a(i8) = b(i8) + 1.0 + end do + !$acc end kernels + + !$acc kernels + do i4 = 1, n + b(i4) = c(i4) + 1.0 + end do + !$acc end kernels + + !$acc kernels + do i2 = 1_2, int(n,2) + c(i2) = d(i2) + 1.0 + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i64) = (%{{.*}} : i64) to (%{{.*}} : i64) step (%{{.*}} : i64) +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) +! CHECK: acc.kernels { +! CHECK: acc.loop {{.*}} control(%{{.*}} : i16) = (%{{.*}} : i16) to (%{{.*}} : i16) step (%{{.*}} : i16) + +end subroutine + +! ----------------------------------------------------------------------------------------- +! Tests for loops that should NOT be converted to acc.loop due to unstructured control flow + +! CHECK-LABEL: func.func @_QPinfinite_loop_no_iv +subroutine infinite_loop_no_iv() + integer :: i + logical :: condition + + ! Infinite loop with no induction variable - should NOT convert to acc.loop + !$acc kernels + do + i = i + 1 + if (i > 100) exit + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK-NOT: acc.loop + +end subroutine + +! CHECK-LABEL: func.func @_QPwhile_like_loop +subroutine while_like_loop() + integer :: i + logical :: condition + + i = 1 + condition = .true. + + ! While-like infinite loop - should NOT convert to acc.loop + !$acc kernels + do while (condition) + i = i + 1 + if (i > 100) condition = .false. + end do + !$acc end kernels + +! CHECK: acc.kernels { +! CHECK-NOT: acc.loop + +end subroutine diff --git a/flang/test/Lower/OpenMP/unroll-heuristic01.f90 b/flang/test/Lower/OpenMP/unroll-heuristic01.f90 index a5f5c00..34020eb 100644 --- a/flang/test/Lower/OpenMP/unroll-heuristic01.f90 +++ b/flang/test/Lower/OpenMP/unroll-heuristic01.f90 @@ -13,27 +13,42 @@ subroutine omp_unroll_heuristic01(lb, ub, inc) end subroutine omp_unroll_heuristic01 -!CHECK-LABEL: func.func @_QPomp_unroll_heuristic01( -!CHECK: %c0_i32 = arith.constant 0 : i32 -!CHECK-NEXT: %c1_i32 = arith.constant 1 : i32 -!CHECK-NEXT: %13 = arith.cmpi slt, %12, %c0_i32 : i32 -!CHECK-NEXT: %14 = arith.subi %c0_i32, %12 : i32 -!CHECK-NEXT: %15 = arith.select %13, %14, %12 : i32 -!CHECK-NEXT: %16 = arith.select %13, %11, %10 : i32 -!CHECK-NEXT: %17 = arith.select %13, %10, %11 : i32 -!CHECK-NEXT: %18 = arith.subi %17, %16 overflow<nuw> : i32 -!CHECK-NEXT: %19 = arith.divui %18, %15 : i32 -!CHECK-NEXT: %20 = arith.addi %19, %c1_i32 overflow<nuw> : i32 -!CHECK-NEXT: %21 = arith.cmpi slt, %17, %16 : i32 -!CHECK-NEXT: %22 = arith.select %21, %c0_i32, %20 : i32 -!CHECK-NEXT: %canonloop_s0 = omp.new_cli -!CHECK-NEXT: omp.canonical_loop(%canonloop_s0) %iv : i32 in range(%22) { -!CHECK-NEXT: %23 = arith.muli %iv, %12 : i32 -!CHECK-NEXT: %24 = arith.addi %10, %23 : i32 -!CHECK-NEXT: hlfir.assign %24 to %9#0 : i32, !fir.ref<i32> -!CHECK-NEXT: %25 = fir.load %9#0 : !fir.ref<i32> -!CHECK-NEXT: hlfir.assign %25 to %6#0 : i32, !fir.ref<i32> -!CHECK-NEXT: omp.terminator -!CHECK-NEXT: } -!CHECK-NEXT: omp.unroll_heuristic(%canonloop_s0) -!CHECK-NEXT: return +! CHECK-LABEL: func.func @_QPomp_unroll_heuristic01( +! CHECK-SAME: %[[ARG0:.*]]: !fir.ref<i32> {fir.bindc_name = "lb"}, +! CHECK-SAME: %[[ARG1:.*]]: !fir.ref<i32> {fir.bindc_name = "ub"}, +! CHECK-SAME: %[[ARG2:.*]]: !fir.ref<i32> {fir.bindc_name = "inc"}) { +! CHECK: %[[VAL_0:.*]] = fir.dummy_scope : !fir.dscope +! CHECK: %[[VAL_1:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFomp_unroll_heuristic01Ei"} +! CHECK: %[[VAL_2:.*]]:2 = hlfir.declare %[[VAL_1]] {uniq_name = "_QFomp_unroll_heuristic01Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_3:.*]]:2 = hlfir.declare %[[ARG2]] dummy_scope %[[VAL_0]] {uniq_name = "_QFomp_unroll_heuristic01Einc"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_4:.*]]:2 = hlfir.declare %[[ARG0]] dummy_scope %[[VAL_0]] {uniq_name = "_QFomp_unroll_heuristic01Elb"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_5:.*]] = fir.alloca i32 {bindc_name = "res", uniq_name = "_QFomp_unroll_heuristic01Eres"} +! CHECK: %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_5]] {uniq_name = "_QFomp_unroll_heuristic01Eres"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_7:.*]]:2 = hlfir.declare %[[ARG1]] dummy_scope %[[VAL_0]] {uniq_name = "_QFomp_unroll_heuristic01Eub"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_8:.*]] = fir.load %[[VAL_4]]#0 : !fir.ref<i32> +! CHECK: %[[VAL_9:.*]] = fir.load %[[VAL_7]]#0 : !fir.ref<i32> +! CHECK: %[[VAL_10:.*]] = fir.load %[[VAL_3]]#0 : !fir.ref<i32> +! CHECK: %[[VAL_11:.*]] = arith.constant 0 : i32 +! CHECK: %[[VAL_12:.*]] = arith.constant 1 : i32 +! CHECK: %[[VAL_13:.*]] = arith.cmpi slt, %[[VAL_10]], %[[VAL_11]] : i32 +! CHECK: %[[VAL_14:.*]] = arith.subi %[[VAL_11]], %[[VAL_10]] : i32 +! CHECK: %[[VAL_15:.*]] = arith.select %[[VAL_13]], %[[VAL_14]], %[[VAL_10]] : i32 +! CHECK: %[[VAL_16:.*]] = arith.select %[[VAL_13]], %[[VAL_9]], %[[VAL_8]] : i32 +! CHECK: %[[VAL_17:.*]] = arith.select %[[VAL_13]], %[[VAL_8]], %[[VAL_9]] : i32 +! CHECK: %[[VAL_18:.*]] = arith.subi %[[VAL_17]], %[[VAL_16]] overflow<nuw> : i32 +! CHECK: %[[VAL_19:.*]] = arith.divui %[[VAL_18]], %[[VAL_15]] : i32 +! CHECK: %[[VAL_20:.*]] = arith.addi %[[VAL_19]], %[[VAL_12]] overflow<nuw> : i32 +! CHECK: %[[VAL_21:.*]] = arith.cmpi slt, %[[VAL_17]], %[[VAL_16]] : i32 +! CHECK: %[[VAL_22:.*]] = arith.select %[[VAL_21]], %[[VAL_11]], %[[VAL_20]] : i32 +! CHECK: %[[VAL_23:.*]] = omp.new_cli +! CHECK: omp.canonical_loop(%[[VAL_23]]) %[[VAL_24:.*]] : i32 in range(%[[VAL_22]]) { +! CHECK: %[[VAL_25:.*]] = arith.muli %[[VAL_24]], %[[VAL_10]] : i32 +! CHECK: %[[VAL_26:.*]] = arith.addi %[[VAL_8]], %[[VAL_25]] : i32 +! CHECK: hlfir.assign %[[VAL_26]] to %[[VAL_2]]#0 : i32, !fir.ref<i32> +! CHECK: %[[VAL_27:.*]] = fir.load %[[VAL_2]]#0 : !fir.ref<i32> +! CHECK: hlfir.assign %[[VAL_27]] to %[[VAL_6]]#0 : i32, !fir.ref<i32> +! CHECK: omp.terminator +! CHECK: } +! CHECK: omp.unroll_heuristic(%[[VAL_23]]) +! CHECK: return +! CHECK: }
\ No newline at end of file diff --git a/flang/test/Lower/OpenMP/unroll-heuristic02.f90 b/flang/test/Lower/OpenMP/unroll-heuristic02.f90 index 14f694d..fdb1366 100644 --- a/flang/test/Lower/OpenMP/unroll-heuristic02.f90 +++ b/flang/test/Lower/OpenMP/unroll-heuristic02.f90 @@ -37,61 +37,55 @@ end subroutine omp_unroll_heuristic_nested02 !CHECK: %[[VAL_10:.*]]:2 = hlfir.declare %[[ARG1]] dummy_scope %[[VAL_0]] {uniq_name = "_QFomp_unroll_heuristic_nested02Eouter_ub"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>) !CHECK: %[[VAL_11:.*]] = fir.alloca i32 {bindc_name = "res", uniq_name = "_QFomp_unroll_heuristic_nested02Eres"} !CHECK: %[[VAL_12:.*]]:2 = hlfir.declare %[[VAL_11]] {uniq_name = "_QFomp_unroll_heuristic_nested02Eres"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) -!CHECK: %[[VAL_13:.*]] = fir.alloca i32 {bindc_name = "i", pinned, uniq_name = "_QFomp_unroll_heuristic_nested02Ei"} -!CHECK: %[[VAL_14:.*]]:2 = hlfir.declare %[[VAL_13]] {uniq_name = "_QFomp_unroll_heuristic_nested02Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) -!CHECK: %[[VAL_15:.*]] = fir.alloca i32 {bindc_name = "j", pinned, uniq_name = "_QFomp_unroll_heuristic_nested02Ej"} -!CHECK: %[[VAL_16:.*]]:2 = hlfir.declare %[[VAL_15]] {uniq_name = "_QFomp_unroll_heuristic_nested02Ej"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) -!CHECK: %[[VAL_17:.*]] = fir.load %[[VAL_9]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_18:.*]] = fir.load %[[VAL_10]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_19:.*]] = fir.load %[[VAL_8]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_20:.*]] = arith.constant 0 : i32 -!CHECK: %[[VAL_21:.*]] = arith.constant 1 : i32 -!CHECK: %[[VAL_22:.*]] = arith.cmpi slt, %[[VAL_19]], %[[VAL_20]] : i32 -!CHECK: %[[VAL_23:.*]] = arith.subi %[[VAL_20]], %[[VAL_19]] : i32 -!CHECK: %[[VAL_24:.*]] = arith.select %[[VAL_22]], %[[VAL_23]], %[[VAL_19]] : i32 -!CHECK: %[[VAL_25:.*]] = arith.select %[[VAL_22]], %[[VAL_18]], %[[VAL_17]] : i32 -!CHECK: %[[VAL_26:.*]] = arith.select %[[VAL_22]], %[[VAL_17]], %[[VAL_18]] : i32 -!CHECK: %[[VAL_27:.*]] = arith.subi %[[VAL_26]], %[[VAL_25]] overflow<nuw> : i32 -!CHECK: %[[VAL_28:.*]] = arith.divui %[[VAL_27]], %[[VAL_24]] : i32 -!CHECK: %[[VAL_29:.*]] = arith.addi %[[VAL_28]], %[[VAL_21]] overflow<nuw> : i32 -!CHECK: %[[VAL_30:.*]] = arith.cmpi slt, %[[VAL_26]], %[[VAL_25]] : i32 -!CHECK: %[[VAL_31:.*]] = arith.select %[[VAL_30]], %[[VAL_20]], %[[VAL_29]] : i32 -!CHECK: %[[VAL_32:.*]] = omp.new_cli -!CHECK: omp.canonical_loop(%[[VAL_32]]) %[[VAL_33:.*]] : i32 in range(%[[VAL_31]]) { -!CHECK: %[[VAL_34:.*]] = arith.muli %[[VAL_33]], %[[VAL_19]] : i32 -!CHECK: %[[VAL_35:.*]] = arith.addi %[[VAL_17]], %[[VAL_34]] : i32 -!CHECK: hlfir.assign %[[VAL_35]] to %[[VAL_14]]#0 : i32, !fir.ref<i32> -!CHECK: %[[VAL_36:.*]] = fir.alloca i32 {bindc_name = "j", pinned, uniq_name = "_QFomp_unroll_heuristic_nested02Ej"} -!CHECK: %[[VAL_37:.*]]:2 = hlfir.declare %[[VAL_36]] {uniq_name = "_QFomp_unroll_heuristic_nested02Ej"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) -!CHECK: %[[VAL_38:.*]] = fir.load %[[VAL_4]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_39:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_40:.*]] = fir.load %[[VAL_3]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_41:.*]] = arith.constant 0 : i32 -!CHECK: %[[VAL_42:.*]] = arith.constant 1 : i32 -!CHECK: %[[VAL_43:.*]] = arith.cmpi slt, %[[VAL_40]], %[[VAL_41]] : i32 -!CHECK: %[[VAL_44:.*]] = arith.subi %[[VAL_41]], %[[VAL_40]] : i32 -!CHECK: %[[VAL_45:.*]] = arith.select %[[VAL_43]], %[[VAL_44]], %[[VAL_40]] : i32 -!CHECK: %[[VAL_46:.*]] = arith.select %[[VAL_43]], %[[VAL_39]], %[[VAL_38]] : i32 -!CHECK: %[[VAL_47:.*]] = arith.select %[[VAL_43]], %[[VAL_38]], %[[VAL_39]] : i32 -!CHECK: %[[VAL_48:.*]] = arith.subi %[[VAL_47]], %[[VAL_46]] overflow<nuw> : i32 -!CHECK: %[[VAL_49:.*]] = arith.divui %[[VAL_48]], %[[VAL_45]] : i32 -!CHECK: %[[VAL_50:.*]] = arith.addi %[[VAL_49]], %[[VAL_42]] overflow<nuw> : i32 -!CHECK: %[[VAL_51:.*]] = arith.cmpi slt, %[[VAL_47]], %[[VAL_46]] : i32 -!CHECK: %[[VAL_52:.*]] = arith.select %[[VAL_51]], %[[VAL_41]], %[[VAL_50]] : i32 -!CHECK: %[[VAL_53:.*]] = omp.new_cli -!CHECK: omp.canonical_loop(%[[VAL_53]]) %[[VAL_54:.*]] : i32 in range(%[[VAL_52]]) { -!CHECK: %[[VAL_55:.*]] = arith.muli %[[VAL_54]], %[[VAL_40]] : i32 -!CHECK: %[[VAL_56:.*]] = arith.addi %[[VAL_38]], %[[VAL_55]] : i32 -!CHECK: hlfir.assign %[[VAL_56]] to %[[VAL_37]]#0 : i32, !fir.ref<i32> -!CHECK: %[[VAL_57:.*]] = fir.load %[[VAL_14]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_58:.*]] = fir.load %[[VAL_37]]#0 : !fir.ref<i32> -!CHECK: %[[VAL_59:.*]] = arith.addi %[[VAL_57]], %[[VAL_58]] : i32 -!CHECK: hlfir.assign %[[VAL_59]] to %[[VAL_12]]#0 : i32, !fir.ref<i32> +!CHECK: %[[VAL_13:.*]] = fir.load %[[VAL_9]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_14:.*]] = fir.load %[[VAL_10]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_15:.*]] = fir.load %[[VAL_8]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_16:.*]] = arith.constant 0 : i32 +!CHECK: %[[VAL_17:.*]] = arith.constant 1 : i32 +!CHECK: %[[VAL_18:.*]] = arith.cmpi slt, %[[VAL_15]], %[[VAL_16]] : i32 +!CHECK: %[[VAL_19:.*]] = arith.subi %[[VAL_16]], %[[VAL_15]] : i32 +!CHECK: %[[VAL_20:.*]] = arith.select %[[VAL_18]], %[[VAL_19]], %[[VAL_15]] : i32 +!CHECK: %[[VAL_21:.*]] = arith.select %[[VAL_18]], %[[VAL_14]], %[[VAL_13]] : i32 +!CHECK: %[[VAL_22:.*]] = arith.select %[[VAL_18]], %[[VAL_13]], %[[VAL_14]] : i32 +!CHECK: %[[VAL_23:.*]] = arith.subi %[[VAL_22]], %[[VAL_21]] overflow<nuw> : i32 +!CHECK: %[[VAL_24:.*]] = arith.divui %[[VAL_23]], %[[VAL_20]] : i32 +!CHECK: %[[VAL_25:.*]] = arith.addi %[[VAL_24]], %[[VAL_17]] overflow<nuw> : i32 +!CHECK: %[[VAL_26:.*]] = arith.cmpi slt, %[[VAL_22]], %[[VAL_21]] : i32 +!CHECK: %[[VAL_27:.*]] = arith.select %[[VAL_26]], %[[VAL_16]], %[[VAL_25]] : i32 +!CHECK: %[[VAL_28:.*]] = omp.new_cli +!CHECK: omp.canonical_loop(%[[VAL_28]]) %[[VAL_29:.*]] : i32 in range(%[[VAL_27]]) { +!CHECK: %[[VAL_30:.*]] = arith.muli %[[VAL_29]], %[[VAL_15]] : i32 +!CHECK: %[[VAL_31:.*]] = arith.addi %[[VAL_13]], %[[VAL_30]] : i32 +!CHECK: hlfir.assign %[[VAL_31]] to %[[VAL_2]]#0 : i32, !fir.ref<i32> +!CHECK: %[[VAL_32:.*]] = fir.load %[[VAL_4]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_33:.*]] = fir.load %[[VAL_5]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_34:.*]] = fir.load %[[VAL_3]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_35:.*]] = arith.constant 0 : i32 +!CHECK: %[[VAL_36:.*]] = arith.constant 1 : i32 +!CHECK: %[[VAL_37:.*]] = arith.cmpi slt, %[[VAL_34]], %[[VAL_35]] : i32 +!CHECK: %[[VAL_38:.*]] = arith.subi %[[VAL_35]], %[[VAL_34]] : i32 +!CHECK: %[[VAL_39:.*]] = arith.select %[[VAL_37]], %[[VAL_38]], %[[VAL_34]] : i32 +!CHECK: %[[VAL_40:.*]] = arith.select %[[VAL_37]], %[[VAL_33]], %[[VAL_32]] : i32 +!CHECK: %[[VAL_41:.*]] = arith.select %[[VAL_37]], %[[VAL_32]], %[[VAL_33]] : i32 +!CHECK: %[[VAL_42:.*]] = arith.subi %[[VAL_41]], %[[VAL_40]] overflow<nuw> : i32 +!CHECK: %[[VAL_43:.*]] = arith.divui %[[VAL_42]], %[[VAL_39]] : i32 +!CHECK: %[[VAL_44:.*]] = arith.addi %[[VAL_43]], %[[VAL_36]] overflow<nuw> : i32 +!CHECK: %[[VAL_45:.*]] = arith.cmpi slt, %[[VAL_41]], %[[VAL_40]] : i32 +!CHECK: %[[VAL_46:.*]] = arith.select %[[VAL_45]], %[[VAL_35]], %[[VAL_44]] : i32 +!CHECK: %[[VAL_47:.*]] = omp.new_cli +!CHECK: omp.canonical_loop(%[[VAL_47]]) %[[VAL_48:.*]] : i32 in range(%[[VAL_46]]) { +!CHECK: %[[VAL_49:.*]] = arith.muli %[[VAL_48]], %[[VAL_34]] : i32 +!CHECK: %[[VAL_50:.*]] = arith.addi %[[VAL_32]], %[[VAL_49]] : i32 +!CHECK: hlfir.assign %[[VAL_50]] to %[[VAL_7]]#0 : i32, !fir.ref<i32> +!CHECK: %[[VAL_51:.*]] = fir.load %[[VAL_2]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_52:.*]] = fir.load %[[VAL_7]]#0 : !fir.ref<i32> +!CHECK: %[[VAL_53:.*]] = arith.addi %[[VAL_51]], %[[VAL_52]] : i32 +!CHECK: hlfir.assign %[[VAL_53]] to %[[VAL_12]]#0 : i32, !fir.ref<i32> !CHECK: omp.terminator !CHECK: } -!CHECK: omp.unroll_heuristic(%[[VAL_53]]) +!CHECK: omp.unroll_heuristic(%[[VAL_47]]) !CHECK: omp.terminator !CHECK: } -!CHECK: omp.unroll_heuristic(%[[VAL_32]]) +!CHECK: omp.unroll_heuristic(%[[VAL_28]]) !CHECK: return !CHECK: } diff --git a/flang/test/Lower/OpenMP/unroll-heuristic03.f90 b/flang/test/Lower/OpenMP/unroll-heuristic03.f90 new file mode 100644 index 0000000..308c149 --- /dev/null +++ b/flang/test/Lower/OpenMP/unroll-heuristic03.f90 @@ -0,0 +1,61 @@ +! RUN: %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=51 -o - %s 2>&1 | FileCheck %s + +! Test implicitly privatized loop variable that is affected by unrolling. + +subroutine omp_unroll_heuristic03(lb, ub, inc) + integer res, i, lb, ub, inc + + !$omp parallel + !$omp unroll + do i = lb, ub, inc + res = i + end do + !$omp end unroll + !$omp end parallel + +end subroutine omp_unroll_heuristic03 + + +! CHECK-LABEL: func.func @_QPomp_unroll_heuristic03( +! CHECK-SAME: %[[ARG0:.*]]: !fir.ref<i32> {fir.bindc_name = "lb"}, +! CHECK-SAME: %[[ARG1:.*]]: !fir.ref<i32> {fir.bindc_name = "ub"}, +! CHECK-SAME: %[[ARG2:.*]]: !fir.ref<i32> {fir.bindc_name = "inc"}) { +! CHECK: %[[VAL_0:.*]] = fir.dummy_scope : !fir.dscope +! CHECK: %[[VAL_1:.*]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFomp_unroll_heuristic03Ei"} +! CHECK: %[[VAL_2:.*]]:2 = hlfir.declare %[[VAL_1]] {uniq_name = "_QFomp_unroll_heuristic03Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_3:.*]]:2 = hlfir.declare %[[ARG2]] dummy_scope %[[VAL_0]] {uniq_name = "_QFomp_unroll_heuristic03Einc"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_4:.*]]:2 = hlfir.declare %[[ARG0]] dummy_scope %[[VAL_0]] {uniq_name = "_QFomp_unroll_heuristic03Elb"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_5:.*]] = fir.alloca i32 {bindc_name = "res", uniq_name = "_QFomp_unroll_heuristic03Eres"} +! CHECK: %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_5]] {uniq_name = "_QFomp_unroll_heuristic03Eres"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_7:.*]]:2 = hlfir.declare %[[ARG1]] dummy_scope %[[VAL_0]] {uniq_name = "_QFomp_unroll_heuristic03Eub"} : (!fir.ref<i32>, !fir.dscope) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: omp.parallel private(@_QFomp_unroll_heuristic03Ei_private_i32 %[[VAL_2]]#0 -> %[[VAL_8:.*]] : !fir.ref<i32>) { +! CHECK: %[[VAL_9:.*]]:2 = hlfir.declare %[[VAL_8]] {uniq_name = "_QFomp_unroll_heuristic03Ei"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) +! CHECK: %[[VAL_10:.*]] = fir.load %[[VAL_4]]#0 : !fir.ref<i32> +! CHECK: %[[VAL_11:.*]] = fir.load %[[VAL_7]]#0 : !fir.ref<i32> +! CHECK: %[[VAL_12:.*]] = fir.load %[[VAL_3]]#0 : !fir.ref<i32> +! CHECK: %[[VAL_13:.*]] = arith.constant 0 : i32 +! CHECK: %[[VAL_14:.*]] = arith.constant 1 : i32 +! CHECK: %[[VAL_15:.*]] = arith.cmpi slt, %[[VAL_12]], %[[VAL_13]] : i32 +! CHECK: %[[VAL_16:.*]] = arith.subi %[[VAL_13]], %[[VAL_12]] : i32 +! CHECK: %[[VAL_17:.*]] = arith.select %[[VAL_15]], %[[VAL_16]], %[[VAL_12]] : i32 +! CHECK: %[[VAL_18:.*]] = arith.select %[[VAL_15]], %[[VAL_11]], %[[VAL_10]] : i32 +! CHECK: %[[VAL_19:.*]] = arith.select %[[VAL_15]], %[[VAL_10]], %[[VAL_11]] : i32 +! CHECK: %[[VAL_20:.*]] = arith.subi %[[VAL_19]], %[[VAL_18]] overflow<nuw> : i32 +! CHECK: %[[VAL_21:.*]] = arith.divui %[[VAL_20]], %[[VAL_17]] : i32 +! CHECK: %[[VAL_22:.*]] = arith.addi %[[VAL_21]], %[[VAL_14]] overflow<nuw> : i32 +! CHECK: %[[VAL_23:.*]] = arith.cmpi slt, %[[VAL_19]], %[[VAL_18]] : i32 +! CHECK: %[[VAL_24:.*]] = arith.select %[[VAL_23]], %[[VAL_13]], %[[VAL_22]] : i32 +! CHECK: %[[VAL_25:.*]] = omp.new_cli +! CHECK: omp.canonical_loop(%[[VAL_25]]) %[[VAL_26:.*]] : i32 in range(%[[VAL_24]]) { +! CHECK: %[[VAL_27:.*]] = arith.muli %[[VAL_26]], %[[VAL_12]] : i32 +! CHECK: %[[VAL_28:.*]] = arith.addi %[[VAL_10]], %[[VAL_27]] : i32 +! CHECK: hlfir.assign %[[VAL_28]] to %[[VAL_9]]#0 : i32, !fir.ref<i32> +! CHECK: %[[VAL_29:.*]] = fir.load %[[VAL_9]]#0 : !fir.ref<i32> +! CHECK: hlfir.assign %[[VAL_29]] to %[[VAL_6]]#0 : i32, !fir.ref<i32> +! CHECK: omp.terminator +! CHECK: } +! CHECK: omp.unroll_heuristic(%[[VAL_25]]) +! CHECK: omp.terminator +! CHECK: } +! CHECK: return +! CHECK: }
\ No newline at end of file |