derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
3722 lines
272 KiB
C++
3722 lines
272 KiB
C++
// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _
|
|
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple x86_64-unknown-unknown -fopenmp-targets=amdgcn-amd-amdhsa -fopenmp-target-ignore-env-vars -fopenmp-assume-no-nested-parallelism -fopenmp-assume-no-thread-state -emit-llvm-bc %s -o %t-ppc-host.bc
|
|
// RUN: %clang_cc1 -fopenmp -x c++ -std=c++11 -triple amdgcn-amd-amdhsa -fopenmp-targets=amdgcn-amd-amdhsa -fopenmp-target-ignore-env-vars -fopenmp-assume-no-nested-parallelism -fopenmp-assume-no-thread-state -emit-llvm %s -fopenmp-is-device -fopenmp-host-ir-file-path %t-ppc-host.bc -o - | FileCheck %s
|
|
// expected-no-diagnostics
|
|
|
|
int compute_sum_res(int j, double *result, double a[]) {
|
|
*result += a[j];
|
|
return 1;
|
|
}
|
|
|
|
void compute_sum(int j, double *result, double a[]) {
|
|
*result += a[j];
|
|
}
|
|
|
|
double compute_sum_rval(int j, double rval, double a[]) {
|
|
return rval + a[j];
|
|
}
|
|
|
|
int foo(int i) { return 2*i; }
|
|
|
|
int main()
|
|
{
|
|
int N = 10000;
|
|
|
|
double a[N];
|
|
|
|
for (int i=0; i<N; i++)
|
|
a[i]=i;
|
|
|
|
double sum1, sum2, sum3, sum4, sum5;
|
|
sum1 = sum2 = sum3 = sum4 = sum5 = 0;
|
|
|
|
int res = 0;
|
|
#pragma omp target teams distribute parallel for map(tofrom:sum1) reduction(+:sum1) map(tofrom:res)
|
|
for (int j = 0; j< N; j=j+1)
|
|
res = compute_sum_res(j, &sum1, a);
|
|
|
|
#pragma omp target teams distribute parallel for reduction(+:sum2)
|
|
for (int j = 0; j< N; j=j+1)
|
|
compute_sum(j, &sum2, a);
|
|
|
|
#pragma omp target teams distribute parallel for reduction(+:sum3)
|
|
for (int j = 0; j< N; j=j+1)
|
|
sum3 = compute_sum_rval(j, *&sum3, a);
|
|
|
|
#pragma omp target teams distribute parallel for reduction(+:sum4)
|
|
for (int j = 0; j< N; j=j+1)
|
|
foo(compute_sum_res(j, &sum4, a));
|
|
|
|
#pragma omp target teams distribute parallel for reduction(+:sum5)
|
|
for (int j = 0; j< N; j=j+1)
|
|
compute_sum_res(j, &sum5, a);
|
|
}
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34
|
|
// CHECK-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[RES:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM1:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0:[0-9]+]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[RES_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM1_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[RES_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RES_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM1_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM1_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[RES]], ptr [[RES_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM1]], ptr [[SUM1_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[RES_ADDR_ASCAST]], align 8, !nonnull [[META9:![0-9]+]], !align [[META10:![0-9]+]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM1_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11:![0-9]+]]
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_kernel_environment to ptr), ptr [[DYN_PTR]])
|
|
// CHECK-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP4]], -1
|
|
// CHECK-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
|
|
// CHECK: user_code.entry:
|
|
// CHECK-NEXT: [[TMP5:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1:[0-9]+]] to ptr))
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP5]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
|
|
// CHECK-NEXT: [[TMP8:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
|
|
// CHECK-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined(ptr [[TMP8]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP7]], ptr [[TMP0]], ptr [[TMP1]], i64 [[TMP2]], ptr [[TMP3]]) #[[ATTR2:[0-9]+]]
|
|
// CHECK-NEXT: call void @__kmpc_target_deinit()
|
|
// CHECK-NEXT: ret void
|
|
// CHECK: worker.exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[RES:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM1:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1:[0-9]+]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[RES_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM1_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM11:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J4:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [7 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[RES_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RES_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM1_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM1_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM11_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM11]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_2]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[J4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J4]] to ptr
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[RES]], ptr [[RES_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM1]], ptr [[SUM1_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[RES_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META10]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM1_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM11_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP7]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[TMP8]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2:[0-9]+]] to ptr), i32 [[TMP9]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP10]], [[TMP11]]
|
|
// CHECK-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
|
|
// CHECK: cond.true:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END:%.*]]
|
|
// CHECK: cond.false:
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END]]
|
|
// CHECK: cond.end:
|
|
// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP12]], [[COND_TRUE]] ], [ [[TMP13]], [[COND_FALSE]] ]
|
|
// CHECK-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP14]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP16]], 1
|
|
// CHECK-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP15]], [[ADD]]
|
|
// CHECK-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP18:%.*]] = zext i32 [[TMP17]] to i64
|
|
// CHECK-NEXT: [[TMP19:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP20:%.*]] = zext i32 [[TMP19]] to i64
|
|
// CHECK-NEXT: [[TMP21:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP21]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP22:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP24:%.*]] = inttoptr i64 [[TMP18]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP24]], ptr [[TMP23]], align 8
|
|
// CHECK-NEXT: [[TMP25:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
|
|
// CHECK-NEXT: [[TMP26:%.*]] = inttoptr i64 [[TMP20]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP26]], ptr [[TMP25]], align 8
|
|
// CHECK-NEXT: [[TMP27:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
|
|
// CHECK-NEXT: [[TMP28:%.*]] = inttoptr i64 [[TMP22]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP28]], ptr [[TMP27]], align 8
|
|
// CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[TMP29]], align 8
|
|
// CHECK-NEXT: [[TMP30:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
|
|
// CHECK-NEXT: store ptr [[SUM11_ASCAST]], ptr [[TMP30]], align 8
|
|
// CHECK-NEXT: [[TMP31:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
|
|
// CHECK-NEXT: [[TMP32:%.*]] = inttoptr i64 [[TMP2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP32]], ptr [[TMP31]], align 8
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [7 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 6
|
|
// CHECK-NEXT: store ptr [[TMP3]], ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[TMP34]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP35]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 7, i32 0)
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
|
|
// CHECK-NEXT: store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP41:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP40]], [[TMP41]]
|
|
// CHECK-NEXT: store i32 [[ADD9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP42:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP42]], [[TMP43]]
|
|
// CHECK-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
|
|
// CHECK: cond.true11:
|
|
// CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13:%.*]]
|
|
// CHECK: cond.false12:
|
|
// CHECK-NEXT: [[TMP45:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13]]
|
|
// CHECK: cond.end13:
|
|
// CHECK-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP44]], [[COND_TRUE11]] ], [ [[TMP45]], [[COND_FALSE12]] ]
|
|
// CHECK-NEXT: store i32 [[COND14]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP46]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP47:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP48:%.*]] = load i32, ptr [[TMP47]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP48]])
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: [[TMP49:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM11_ASCAST]], ptr [[TMP49]], align 8
|
|
// CHECK-NEXT: %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
|
|
// CHECK-NEXT: [[TMP50:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.1, ptr @_omp_reduction_inter_warp_copy_func.2, ptr @_omp_reduction_list_to_global_copy_func, ptr @_omp_reduction_list_to_global_reduce_func, ptr @_omp_reduction_global_to_list_copy_func, ptr @_omp_reduction_global_to_list_reduce_func)
|
|
// CHECK-NEXT: [[TMP51:%.*]] = icmp eq i32 [[TMP50]], 1
|
|
// CHECK-NEXT: br i1 [[TMP51]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP52:%.*]] = load double, ptr [[TMP1]], align 8
|
|
// CHECK-NEXT: [[TMP53:%.*]] = load double, ptr [[SUM11_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD15:%.*]] = fadd double [[TMP52]], [[TMP53]]
|
|
// CHECK-NEXT: store double [[ADD15]], ptr [[TMP1]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 4 dereferenceable(4) [[RES:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM1:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[RES_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM1_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[SUM14:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J5:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[RES_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RES_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM1_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM1_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_1]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[SUM14_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM14]] to ptr
|
|
// CHECK-NEXT: [[J5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J5]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[RES]], ptr [[RES_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM1]], ptr [[SUM1_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[RES_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META10]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[SUM1_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP4]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP5]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP6]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP7]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV:%.*]] = trunc i64 [[TMP8]] to i32
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP9]] to i32
|
|
// CHECK-NEXT: store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[CONV3]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM14_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3:[0-9]+]] to ptr), i32 [[TMP11]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP12]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CONV6:%.*]] = sext i32 [[TMP13]] to i64
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CMP7:%.*]] = icmp ule i64 [[CONV6]], [[TMP14]]
|
|
// CHECK-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP15]], 1
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
|
|
// CHECK-NEXT: store i32 [[ADD]], ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CALL:%.*]] = call noundef i32 @_Z15compute_sum_resiPdS_(i32 noundef [[TMP16]], ptr noundef [[SUM14_ASCAST]], ptr noundef [[TMP3]]) #[[ATTR7:[0-9]+]]
|
|
// CHECK-NEXT: store i32 [[CALL]], ptr [[TMP0]], align 4
|
|
// CHECK-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
|
|
// CHECK: omp.body.continue:
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP20]])
|
|
// CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM14_ASCAST]], ptr [[TMP21]], align 8
|
|
// CHECK-NEXT: [[TMP22:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func, ptr @_omp_reduction_inter_warp_copy_func)
|
|
// CHECK-NEXT: [[TMP23:%.*]] = icmp eq i32 [[TMP22]], 1
|
|
// CHECK-NEXT: br i1 [[TMP23]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP24:%.*]] = load double, ptr [[TMP1]], align 8
|
|
// CHECK-NEXT: [[TMP25:%.*]] = load double, ptr [[SUM14_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD9:%.*]] = fadd double [[TMP24]], [[TMP25]]
|
|
// CHECK-NEXT: store double [[ADD9]], ptr [[TMP1]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_Z15compute_sum_resiPdS_
|
|
// CHECK-SAME: (i32 noundef [[J:%.*]], ptr noundef [[RESULT:%.*]], ptr noundef [[A:%.*]]) #[[ATTR3:[0-9]+]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[J_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[RESULT_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_ADDR]] to ptr
|
|
// CHECK-NEXT: [[RESULT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: store i32 [[J]], ptr [[J_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[RESULT]], ptr [[RESULT_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[J_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
|
|
// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i64 [[IDXPROM]]
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load double, ptr [[ARRAYIDX]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[RESULT_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load double, ptr [[TMP3]], align 8
|
|
// CHECK-NEXT: [[ADD:%.*]] = fadd double [[TMP4]], [[TMP2]]
|
|
// CHECK-NEXT: store double [[ADD]], ptr [[TMP3]], align 8
|
|
// CHECK-NEXT: ret i32 1
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4:[0-9]+]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4:[0-9]+]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.1
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.2
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l34_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38
|
|
// CHECK-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM2:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM2_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM2_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM2_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM2]], ptr [[SUM2_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM2_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_kernel_environment to ptr), ptr [[DYN_PTR]])
|
|
// CHECK-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
|
|
// CHECK-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
|
|
// CHECK: user_code.entry:
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP5]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP4]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
|
|
// CHECK-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined(ptr [[TMP7]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP6]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]]) #[[ATTR2]]
|
|
// CHECK-NEXT: call void @__kmpc_target_deinit()
|
|
// CHECK-NEXT: ret void
|
|
// CHECK: worker.exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM2:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM2_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM21:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J4:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [6 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM2_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM2_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM21_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM21]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_2]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[J4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J4]] to ptr
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM2]], ptr [[SUM2_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM2_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM21_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP8]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
|
|
// CHECK-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
|
|
// CHECK: cond.true:
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END:%.*]]
|
|
// CHECK: cond.false:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END]]
|
|
// CHECK: cond.end:
|
|
// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
|
|
// CHECK-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP13]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP15]], 1
|
|
// CHECK-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP14]], [[ADD]]
|
|
// CHECK-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
|
|
// CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP20]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP21:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP23:%.*]] = inttoptr i64 [[TMP17]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP23]], ptr [[TMP22]], align 8
|
|
// CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
|
|
// CHECK-NEXT: [[TMP25:%.*]] = inttoptr i64 [[TMP19]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP25]], ptr [[TMP24]], align 8
|
|
// CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
|
|
// CHECK-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP21]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP27]], ptr [[TMP26]], align 8
|
|
// CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
|
|
// CHECK-NEXT: store ptr [[SUM21_ASCAST]], ptr [[TMP28]], align 8
|
|
// CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
|
|
// CHECK-NEXT: [[TMP30:%.*]] = inttoptr i64 [[TMP1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP30]], ptr [[TMP29]], align 8
|
|
// CHECK-NEXT: [[TMP31:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[TMP31]], align 8
|
|
// CHECK-NEXT: [[TMP32:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP33:%.*]] = load i32, ptr [[TMP32]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP33]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 6, i32 0)
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
|
|
// CHECK-NEXT: store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
|
|
// CHECK-NEXT: store i32 [[ADD9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP41:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP40]], [[TMP41]]
|
|
// CHECK-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
|
|
// CHECK: cond.true11:
|
|
// CHECK-NEXT: [[TMP42:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13:%.*]]
|
|
// CHECK: cond.false12:
|
|
// CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13]]
|
|
// CHECK: cond.end13:
|
|
// CHECK-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP42]], [[COND_TRUE11]] ], [ [[TMP43]], [[COND_FALSE12]] ]
|
|
// CHECK-NEXT: store i32 [[COND14]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP44]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP45:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[TMP45]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP46]])
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM21_ASCAST]], ptr [[TMP47]], align 8
|
|
// CHECK-NEXT: %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
|
|
// CHECK-NEXT: [[TMP48:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.5, ptr @_omp_reduction_inter_warp_copy_func.6, ptr @_omp_reduction_list_to_global_copy_func.7, ptr @_omp_reduction_list_to_global_reduce_func.8, ptr @_omp_reduction_global_to_list_copy_func.9, ptr @_omp_reduction_global_to_list_reduce_func.10)
|
|
// CHECK-NEXT: [[TMP49:%.*]] = icmp eq i32 [[TMP48]], 1
|
|
// CHECK-NEXT: br i1 [[TMP49]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP50:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP51:%.*]] = load double, ptr [[SUM21_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD15:%.*]] = fadd double [[TMP50]], [[TMP51]]
|
|
// CHECK-NEXT: store double [[ADD15]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM2:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM2_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[SUM24:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J5:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM2_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM2_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_1]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[SUM24_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM24]] to ptr
|
|
// CHECK-NEXT: [[J5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J5]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM2]], ptr [[SUM2_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM2_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
|
|
// CHECK-NEXT: store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[CONV3]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM24_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP10]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP11]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CONV6:%.*]] = sext i32 [[TMP12]] to i64
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CMP7:%.*]] = icmp ule i64 [[CONV6]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
|
|
// CHECK-NEXT: store i32 [[ADD]], ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: call void @_Z11compute_sumiPdS_(i32 noundef [[TMP15]], ptr noundef [[SUM24_ASCAST]], ptr noundef [[TMP2]]) #[[ATTR7]]
|
|
// CHECK-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
|
|
// CHECK: omp.body.continue:
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP19]])
|
|
// CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM24_ASCAST]], ptr [[TMP20]], align 8
|
|
// CHECK-NEXT: [[TMP21:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.3, ptr @_omp_reduction_inter_warp_copy_func.4)
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i32 [[TMP21]], 1
|
|
// CHECK-NEXT: br i1 [[TMP22]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP23:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP24:%.*]] = load double, ptr [[SUM24_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD9:%.*]] = fadd double [[TMP23]], [[TMP24]]
|
|
// CHECK-NEXT: store double [[ADD9]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_Z11compute_sumiPdS_
|
|
// CHECK-SAME: (i32 noundef [[J:%.*]], ptr noundef [[RESULT:%.*]], ptr noundef [[A:%.*]]) #[[ATTR3]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[J_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[RESULT_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_ADDR]] to ptr
|
|
// CHECK-NEXT: [[RESULT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RESULT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: store i32 [[J]], ptr [[J_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[RESULT]], ptr [[RESULT_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i32, ptr [[J_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP1]] to i64
|
|
// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i64 [[IDXPROM]]
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load double, ptr [[ARRAYIDX]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[RESULT_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load double, ptr [[TMP3]], align 8
|
|
// CHECK-NEXT: [[ADD:%.*]] = fadd double [[TMP4]], [[TMP2]]
|
|
// CHECK-NEXT: store double [[ADD]], ptr [[TMP3]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.3
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.4
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.5
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.6
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.7
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.8
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.9
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.10
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_1]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l38_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42
|
|
// CHECK-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM3:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM3_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM3_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM3_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM3]], ptr [[SUM3_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM3_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_kernel_environment to ptr), ptr [[DYN_PTR]])
|
|
// CHECK-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
|
|
// CHECK-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
|
|
// CHECK: user_code.entry:
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP5]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP4]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
|
|
// CHECK-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined(ptr [[TMP7]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP6]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]]) #[[ATTR2]]
|
|
// CHECK-NEXT: call void @__kmpc_target_deinit()
|
|
// CHECK-NEXT: ret void
|
|
// CHECK: worker.exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM3:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM3_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM31:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J4:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [6 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM3_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM3_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM31_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM31]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_2]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[J4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J4]] to ptr
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM3]], ptr [[SUM3_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM3_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM31_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP8]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
|
|
// CHECK-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
|
|
// CHECK: cond.true:
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END:%.*]]
|
|
// CHECK: cond.false:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END]]
|
|
// CHECK: cond.end:
|
|
// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
|
|
// CHECK-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP13]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP15]], 1
|
|
// CHECK-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP14]], [[ADD]]
|
|
// CHECK-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
|
|
// CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP20]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP21:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP23:%.*]] = inttoptr i64 [[TMP17]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP23]], ptr [[TMP22]], align 8
|
|
// CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
|
|
// CHECK-NEXT: [[TMP25:%.*]] = inttoptr i64 [[TMP19]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP25]], ptr [[TMP24]], align 8
|
|
// CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
|
|
// CHECK-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP21]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP27]], ptr [[TMP26]], align 8
|
|
// CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
|
|
// CHECK-NEXT: store ptr [[SUM31_ASCAST]], ptr [[TMP28]], align 8
|
|
// CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
|
|
// CHECK-NEXT: [[TMP30:%.*]] = inttoptr i64 [[TMP1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP30]], ptr [[TMP29]], align 8
|
|
// CHECK-NEXT: [[TMP31:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[TMP31]], align 8
|
|
// CHECK-NEXT: [[TMP32:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP33:%.*]] = load i32, ptr [[TMP32]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP33]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 6, i32 0)
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
|
|
// CHECK-NEXT: store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
|
|
// CHECK-NEXT: store i32 [[ADD9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP41:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP40]], [[TMP41]]
|
|
// CHECK-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
|
|
// CHECK: cond.true11:
|
|
// CHECK-NEXT: [[TMP42:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13:%.*]]
|
|
// CHECK: cond.false12:
|
|
// CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13]]
|
|
// CHECK: cond.end13:
|
|
// CHECK-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP42]], [[COND_TRUE11]] ], [ [[TMP43]], [[COND_FALSE12]] ]
|
|
// CHECK-NEXT: store i32 [[COND14]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP44]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP45:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[TMP45]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP46]])
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM31_ASCAST]], ptr [[TMP47]], align 8
|
|
// CHECK-NEXT: %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
|
|
// CHECK-NEXT: [[TMP48:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.13, ptr @_omp_reduction_inter_warp_copy_func.14, ptr @_omp_reduction_list_to_global_copy_func.15, ptr @_omp_reduction_list_to_global_reduce_func.16, ptr @_omp_reduction_global_to_list_copy_func.17, ptr @_omp_reduction_global_to_list_reduce_func.18)
|
|
// CHECK-NEXT: [[TMP49:%.*]] = icmp eq i32 [[TMP48]], 1
|
|
// CHECK-NEXT: br i1 [[TMP49]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP50:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP51:%.*]] = load double, ptr [[SUM31_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD15:%.*]] = fadd double [[TMP50]], [[TMP51]]
|
|
// CHECK-NEXT: store double [[ADD15]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM3:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM3_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[SUM34:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J5:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM3_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM3_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_1]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[SUM34_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM34]] to ptr
|
|
// CHECK-NEXT: [[J5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J5]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM3]], ptr [[SUM3_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM3_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
|
|
// CHECK-NEXT: store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[CONV3]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM34_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP10]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP11]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CONV6:%.*]] = sext i32 [[TMP12]] to i64
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CMP7:%.*]] = icmp ule i64 [[CONV6]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
|
|
// CHECK-NEXT: store i32 [[ADD]], ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load double, ptr [[SUM34_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CALL:%.*]] = call noundef double @_Z16compute_sum_rvalidPd(i32 noundef [[TMP15]], double noundef [[TMP16]], ptr noundef [[TMP2]]) #[[ATTR7]]
|
|
// CHECK-NEXT: store double [[CALL]], ptr [[SUM34_ASCAST]], align 8
|
|
// CHECK-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
|
|
// CHECK: omp.body.continue:
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP17]], [[TMP18]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP20]])
|
|
// CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM34_ASCAST]], ptr [[TMP21]], align 8
|
|
// CHECK-NEXT: [[TMP22:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.11, ptr @_omp_reduction_inter_warp_copy_func.12)
|
|
// CHECK-NEXT: [[TMP23:%.*]] = icmp eq i32 [[TMP22]], 1
|
|
// CHECK-NEXT: br i1 [[TMP23]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP24:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP25:%.*]] = load double, ptr [[SUM34_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD9:%.*]] = fadd double [[TMP24]], [[TMP25]]
|
|
// CHECK-NEXT: store double [[ADD9]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_Z16compute_sum_rvalidPd
|
|
// CHECK-SAME: (i32 noundef [[J:%.*]], double noundef [[RVAL:%.*]], ptr noundef [[A:%.*]]) #[[ATTR3]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[J_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[RVAL_ADDR:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J_ADDR]] to ptr
|
|
// CHECK-NEXT: [[RVAL_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[RVAL_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: store i32 [[J]], ptr [[J_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store double [[RVAL]], ptr [[RVAL_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load double, ptr [[RVAL_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[J_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IDXPROM:%.*]] = sext i32 [[TMP2]] to i64
|
|
// CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i64 [[IDXPROM]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load double, ptr [[ARRAYIDX]], align 8
|
|
// CHECK-NEXT: [[ADD:%.*]] = fadd double [[TMP0]], [[TMP3]]
|
|
// CHECK-NEXT: ret double [[ADD]]
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.11
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.12
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.13
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.14
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.15
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.16
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.17
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.18
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_4]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l42_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46
|
|
// CHECK-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM4:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM4_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM4_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM4_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM4]], ptr [[SUM4_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM4_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_kernel_environment to ptr), ptr [[DYN_PTR]])
|
|
// CHECK-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
|
|
// CHECK-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
|
|
// CHECK: user_code.entry:
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP5]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP4]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
|
|
// CHECK-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined(ptr [[TMP7]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP6]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]]) #[[ATTR2]]
|
|
// CHECK-NEXT: call void @__kmpc_target_deinit()
|
|
// CHECK-NEXT: ret void
|
|
// CHECK: worker.exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM4:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM4_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM41:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J4:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [6 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM4_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM4_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM41_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM41]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_2]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[J4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J4]] to ptr
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM4]], ptr [[SUM4_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM4_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM41_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP8]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
|
|
// CHECK-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
|
|
// CHECK: cond.true:
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END:%.*]]
|
|
// CHECK: cond.false:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END]]
|
|
// CHECK: cond.end:
|
|
// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
|
|
// CHECK-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP13]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP15]], 1
|
|
// CHECK-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP14]], [[ADD]]
|
|
// CHECK-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
|
|
// CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP20]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP21:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP23:%.*]] = inttoptr i64 [[TMP17]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP23]], ptr [[TMP22]], align 8
|
|
// CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
|
|
// CHECK-NEXT: [[TMP25:%.*]] = inttoptr i64 [[TMP19]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP25]], ptr [[TMP24]], align 8
|
|
// CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
|
|
// CHECK-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP21]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP27]], ptr [[TMP26]], align 8
|
|
// CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
|
|
// CHECK-NEXT: store ptr [[SUM41_ASCAST]], ptr [[TMP28]], align 8
|
|
// CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
|
|
// CHECK-NEXT: [[TMP30:%.*]] = inttoptr i64 [[TMP1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP30]], ptr [[TMP29]], align 8
|
|
// CHECK-NEXT: [[TMP31:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[TMP31]], align 8
|
|
// CHECK-NEXT: [[TMP32:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP33:%.*]] = load i32, ptr [[TMP32]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP33]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 6, i32 0)
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
|
|
// CHECK-NEXT: store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
|
|
// CHECK-NEXT: store i32 [[ADD9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP41:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP40]], [[TMP41]]
|
|
// CHECK-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
|
|
// CHECK: cond.true11:
|
|
// CHECK-NEXT: [[TMP42:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13:%.*]]
|
|
// CHECK: cond.false12:
|
|
// CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13]]
|
|
// CHECK: cond.end13:
|
|
// CHECK-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP42]], [[COND_TRUE11]] ], [ [[TMP43]], [[COND_FALSE12]] ]
|
|
// CHECK-NEXT: store i32 [[COND14]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP44]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP45:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[TMP45]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP46]])
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM41_ASCAST]], ptr [[TMP47]], align 8
|
|
// CHECK-NEXT: %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
|
|
// CHECK-NEXT: [[TMP48:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.21, ptr @_omp_reduction_inter_warp_copy_func.22, ptr @_omp_reduction_list_to_global_copy_func.23, ptr @_omp_reduction_list_to_global_reduce_func.24, ptr @_omp_reduction_global_to_list_copy_func.25, ptr @_omp_reduction_global_to_list_reduce_func.26)
|
|
// CHECK-NEXT: [[TMP49:%.*]] = icmp eq i32 [[TMP48]], 1
|
|
// CHECK-NEXT: br i1 [[TMP49]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP50:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP51:%.*]] = load double, ptr [[SUM41_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD15:%.*]] = fadd double [[TMP50]], [[TMP51]]
|
|
// CHECK-NEXT: store double [[ADD15]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM4:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM4_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[SUM44:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J5:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM4_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM4_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_1]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[SUM44_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM44]] to ptr
|
|
// CHECK-NEXT: [[J5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J5]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM4]], ptr [[SUM4_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM4_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
|
|
// CHECK-NEXT: store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[CONV3]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM44_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP10]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP11]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CONV6:%.*]] = sext i32 [[TMP12]] to i64
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CMP7:%.*]] = icmp ule i64 [[CONV6]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
|
|
// CHECK-NEXT: store i32 [[ADD]], ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CALL:%.*]] = call noundef i32 @_Z15compute_sum_resiPdS_(i32 noundef [[TMP15]], ptr noundef [[SUM44_ASCAST]], ptr noundef [[TMP2]]) #[[ATTR7]]
|
|
// CHECK-NEXT: [[CALL8:%.*]] = call noundef i32 @_Z3fooi(i32 noundef [[CALL]]) #[[ATTR7]]
|
|
// CHECK-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
|
|
// CHECK: omp.body.continue:
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
|
|
// CHECK-NEXT: store i32 [[ADD9]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP19]])
|
|
// CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM44_ASCAST]], ptr [[TMP20]], align 8
|
|
// CHECK-NEXT: [[TMP21:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.19, ptr @_omp_reduction_inter_warp_copy_func.20)
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i32 [[TMP21]], 1
|
|
// CHECK-NEXT: br i1 [[TMP22]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP23:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP24:%.*]] = load double, ptr [[SUM44_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD10:%.*]] = fadd double [[TMP23]], [[TMP24]]
|
|
// CHECK-NEXT: store double [[ADD10]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_Z3fooi
|
|
// CHECK-SAME: (i32 noundef [[I:%.*]]) #[[ATTR3]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[I_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[I_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[I_ADDR]] to ptr
|
|
// CHECK-NEXT: store i32 [[I]], ptr [[I_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[I_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 2, [[TMP0]]
|
|
// CHECK-NEXT: ret i32 [[MUL]]
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.19
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.20
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.21
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.22
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.23
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.24
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.25
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.26
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_7]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l46_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50
|
|
// CHECK-SAME: (i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM5:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]], ptr noalias noundef [[DYN_PTR:%.*]]) #[[ATTR0]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM5_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM5_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM5_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DYN_PTR_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DYN_PTR_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTZERO_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTZERO_ADDR]] to ptr
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM5]], ptr [[SUM5_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DYN_PTR]], ptr [[DYN_PTR_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM5_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_kernel_environment to ptr), ptr [[DYN_PTR]])
|
|
// CHECK-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP3]], -1
|
|
// CHECK-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]]
|
|
// CHECK: user_code.entry:
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP5]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTZERO_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP4]], ptr addrspace(5) [[DOTTHREADID_TEMP_]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = addrspacecast ptr addrspace(5) [[DOTTHREADID_TEMP_]] to ptr
|
|
// CHECK-NEXT: call void @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined(ptr [[TMP7]], ptr [[DOTZERO_ADDR_ASCAST]], i64 [[TMP6]], ptr [[TMP0]], i64 [[TMP1]], ptr [[TMP2]]) #[[ATTR2]]
|
|
// CHECK-NEXT: call void @__kmpc_target_deinit()
|
|
// CHECK-NEXT: ret void
|
|
// CHECK: worker.exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM5:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM5_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM51:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J4:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[N_CASTED:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [6 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM5_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM5_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM51_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM51]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_2]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_COMB_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_COMB_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[J4_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J4]] to ptr
|
|
// CHECK-NEXT: [[CAPTURED_VARS_ADDRS_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[CAPTURED_VARS_ADDRS]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM5]], ptr [[SUM5_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM5_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM51_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB3:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB3]], ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: [[NVPTX_NUM_THREADS:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block()
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[TMP7]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP8]], i32 91, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_COMB_LB_ASCAST]], ptr [[DOTOMP_COMB_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 [[NVPTX_NUM_THREADS]])
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP5:%.*]] = icmp sgt i32 [[TMP9]], [[TMP10]]
|
|
// CHECK-NEXT: br i1 [[CMP5]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]]
|
|
// CHECK: cond.true:
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END:%.*]]
|
|
// CHECK: cond.false:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END]]
|
|
// CHECK: cond.end:
|
|
// CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[TMP11]], [[COND_TRUE]] ], [ [[TMP12]], [[COND_FALSE]] ]
|
|
// CHECK-NEXT: store i32 [[COND]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP13]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 [[TMP15]], 1
|
|
// CHECK-NEXT: [[CMP6:%.*]] = icmp slt i32 [[TMP14]], [[ADD]]
|
|
// CHECK-NEXT: br i1 [[CMP6]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP17:%.*]] = zext i32 [[TMP16]] to i64
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP19:%.*]] = zext i32 [[TMP18]] to i64
|
|
// CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP20]], ptr addrspace(5) [[N_CASTED]], align 4
|
|
// CHECK-NEXT: [[TMP21:%.*]] = load i64, ptr addrspace(5) [[N_CASTED]], align 8
|
|
// CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP23:%.*]] = inttoptr i64 [[TMP17]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP23]], ptr [[TMP22]], align 8
|
|
// CHECK-NEXT: [[TMP24:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 1
|
|
// CHECK-NEXT: [[TMP25:%.*]] = inttoptr i64 [[TMP19]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP25]], ptr [[TMP24]], align 8
|
|
// CHECK-NEXT: [[TMP26:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 2
|
|
// CHECK-NEXT: [[TMP27:%.*]] = inttoptr i64 [[TMP21]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP27]], ptr [[TMP26]], align 8
|
|
// CHECK-NEXT: [[TMP28:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 3
|
|
// CHECK-NEXT: store ptr [[SUM51_ASCAST]], ptr [[TMP28]], align 8
|
|
// CHECK-NEXT: [[TMP29:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 4
|
|
// CHECK-NEXT: [[TMP30:%.*]] = inttoptr i64 [[TMP1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP30]], ptr [[TMP29]], align 8
|
|
// CHECK-NEXT: [[TMP31:%.*]] = getelementptr inbounds [6 x ptr], ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 0, i64 5
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[TMP31]], align 8
|
|
// CHECK-NEXT: [[TMP32:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP33:%.*]] = load i32, ptr [[TMP32]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_parallel_60(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i32 [[TMP33]], i32 1, i32 -1, i32 -1, ptr @{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined_omp_outlined, ptr null, ptr [[CAPTURED_VARS_ADDRS_ASCAST]], i64 6, i32 0)
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP35:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD7:%.*]] = add nsw i32 [[TMP34]], [[TMP35]]
|
|
// CHECK-NEXT: store i32 [[ADD7]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP36]], [[TMP37]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP38:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP39:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD9:%.*]] = add nsw i32 [[TMP38]], [[TMP39]]
|
|
// CHECK-NEXT: store i32 [[ADD9]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP40:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP41:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP10:%.*]] = icmp sgt i32 [[TMP40]], [[TMP41]]
|
|
// CHECK-NEXT: br i1 [[CMP10]], label [[COND_TRUE11:%.*]], label [[COND_FALSE12:%.*]]
|
|
// CHECK: cond.true11:
|
|
// CHECK-NEXT: [[TMP42:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_2_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13:%.*]]
|
|
// CHECK: cond.false12:
|
|
// CHECK-NEXT: [[TMP43:%.*]] = load i32, ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[COND_END13]]
|
|
// CHECK: cond.end13:
|
|
// CHECK-NEXT: [[COND14:%.*]] = phi i32 [ [[TMP42]], [[COND_TRUE11]] ], [ [[TMP43]], [[COND_FALSE12]] ]
|
|
// CHECK-NEXT: store i32 [[COND14]], ptr [[DOTOMP_COMB_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP44:%.*]] = load i32, ptr [[DOTOMP_COMB_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP44]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP45:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP46:%.*]] = load i32, ptr [[TMP45]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_distribute_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB2]] to ptr), i32 [[TMP46]])
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: [[TMP47:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM51_ASCAST]], ptr [[TMP47]], align 8
|
|
// CHECK-NEXT: %"_openmp_teams_reductions_buffer_$_$ptr" = call ptr @__kmpc_reduction_get_fixed_buffer()
|
|
// CHECK-NEXT: [[TMP48:%.*]] = call i32 @__kmpc_nvptx_teams_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), ptr %"_openmp_teams_reductions_buffer_$_$ptr", i32 1024, i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.29, ptr @_omp_reduction_inter_warp_copy_func.30, ptr @_omp_reduction_list_to_global_copy_func.31, ptr @_omp_reduction_list_to_global_reduce_func.32, ptr @_omp_reduction_global_to_list_copy_func.33, ptr @_omp_reduction_global_to_list_reduce_func.34)
|
|
// CHECK-NEXT: [[TMP49:%.*]] = icmp eq i32 [[TMP48]], 1
|
|
// CHECK-NEXT: br i1 [[TMP49]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP50:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP51:%.*]] = load double, ptr [[SUM51_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD15:%.*]] = fadd double [[TMP50]], [[TMP51]]
|
|
// CHECK-NEXT: store double [[ADD15]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined_omp_outlined
|
|
// CHECK-SAME: (ptr noalias noundef [[DOTGLOBAL_TID_:%.*]], ptr noalias noundef [[DOTBOUND_TID_:%.*]], i64 noundef [[DOTPREVIOUS_LB_:%.*]], i64 noundef [[DOTPREVIOUS_UB_:%.*]], i64 noundef [[N:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[SUM5:%.*]], i64 noundef [[VLA:%.*]], ptr noundef nonnull align 8 dereferenceable(8) [[A:%.*]]) #[[ATTR1]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[SUM5_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[VLA_ADDR:%.*]] = alloca i64, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IV:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[TMP:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[J:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_LB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_UB:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[SUM54:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[J5:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: [[DOTGLOBAL_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTGLOBAL_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTBOUND_TID__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTBOUND_TID__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_LB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_LB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTPREVIOUS_UB__ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTPREVIOUS_UB__ADDR]] to ptr
|
|
// CHECK-NEXT: [[N_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[N_ADDR]] to ptr
|
|
// CHECK-NEXT: [[SUM5_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM5_ADDR]] to ptr
|
|
// CHECK-NEXT: [[VLA_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[VLA_ADDR]] to ptr
|
|
// CHECK-NEXT: [[A_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[A_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IV_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IV]] to ptr
|
|
// CHECK-NEXT: [[TMP_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[TMP]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR__ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_]] to ptr
|
|
// CHECK-NEXT: [[DOTCAPTURE_EXPR_1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCAPTURE_EXPR_1]] to ptr
|
|
// CHECK-NEXT: [[J_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_LB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_LB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_UB_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_UB]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_STRIDE_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_STRIDE]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_IS_LAST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_IS_LAST]] to ptr
|
|
// CHECK-NEXT: [[SUM54_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[SUM54]] to ptr
|
|
// CHECK-NEXT: [[J5_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[J5]] to ptr
|
|
// CHECK-NEXT: store ptr [[DOTGLOBAL_TID_]], ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[DOTBOUND_TID_]], ptr [[DOTBOUND_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_LB_]], ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[DOTPREVIOUS_UB_]], ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[N]], ptr [[N_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[SUM5]], ptr [[SUM5_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i64 [[VLA]], ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[SUM5_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr [[VLA_ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[A_ADDR_ASCAST]], align 8, !nonnull [[META9]], !align [[META11]]
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[N_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[SUB:%.*]] = sub nsw i32 [[TMP4]], 0
|
|
// CHECK-NEXT: [[DIV:%.*]] = sdiv i32 [[SUB]], 1
|
|
// CHECK-NEXT: [[SUB2:%.*]] = sub nsw i32 [[DIV]], 1
|
|
// CHECK-NEXT: store i32 [[SUB2]], ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[J_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR__ASCAST]], align 4
|
|
// CHECK-NEXT: [[CMP:%.*]] = icmp slt i32 0, [[TMP5]]
|
|
// CHECK-NEXT: br i1 [[CMP]], label [[OMP_PRECOND_THEN:%.*]], label [[OMP_PRECOND_END:%.*]]
|
|
// CHECK: omp.precond.then:
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCAPTURE_EXPR_1_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP6]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr [[DOTPREVIOUS_LB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV:%.*]] = trunc i64 [[TMP7]] to i32
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CONV3:%.*]] = trunc i64 [[TMP8]] to i32
|
|
// CHECK-NEXT: store i32 [[CONV]], ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[CONV3]], ptr [[DOTOMP_UB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 1, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTOMP_IS_LAST_ASCAST]], align 4
|
|
// CHECK-NEXT: store double 0.000000e+00, ptr [[SUM54_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr [[TMP9]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_init_4(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP10]], i32 33, ptr [[DOTOMP_IS_LAST_ASCAST]], ptr [[DOTOMP_LB_ASCAST]], ptr [[DOTOMP_UB_ASCAST]], ptr [[DOTOMP_STRIDE_ASCAST]], i32 1, i32 1)
|
|
// CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[DOTOMP_LB_ASCAST]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP11]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND:%.*]]
|
|
// CHECK: omp.inner.for.cond:
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CONV6:%.*]] = sext i32 [[TMP12]] to i64
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i64, ptr [[DOTPREVIOUS_UB__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[CMP7:%.*]] = icmp ule i64 [[CONV6]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[CMP7]], label [[OMP_INNER_FOR_BODY:%.*]], label [[OMP_INNER_FOR_END:%.*]]
|
|
// CHECK: omp.inner.for.body:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[MUL:%.*]] = mul nsw i32 [[TMP14]], 1
|
|
// CHECK-NEXT: [[ADD:%.*]] = add nsw i32 0, [[MUL]]
|
|
// CHECK-NEXT: store i32 [[ADD]], ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP15:%.*]] = load i32, ptr [[J5_ASCAST]], align 4
|
|
// CHECK-NEXT: [[CALL:%.*]] = call noundef i32 @_Z15compute_sum_resiPdS_(i32 noundef [[TMP15]], ptr noundef [[SUM54_ASCAST]], ptr noundef [[TMP2]]) #[[ATTR7]]
|
|
// CHECK-NEXT: br label [[OMP_BODY_CONTINUE:%.*]]
|
|
// CHECK: omp.body.continue:
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_INC:%.*]]
|
|
// CHECK: omp.inner.for.inc:
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load i32, ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP17:%.*]] = load i32, ptr [[DOTOMP_STRIDE_ASCAST]], align 4
|
|
// CHECK-NEXT: [[ADD8:%.*]] = add nsw i32 [[TMP16]], [[TMP17]]
|
|
// CHECK-NEXT: store i32 [[ADD8]], ptr [[DOTOMP_IV_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[OMP_INNER_FOR_COND]]
|
|
// CHECK: omp.inner.for.end:
|
|
// CHECK-NEXT: br label [[OMP_LOOP_EXIT:%.*]]
|
|
// CHECK: omp.loop.exit:
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load ptr, ptr [[DOTGLOBAL_TID__ADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
|
|
// CHECK-NEXT: call void @__kmpc_for_static_fini(ptr addrspacecast (ptr addrspace(1) @[[GLOB3]] to ptr), i32 [[TMP19]])
|
|
// CHECK-NEXT: [[TMP20:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: store ptr [[SUM54_ASCAST]], ptr [[TMP20]], align 8
|
|
// CHECK-NEXT: [[TMP21:%.*]] = call i32 @__kmpc_nvptx_parallel_reduce_nowait_v2(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr), i64 8, ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr @_omp_reduction_shuffle_and_reduce_func.27, ptr @_omp_reduction_inter_warp_copy_func.28)
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i32 [[TMP21]], 1
|
|
// CHECK-NEXT: br i1 [[TMP22]], label [[DOTOMP_REDUCTION_THEN:%.*]], label [[DOTOMP_REDUCTION_DONE:%.*]]
|
|
// CHECK: .omp.reduction.then:
|
|
// CHECK-NEXT: [[TMP23:%.*]] = load double, ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: [[TMP24:%.*]] = load double, ptr [[SUM54_ASCAST]], align 8
|
|
// CHECK-NEXT: [[ADD9:%.*]] = fadd double [[TMP23]], [[TMP24]]
|
|
// CHECK-NEXT: store double [[ADD9]], ptr [[TMP0]], align 8
|
|
// CHECK-NEXT: br label [[DOTOMP_REDUCTION_DONE]]
|
|
// CHECK: .omp.reduction.done:
|
|
// CHECK-NEXT: br label [[OMP_PRECOND_END]]
|
|
// CHECK: omp.precond.end:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.27
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.28
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_shuffle_and_reduce_func.29
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i16 noundef signext [[TMP1:%.*]], i16 noundef signext [[TMP2:%.*]], i16 noundef signext [[TMP3:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR3:%.*]] = alloca i16, align 2, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT:%.*]] = alloca double, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_ELEMENT_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_ELEMENT]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR3_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR3]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i16 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: store i16 [[TMP3]], ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i16, ptr [[DOTADDR1_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i16, ptr [[DOTADDR2_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load i16, ptr [[DOTADDR3_ASCAST]], align 2
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr double, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i64, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: [[TMP13:%.*]] = call i32 @__kmpc_get_warp_size()
|
|
// CHECK-NEXT: [[TMP14:%.*]] = trunc i32 [[TMP13]] to i16
|
|
// CHECK-NEXT: [[TMP15:%.*]] = call i64 @__kmpc_shuffle_int64(i64 [[TMP12]], i16 [[TMP6]], i16 [[TMP14]])
|
|
// CHECK-NEXT: store i64 [[TMP15]], ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[TMP9]], i64 1
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i64, ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], i64 1
|
|
// CHECK-NEXT: store ptr [[DOTOMP_REDUCTION_ELEMENT_ASCAST]], ptr [[TMP10]], align 8
|
|
// CHECK-NEXT: [[TMP18:%.*]] = icmp eq i16 [[TMP7]], 0
|
|
// CHECK-NEXT: [[TMP19:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP20:%.*]] = icmp ult i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP21:%.*]] = and i1 [[TMP19]], [[TMP20]]
|
|
// CHECK-NEXT: [[TMP22:%.*]] = icmp eq i16 [[TMP7]], 2
|
|
// CHECK-NEXT: [[TMP23:%.*]] = and i16 [[TMP5]], 1
|
|
// CHECK-NEXT: [[TMP24:%.*]] = icmp eq i16 [[TMP23]], 0
|
|
// CHECK-NEXT: [[TMP25:%.*]] = and i1 [[TMP22]], [[TMP24]]
|
|
// CHECK-NEXT: [[TMP26:%.*]] = icmp sgt i16 [[TMP6]], 0
|
|
// CHECK-NEXT: [[TMP27:%.*]] = and i1 [[TMP25]], [[TMP26]]
|
|
// CHECK-NEXT: [[TMP28:%.*]] = or i1 [[TMP18]], [[TMP21]]
|
|
// CHECK-NEXT: [[TMP29:%.*]] = or i1 [[TMP28]], [[TMP27]]
|
|
// CHECK-NEXT: br i1 [[TMP29]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP4]], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[TMP30:%.*]] = icmp eq i16 [[TMP7]], 1
|
|
// CHECK-NEXT: [[TMP31:%.*]] = icmp uge i16 [[TMP5]], [[TMP6]]
|
|
// CHECK-NEXT: [[TMP32:%.*]] = and i1 [[TMP30]], [[TMP31]]
|
|
// CHECK-NEXT: br i1 [[TMP32]], label [[THEN4:%.*]], label [[ELSE5:%.*]]
|
|
// CHECK: then4:
|
|
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_REMOTE_REDUCE_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP34:%.*]] = load ptr, ptr [[TMP33]], align 8
|
|
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP4]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP36:%.*]] = load ptr, ptr [[TMP35]], align 8
|
|
// CHECK-NEXT: [[TMP37:%.*]] = load double, ptr [[TMP34]], align 8
|
|
// CHECK-NEXT: store double [[TMP37]], ptr [[TMP36]], align 8
|
|
// CHECK-NEXT: br label [[IFCONT6:%.*]]
|
|
// CHECK: else5:
|
|
// CHECK-NEXT: br label [[IFCONT6]]
|
|
// CHECK: ifcont6:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_inter_warp_copy_func.30
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTCNT_ADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTCNT_ADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[TMP3:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_LANE_ID:%.*]] = and i32 [[TMP3]], 63
|
|
// CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block()
|
|
// CHECK-NEXT: [[NVPTX_WARP_ID:%.*]] = ashr i32 [[TMP4]], 6
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 0, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND:%.*]]
|
|
// CHECK: precond:
|
|
// CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP7:%.*]] = icmp ult i32 [[TMP6]], 2
|
|
// CHECK-NEXT: br i1 [[TMP7]], label [[BODY:%.*]], label [[EXIT:%.*]]
|
|
// CHECK: body:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM]])
|
|
// CHECK-NEXT: [[WARP_MASTER:%.*]] = icmp eq i32 [[NVPTX_LANE_ID]], 0
|
|
// CHECK-NEXT: br i1 [[WARP_MASTER]], label [[THEN:%.*]], label [[ELSE:%.*]]
|
|
// CHECK: then:
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP9:%.*]] = load ptr, ptr [[TMP8]], align 8
|
|
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[NVPTX_WARP_ID]]
|
|
// CHECK-NEXT: [[TMP12:%.*]] = load i32, ptr [[TMP10]], align 4
|
|
// CHECK-NEXT: store volatile i32 [[TMP12]], ptr addrspace(3) [[TMP11]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT:%.*]]
|
|
// CHECK: else:
|
|
// CHECK-NEXT: br label [[IFCONT]]
|
|
// CHECK: ifcont:
|
|
// CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr addrspacecast (ptr addrspace(1) @[[GLOB1]] to ptr))
|
|
// CHECK-NEXT: call void @__kmpc_barrier(ptr addrspacecast (ptr addrspace(1) @[[GLOB4]] to ptr), i32 [[OMP_GLOBAL_THREAD_NUM2]])
|
|
// CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[IS_ACTIVE_THREAD:%.*]] = icmp ult i32 [[TMP2]], [[TMP13]]
|
|
// CHECK-NEXT: br i1 [[IS_ACTIVE_THREAD]], label [[THEN3:%.*]], label [[ELSE4:%.*]]
|
|
// CHECK: then3:
|
|
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [64 x i32], ptr addrspace(3) @__openmp_nvptx_data_transfer_temporary_storage, i64 0, i32 [[TMP2]]
|
|
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP5]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP16:%.*]] = load ptr, ptr [[TMP15]], align 8
|
|
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 [[TMP6]]
|
|
// CHECK-NEXT: [[TMP18:%.*]] = load volatile i32, ptr addrspace(3) [[TMP14]], align 4
|
|
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP17]], align 4
|
|
// CHECK-NEXT: br label [[IFCONT5:%.*]]
|
|
// CHECK: else4:
|
|
// CHECK-NEXT: br label [[IFCONT5]]
|
|
// CHECK: ifcont5:
|
|
// CHECK-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP6]], 1
|
|
// CHECK-NEXT: store i32 [[TMP19]], ptr [[DOTCNT_ADDR_ASCAST]], align 4
|
|
// CHECK-NEXT: br label [[PRECOND]]
|
|
// CHECK: exit:
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_copy_func.31
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_list_to_global_reduce_func.32
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined_omp$reduction$reduction_func"(ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], ptr [[TMP8]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_copy_func.33
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [1 x ptr], ptr [[TMP3]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP7:%.*]] = load ptr, ptr [[TMP6]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP4]], i32 [[TMP5]]
|
|
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP8]], i32 0, i32 0
|
|
// CHECK-NEXT: [[TMP10:%.*]] = load double, ptr [[TMP9]], align 8
|
|
// CHECK-NEXT: store double [[TMP10]], ptr [[TMP7]], align 8
|
|
// CHECK-NEXT: ret void
|
|
//
|
|
//
|
|
// CHECK-LABEL: define {{[^@]+}}@_omp_reduction_global_to_list_reduce_func.34
|
|
// CHECK-SAME: (ptr noundef [[TMP0:%.*]], i32 noundef [[TMP1:%.*]], ptr noundef [[TMP2:%.*]]) #[[ATTR4]] {
|
|
// CHECK-NEXT: entry:
|
|
// CHECK-NEXT: [[DOTADDR:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR2:%.*]] = alloca ptr, align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST:%.*]] = alloca [1 x ptr], align 8, addrspace(5)
|
|
// CHECK-NEXT: [[DOTADDR_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR1_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR1]] to ptr
|
|
// CHECK-NEXT: [[DOTADDR2_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTADDR2]] to ptr
|
|
// CHECK-NEXT: [[DOTOMP_REDUCTION_RED_LIST_ASCAST:%.*]] = addrspacecast ptr addrspace(5) [[DOTOMP_REDUCTION_RED_LIST]] to ptr
|
|
// CHECK-NEXT: store ptr [[TMP0]], ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: store i32 [[TMP1]], ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: store ptr [[TMP2]], ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP3:%.*]] = load ptr, ptr [[DOTADDR_ASCAST]], align 8
|
|
// CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTADDR1_ASCAST]], align 4
|
|
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds [1 x ptr], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]], i64 0, i64 0
|
|
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10:%.*]], ptr [[TMP3]], i32 [[TMP4]]
|
|
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds [[STRUCT__GLOBALIZED_LOCALS_TY_10]], ptr [[TMP6]], i32 0, i32 0
|
|
// CHECK-NEXT: store ptr [[TMP7]], ptr [[TMP5]], align 8
|
|
// CHECK-NEXT: [[TMP8:%.*]] = load ptr, ptr [[DOTADDR2_ASCAST]], align 8
|
|
// CHECK-NEXT: call void @"{{__omp_offloading_[0-9a-z]+_[0-9a-z]+}}_main_l50_omp_outlined_omp$reduction$reduction_func"(ptr [[TMP8]], ptr [[DOTOMP_REDUCTION_RED_LIST_ASCAST]]) #[[ATTR2]]
|
|
// CHECK-NEXT: ret void
|
|
//
|