Skip to content

Commit 00de635

Browse files
agozillonagozillon
committed
[MLIR][OpenMP] Introduce overlapped record type map support
This PR introduces a new additional type of map lowering for record types that Clang currently supports, in which a user can map a top-level record type and then individual members with different mapping, effectively creating a sort of "overlapping" mapping that we attempt to cut around. This is currently most predominantly used in Fortran, when mapping descriptors and there data, we map the descriptor and its data with separate map modifiers and "cut around" the pointer data, so that wedo not overwrite it unless the runtime deems it a neccesary action based on its reference counting mechanism. However, it is a mechanism that will come in handy/trigger when a user explitily maps a record type (derived type or structure) and then explicitly maps a member with a different map type. These additions were predominantly in the OpenMPToLLVMIRTranslation.cpp file and phase, however, one Flang test that checks end-to-end IR compilation (as far as we care for now at least) was altered. 2/3 required PRs to enable declare target to mapping, should look at PR 3/3 to check for full green passes (this one will fail a number due to some dependencies). Co-authored-by: Raghu Maddhipatla [email protected]
1 parent 83765f4 commit 00de635

File tree

8 files changed

+551
-197
lines changed

8 files changed

+551
-197
lines changed

‎flang/test/Integration/OpenMP/map-types-and-sizes.f90‎

Lines changed: 84 additions & 59 deletions
Large diffs are not rendered by default.

‎mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp‎

Lines changed: 225 additions & 73 deletions
Large diffs are not rendered by default.

‎mlir/test/Target/LLVMIR/omptarget-data-use-dev-ordering.mlir‎

Lines changed: 10 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -67,18 +67,18 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["amdgcn-a
6767

6868
// CHECK: define void @mix_use_device_ptr_and_addr_and_map_(ptr %[[ARG_0:.*]], ptr %[[ARG_1:.*]], ptr %[[ARG_2:.*]], ptr %[[ARG_3:.*]], ptr %[[ARG_4:.*]], ptr %[[ARG_5:.*]], ptr %[[ARG_6:.*]], ptr %[[ARG_7:.*]]) {
6969
// CHECK: %[[ALLOCA:.*]] = alloca ptr, align 8
70-
// CHECK: %[[BASEPTR_0_GEP:.*]] = getelementptr inbounds [10 x ptr], ptr %.offload_baseptrs, i32 0, i32 0
70+
// CHECK: %[[BASEPTR_0_GEP:.*]] = getelementptr inbounds [12 x ptr], ptr %.offload_baseptrs, i32 0, i32 0
7171
// CHECK: store ptr %[[ARG_0]], ptr %[[BASEPTR_0_GEP]], align 8
72-
// CHECK: %[[BASEPTR_2_GEP:.*]] = getelementptr inbounds [10 x ptr], ptr %.offload_baseptrs, i32 0, i32 2
72+
// CHECK: %[[BASEPTR_2_GEP:.*]] = getelementptr inbounds [12 x ptr], ptr %.offload_baseptrs, i32 0, i32 4
7373
// CHECK: store ptr %[[ARG_2]], ptr %[[BASEPTR_2_GEP]], align 8
74-
// CHECK: %[[BASEPTR_6_GEP:.*]] = getelementptr inbounds [10 x ptr], ptr %.offload_baseptrs, i32 0, i32 6
75-
// CHECK: store ptr %[[ARG_4]], ptr %[[BASEPTR_6_GEP]], align 8
74+
// CHECK: %[[BASEPTR_3_GEP:.*]] = getelementptr inbounds [12 x ptr], ptr %.offload_baseptrs, i32 0, i32 9
75+
// CHECK: store ptr %[[ARG_4]], ptr %[[BASEPTR_3_GEP]], align 8
7676

7777
// CHECK: call void @__tgt_target_data_begin_mapper({{.*}})
7878
// CHECK: %[[LOAD_BASEPTR_0:.*]] = load ptr, ptr %[[BASEPTR_0_GEP]], align 8
7979
// store ptr %[[LOAD_BASEPTR_0]], ptr %[[ALLOCA]], align 8
8080
// CHECK: %[[LOAD_BASEPTR_2:.*]] = load ptr, ptr %[[BASEPTR_2_GEP]], align 8
81-
// CHECK: %[[LOAD_BASEPTR_6:.*]] = load ptr, ptr %[[BASEPTR_6_GEP]], align 8
81+
// CHECK: %[[LOAD_BASEPTR_3:.*]] = load ptr, ptr %[[BASEPTR_3_GEP]], align 8
8282
// CHECK: %[[GEP_A4:.*]] = getelementptr { i64 }, ptr %[[ARG_4]], i32 0, i32 0
8383
// CHECK: %[[GEP_A7:.*]] = getelementptr { i64 }, ptr %[[ARG_7]], i32 0, i32 0
8484
// CHECK: %[[LOAD_A4:.*]] = load i64, ptr %[[GEP_A4]], align 4
@@ -93,17 +93,17 @@ module attributes {omp.is_target_device = false, omp.target_triples = ["amdgcn-a
9393

9494
// CHECK: define void @mix_use_device_ptr_and_addr_and_map_2(ptr %[[ARG_0:.*]], ptr %[[ARG_1:.*]], ptr %[[ARG_2:.*]], ptr %[[ARG_3:.*]], ptr %[[ARG_4:.*]], ptr %[[ARG_5:.*]], ptr %[[ARG_6:.*]], ptr %[[ARG_7:.*]]) {
9595
// CHECK: %[[ALLOCA:.*]] = alloca ptr, align 8
96-
// CHECK: %[[BASEPTR_1_GEP:.*]] = getelementptr inbounds [10 x ptr], ptr %.offload_baseptrs, i32 0, i32 1
96+
// CHECK: %[[BASEPTR_1_GEP:.*]] = getelementptr inbounds [12 x ptr], ptr %.offload_baseptrs, i32 0, i32 1
9797
// CHECK: store ptr %[[ARG_0]], ptr %[[BASEPTR_1_GEP]], align 8
98-
// CHECK: %[[BASEPTR_2_GEP:.*]] = getelementptr inbounds [10 x ptr], ptr %.offload_baseptrs, i32 0, i32 2
98+
// CHECK: %[[BASEPTR_2_GEP:.*]] = getelementptr inbounds [12 x ptr], ptr %.offload_baseptrs, i32 0, i32 4
9999
// CHECK: store ptr %[[ARG_2]], ptr %[[BASEPTR_2_GEP]], align 8
100-
// CHECK: %[[BASEPTR_6_GEP:.*]] = getelementptr inbounds [10 x ptr], ptr %.offload_baseptrs, i32 0, i32 6
101-
// CHECK: store ptr %[[ARG_4]], ptr %[[BASEPTR_6_GEP]], align 8
100+
// CHECK: %[[BASEPTR_3_GEP:.*]] = getelementptr inbounds [12 x ptr], ptr %.offload_baseptrs, i32 0, i32 9
101+
// CHECK: store ptr %[[ARG_4]], ptr %[[BASEPTR_3_GEP]], align 8
102102
// CHECK: call void @__tgt_target_data_begin_mapper({{.*}})
103103
// CHECK: %[[LOAD_BASEPTR_1:.*]] = load ptr, ptr %[[BASEPTR_1_GEP]], align 8
104104
// store ptr %[[LOAD_BASEPTR_1]], ptr %[[ALLOCA]], align 8
105105
// CHECK: %[[LOAD_BASEPTR_2:.*]] = load ptr, ptr %[[BASEPTR_2_GEP]], align 8
106-
// CHECK: %[[LOAD_BASEPTR_6:.*]] = load ptr, ptr %[[BASEPTR_6_GEP]], align 8
106+
// CHECK: %[[LOAD_BASEPTR_3:.*]] = load ptr, ptr %[[BASEPTR_3_GEP]], align 8
107107
// CHECK: %[[GEP_A4:.*]] = getelementptr { i64 }, ptr %[[ARG_4]], i32 0, i32 0
108108
// CHECK: %[[GEP_A7:.*]] = getelementptr { i64 }, ptr %[[ARG_7]], i32 0, i32 0
109109
// CHECK: %[[LOAD_A4:.*]] = load i64, ptr %[[GEP_A4]], align 4

‎mlir/test/Target/LLVMIR/omptarget-declare-target-to-host.mlir‎

Lines changed: 34 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,34 @@
1+
// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
2+
3+
module attributes {llvm.target_triple = "x86_64-unknown-linux-gnu", omp.is_gpu = false, omp.is_target_device = false, omp.target_triples = ["amdgcn-amd-amdhsa"]} {
4+
// CHECK-DAG: @_QMtest_0Ezii = global [11 x float] zeroinitializer
5+
// CHECK-DAG: @.offload_sizes = private unnamed_addr constant [1 x i64] [i64 48]
6+
// CHECK-DAG: @.offload_maptypes = private unnamed_addr constant [1 x i64] [i64 3]
7+
// CHECK-DAG: @.offloading.entry._QMtest_0Ezii = weak constant %struct.__tgt_offload_entry {{.*}} ptr @_QMtest_0Ezii, {{.*}}, i64 44,{{.*}}
8+
llvm.mlir.global external @_QMtest_0Ezii() {addr_space = 0 : i32, omp.declare_target = #omp.declaretarget<device_type = (any), capture_clause = (to)>} : !llvm.array<11 x f32> {
9+
%0 = llvm.mlir.zero : !llvm.array<11 x f32>
10+
llvm.return %0 : !llvm.array<11 x f32>
11+
}
12+
13+
// CHECK-DAG: %[[BASEPTR:.*]] = getelementptr inbounds [1 x ptr], ptr %.offload_baseptrs, i32 0, i32 0
14+
// CHECK-DAG: store ptr @_QMtest_0Ezii, ptr %[[BASEPTR]], align 8
15+
// CHECK-DAG: %[[OFFLOADPTR:.*]] = getelementptr inbounds [1 x ptr], ptr %.offload_ptrs, i32 0, i32 0
16+
// CHECK-DAG: store ptr @_QMtest_0Ezii, ptr %[[OFFLOADPTR]], align 8
17+
llvm.func @_QQmain() {
18+
%0 = llvm.mlir.constant(1 : index) : i64
19+
%1 = llvm.mlir.constant(0 : index) : i64
20+
%2 = llvm.mlir.constant(11 : index) : i64
21+
%3 = llvm.mlir.addressof @_QMtest_0Ezii : !llvm.ptr
22+
%4 = omp.map.bounds lower_bound(%1 : i64) upper_bound(%2 : i64) extent(%2 : i64) stride(%0 : i64) start_idx(%1 : i64) {stride_in_bytes = true}
23+
%5 = omp.map.info var_ptr(%3 : !llvm.ptr, !llvm.array<11 x f32>) map_clauses(tofrom) capture(ByRef) bounds(%4) -> !llvm.ptr
24+
omp.target map_entries(%5 -> %arg0 : !llvm.ptr) {
25+
%6 = llvm.mlir.constant(1.0 : f32) : f32
26+
%7 = llvm.mlir.constant(0 : i64) : i64
27+
%8 = llvm.getelementptr %arg0[%7] : (!llvm.ptr, i64) -> !llvm.ptr, f32
28+
llvm.store %6, %8 : f32, !llvm.ptr
29+
omp.terminator
30+
}
31+
llvm.return
32+
}
33+
// CHEKC-DAG: !{{.*}} = !{i32 {{.*}}, !"_QMtest_0Ezii", i32 {{.*}}, i32 {{.*}}}
34+
}

‎mlir/test/Target/LLVMIR/omptarget-nowait.mlir‎

Lines changed: 13 additions & 14 deletions
Original file line numberDiff line numberDiff line change
@@ -25,34 +25,33 @@ module attributes {omp.target_triples = ["amdgcn-amd-amdhsa"]} {
2525

2626
// CHECK: %struct.[[TSK_WTH_PRVTS:.*]] = type { %struct.kmp_task_ompbuilder_t, %struct.[[PRVTS:.*]] }
2727
// CHECK: %struct.kmp_task_ompbuilder_t = type { ptr, ptr, i32, ptr, ptr }
28-
// CHECK: %struct.[[PRVTS]] = type { [5 x ptr], [5 x ptr], [5 x i64] }
28+
// CHECK: %struct.[[PRVTS]] = type { [6 x ptr], [6 x ptr], [6 x i64] }
2929

3030
// CHECK: define void @launch_(ptr captures(none) %0)
3131
// CHECK: %[[STRUCTARG:.*]] = alloca { ptr, ptr }, align 8
32-
// CHECK: %[[BASEPTRS:.*]] = alloca [5 x ptr], align 8
33-
// CHECK: %[[PTRS:.*]] = alloca [5 x ptr], align 8
34-
// CHECK: %[[MAPPERS:.*]] = alloca [5 x ptr], align 8
35-
// CHECK: %[[SIZES:.*]] = alloca [5 x i64], align 4
32+
// CHECK: %[[BASEPTRS:.*]] = alloca [6 x ptr], align 8
33+
// CHECK: %[[PTRS:.*]] = alloca [6 x ptr], align 8
34+
// CHECK: %[[MAPPERS:.*]] = alloca [6 x ptr], align 8
35+
// CHECK: %[[SIZES:.*]] = alloca [6 x i64], align 4
3636

37-
38-
// CHECK: %[[VAL_20:.*]] = getelementptr inbounds [5 x ptr], ptr %[[BASEPTRS]], i32 0, i32 0
39-
// CHECK: %[[BASEPTRS_GEP:.*]] = getelementptr inbounds [5 x ptr], ptr %[[BASEPTRS]], i32 0, i32 0
40-
// CHECK: %[[PTRS_GEP:.*]] = getelementptr inbounds [5 x ptr], ptr %[[PTRS]], i32 0, i32 0
41-
// CHECK: %[[SIZES_GEP:.*]] = getelementptr inbounds [5 x i64], ptr %[[SIZES]], i32 0, i32 0
37+
// CHECK: %[[VAL_20:.*]] = getelementptr inbounds [6 x ptr], ptr %[[BASEPTRS]], i32 0, i32 0
38+
// CHECK: %[[BASEPTRS_GEP:.*]] = getelementptr inbounds [6 x ptr], ptr %[[BASEPTRS]], i32 0, i32 0
39+
// CHECK: %[[PTRS_GEP:.*]] = getelementptr inbounds [6 x ptr], ptr %[[PTRS]], i32 0, i32 0
40+
// CHECK: %[[SIZES_GEP:.*]] = getelementptr inbounds [6 x i64], ptr %[[SIZES]], i32 0, i32 0
4241

4342
// CHECK: %[[GL_THRD_NUM:.*]] = call i32 @__kmpc_global_thread_num
44-
// CHECK: %[[TASK_DESC:.*]] = call ptr @__kmpc_omp_target_task_alloc(ptr @4, i32 {{.*}}, i32 0, i64 160, i64 16, ptr [[TGT_TSK_PRXY_FNC:.*]], i64 -1)
43+
// CHECK: %[[TASK_DESC:.*]] = call ptr @__kmpc_omp_target_task_alloc(ptr @4, i32 {{.*}}, i32 0, i64 184, i64 16, ptr [[TGT_TSK_PRXY_FNC:.*]], i64 -1)
4544
// CHECK: %[[TSK_PTR:.*]] = getelementptr inbounds nuw %struct.[[TSK_WTH_PRVTS]], ptr %[[TASK_DESC]], i32 0, i32 0
4645
// CHECK: %[[SHAREDS:.*]] = getelementptr inbounds nuw %struct.kmp_task_ompbuilder_t, ptr %[[TSK_PTR]], i32 0, i32 0
4746
// CHECK: %[[SHAREDS_PTR:.*]] = load ptr, ptr %[[SHAREDS]], align 8
4847
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[SHAREDS_PTR]], ptr align 1 %[[STRUCTARG]], i64 16, i1 false)
4948
// CHECK: %[[VAL_50:.*]] = getelementptr inbounds nuw %struct.[[TSK_WTH_PRVTS]], ptr %[[TASK_DESC]], i32 0, i32 1
5049
// CHECK: %[[VAL_51:.*]] = getelementptr inbounds nuw %struct.[[PRVTS]], ptr %[[VAL_50]], i32 0, i32 0
51-
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_51]], ptr align 1 %[[BASEPTRS_GEP]], i64 40, i1 false)
50+
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_51]], ptr align 1 %[[BASEPTRS_GEP]], i64 48, i1 false)
5251
// CHECK: %[[VAL_53:.*]] = getelementptr inbounds nuw %struct.[[PRVTS]], ptr %[[VAL_50]], i32 0, i32 1
53-
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_53]], ptr align 1 %[[PTRS_GEP]], i64 40, i1 false)
52+
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_53]], ptr align 1 %[[PTRS_GEP]], i64 48, i1 false)
5453
// CHECK: %[[VAL_54:.*]] = getelementptr inbounds nuw %struct.[[PRVTS]], ptr %[[VAL_50]], i32 0, i32 2
55-
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_54]], ptr align 1 %[[SIZES_GEP]], i64 40, i1 false)
54+
// CHECK: call void @llvm.memcpy.p0.p0.i64(ptr align 1 %[[VAL_54]], ptr align 1 %[[SIZES_GEP]], i64 48, i1 false)
5655
// CHECK: %[[VAL_55:.*]] = call i32 @__kmpc_omp_task(ptr @4, i32 %[[GL_THRD_NUM]], ptr %[[TASK_DESC]])
5756

5857
// CHECK: define internal void @[[WORKER:.*]](i32 {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}, ptr {{.*}}) {

‎mlir/test/Target/LLVMIR/omptarget-overlapping-record-member-map.mlir‎

Lines changed: 65 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,65 @@
1+
// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
2+
3+
module attributes {llvm.target_triple = "x86_64-unknown-linux-gnu", omp.is_gpu = false, omp.is_target_device = false, omp.target_triples = ["amdgcn-amd-amdhsa"]} {
4+
llvm.func @_QQmain() attributes {fir.bindc_name = "main"} {
5+
%0 = llvm.mlir.constant(1 : i64) : i64
6+
%1 = llvm.alloca %0 x !llvm.struct<"_QFTdtype", (f32, i32)> {bindc_name = "dtypev"} : (i64) -> !llvm.ptr
7+
%2 = llvm.getelementptr %1[0, 1] : (!llvm.ptr) -> !llvm.ptr, !llvm.struct<"_QFTdtype", (f32, i32)>
8+
%3 = omp.map.info var_ptr(%2 : !llvm.ptr, i32) map_clauses(tofrom) capture(ByRef) -> !llvm.ptr {name = "dtypev%value2"}
9+
%4 = omp.map.info var_ptr(%1 : !llvm.ptr, !llvm.struct<"_QFTdtype", (f32, i32)>) map_clauses(to) capture(ByRef) members(%3 : [1] : !llvm.ptr) -> !llvm.ptr {name = "dtypev"}
10+
omp.target map_entries(%4 -> %arg0, %3 -> %arg1 : !llvm.ptr, !llvm.ptr) {
11+
omp.terminator
12+
}
13+
llvm.return
14+
}
15+
}
16+
17+
// CHECK: @.offload_sizes = private unnamed_addr constant [4 x i64] [i64 0, i64 0, i64 0, i64 4]
18+
// CHECK: @.offload_maptypes = private unnamed_addr constant [4 x i64] [i64 32, i64 281474976710657, i64 281474976710657, i64 281474976710659]
19+
20+
// CHECK: %[[ALLOCA:.*]] = alloca %_QFTdtype, i64 1, align 8
21+
// CHECK: %[[ELEMENT_ACC:.*]] = getelementptr %_QFTdtype, ptr %[[ALLOCA]], i32 0, i32 1
22+
23+
// CHECK: %[[SIZE1_CALC_1:.*]] = getelementptr %_QFTdtype, ptr %[[ALLOCA]], i32 1
24+
// CHECK: %[[SIZE1_CALC_2:.*]] = ptrtoint ptr %[[SIZE1_CALC_1]] to i64
25+
// CHECK: %[[SIZE1_CALC_3:.*]] = ptrtoint ptr %[[ALLOCA]] to i64
26+
// CHECK: %[[SIZE1_CALC_4:.*]] = sub i64 %[[SIZE1_CALC_2]], %[[SIZE1_CALC_3]]
27+
// CHECK: %[[SIZE1_CALC_5:.*]] = sdiv exact i64 %[[SIZE1_CALC_4]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
28+
29+
// CHECK: %[[SIZE2_CALC_1:.*]] = getelementptr %_QFTdtype, ptr %[[ALLOCA]], i32 1
30+
// CHECK: %[[SIZE2_CALC_2:.*]] = ptrtoint ptr %[[ELEMENT_ACC]] to i64
31+
// CHECK: %[[SIZE2_CALC_3:.*]] = ptrtoint ptr %[[ALLOCA]] to i64
32+
// CHECK: %[[SIZE2_CALC_4:.*]] = sub i64 %[[SIZE2_CALC_2]], %[[SIZE2_CALC_3]]
33+
// CHECK: %[[SIZE2_CALC_5:.*]] = sdiv exact i64 %[[SIZE2_CALC_4]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
34+
35+
// CHECK: %[[SIZE3_CALC_1:.*]] = getelementptr i32, ptr %[[ELEMENT_ACC]], i32 1
36+
// CHECK: %[[SIZE3_CALC_2:.*]] = ptrtoint ptr %[[SIZE2_CALC_1]] to i64
37+
// CHECK: %[[SIZE3_CALC_3:.*]] = ptrtoint ptr %[[SIZE3_CALC_1]] to i64
38+
// CHECK: %[[SIZE3_CALC_4:.*]] = sub i64 %[[SIZE3_CALC_2]], %[[SIZE3_CALC_3]]
39+
// CHECK: %[[SIZE3_CALC_5:.*]] = sdiv exact i64 %[[SIZE3_CALC_4]], ptrtoint (ptr getelementptr (i8, ptr null, i32 1) to i64)
40+
41+
// CHECK: %[[BASEPTR:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_baseptrs, i32 0, i32 0
42+
// CHECK: store ptr %[[ALLOCA]], ptr %[[BASEPTR]], align 8
43+
// CHECK: %[[PTRS:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_ptrs, i32 0, i32 0
44+
// CHECK: store ptr %[[ALLOCA]], ptr %[[PTRS]], align 8
45+
// CHECK: %[[SIZES:.*]] = getelementptr inbounds [4 x i64], ptr %.offload_sizes, i32 0, i32 0
46+
// CHECK: store i64 %[[SIZE1_CALC_5]], ptr %[[SIZES]], align 8
47+
48+
// CHECK: %[[BASEPTR:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_baseptrs, i32 0, i32 1
49+
// CHECK: store ptr %[[ALLOCA]], ptr %[[BASEPTR]], align 8
50+
// CHECK: %[[PTRS:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_ptrs, i32 0, i32 1
51+
// CHECK: store ptr %[[ALLOCA]], ptr %[[PTRS]], align 8
52+
// CHECK: %[[SIZES:.*]] = getelementptr inbounds [4 x i64], ptr %.offload_sizes, i32 0, i32 1
53+
// CHECK: store i64 %[[SIZE2_CALC_5]], ptr %[[SIZES]], align 8
54+
55+
// CHECK: %[[BASEPTR:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_baseptrs, i32 0, i32 2
56+
// CHECK: store ptr %[[ALLOCA]], ptr %[[BASEPTR]], align 8
57+
// CHECK: %[[PTRS:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_ptrs, i32 0, i32 2
58+
// CHECK: store ptr %13, ptr %[[PTRS]], align 8
59+
// CHECK: %[[SIZES:.*]] = getelementptr inbounds [4 x i64], ptr %.offload_sizes, i32 0, i32 2
60+
// CHECK: store i64 %[[SIZE3_CALC_5]], ptr %[[SIZES]], align 8
61+
62+
// CHECK: %[[BASEPTR:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_baseptrs, i32 0, i32 3
63+
// CHECK: store ptr %[[ALLOCA]], ptr %[[BASEPTR]], align 8
64+
// CHECK: %[[PTRS:.*]] = getelementptr inbounds [4 x ptr], ptr %.offload_ptrs, i32 0, i32 3
65+
// CHECK: store ptr %[[ELEMENT_ACC]], ptr %[[PTRS]], align 8

0 commit comments

Comments
 (0)