[clang] [OpenACC][CIR] Implement member exprs for 'copy' lowering (PR #142998)

[Erich Keane via cfe-commits]

================
@@ -754,12 +754,320 @@ void acc_compute(int parmVar) {
   // CHECK-NEXT: %[[ZERO_CONST:.*]] = arith.constant 0 : i64
   // CHECK-NEXT: %[[ONE_CONST2:.*]] = arith.constant 1 : i64
   // CHECK-NEXT: %[[BOUNDS3:.*]] = acc.bounds lowerbound(%[[ONE_CAST]] : si32) extent(%[[ONE_CAST2]] : si32) stride(%[[ONE_CONST2]] : i64) startIdx(%[[ZERO_CONST]] : i64) loc
-  // CHECK-NEXT: %[[COPYIN1:.*]] = acc.copyin varPtr(%[[THREEDARRAY]] : !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>>) bounds(%[[BOUNDS]], %[[BOUNDS2]], %[[BOUNDS3]]) -> !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>> {dataClause = #acc<data_clause acc_copy>, name = "threeDArray"} loc
+  // CHECK-NEXT: %[[COPYIN1:.*]] = acc.copyin varPtr(%[[THREEDARRAY]] : !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>>) bounds(%[[BOUNDS]], %[[BOUNDS2]], %[[BOUNDS3]]) -> !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>> {dataClause = #acc<data_clause acc_copy>, name = "threeDArray[1:1][2:1][3:1]"} loc
   // CHECK-NEXT: acc.parallel combined(loop) dataOperands(%[[COPYIN1]] : !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>>) {
   // CHECK-NEXT: acc.loop combined(parallel) {
   // CHECK: acc.yield
   // CHECK-NEXT: }
   // CHECK-NEXT: acc.yield
   // CHECK-NEXT: } loc
-  // CHECK-NEXT: acc.copyout accPtr(%[[COPYIN1]] : !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>>) bounds(%[[BOUNDS]], %[[BOUNDS2]], %[[BOUNDS3]]) to varPtr(%[[THREEDARRAY]] : !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>>) {dataClause = #acc<data_clause acc_copy>, name = "threeDArray"} loc
+  // CHECK-NEXT: acc.copyout accPtr(%[[COPYIN1]] : !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>>) bounds(%[[BOUNDS]], %[[BOUNDS2]], %[[BOUNDS3]]) to varPtr(%[[THREEDARRAY]] : !cir.ptr<!cir.array<!cir.array<!cir.array<!cir.double x 7> x 6> x 5>>) {dataClause = #acc<data_clause acc_copy>, name = "threeDArray[1:1][2:1][3:1]"} loc
+}
+
+typedef struct StructTy {
+  int scalarMember;
+  int arrayMember[5];
+  short twoDArrayMember[5][3];
+  float *ptrArrayMember[5];
+  double **ptrPtrMember;
+} Struct ;
+
+void acc_compute_members() {
+  // CHECK: cir.func @acc_compute_members()
+  Struct localStruct;
+  // CHECK-NEXT: %[[LOCALSTRUCT:.*]] = cir.alloca !rec_StructTy, !cir.ptr<!rec_StructTy>, ["localStruct"]
+
+#pragma acc parallel loop copy(localStruct)
+  for(int i = 0; i < 5; ++i);
+  // CHECK-NEXT: %[[COPYIN1:.*]] = acc.copyin varPtr(%[[LOCALSTRUCT]] : !cir.ptr<!rec_StructTy>) -> !cir.ptr<!rec_StructTy> {dataClause = #acc<data_clause acc_copy>, name = "localStruct"}
+  // CHECK-NEXT: acc.parallel combined(loop) dataOperands(%[[COPYIN1]] : !cir.ptr<!rec_StructTy>) {
+  // CHECK-NEXT: acc.loop combined(parallel) {
+  // CHECK: acc.yield
+  // CHECK-NEXT: }
+  // CHECK-NEXT: acc.yield
+  // CHECK-NEXT: } loc
+  // CHECK-NEXT: acc.copyout accPtr(%[[COPYIN1]] : !cir.ptr<!rec_StructTy>) to varPtr(%[[LOCALSTRUCT]] : !cir.ptr<!rec_StructTy>) {dataClause = #acc<data_clause acc_copy>, name = "localStruct"}
+
+#pragma acc serial loop copy(localStruct.scalarMember)
+  for(int i = 0; i < 5; ++i);
+  // CHECK-NEXT: %[[GETMEMBER:.*]] = cir.get_member %[[LOCALSTRUCT]][0] {name = "scalarMember"} : !cir.ptr<!rec_StructTy> -> !cir.ptr<!s32i>
+  // CHECK-NEXT: %[[COPYIN1:.*]] = acc.copyin varPtr(%[[GETMEMBER]] : !cir.ptr<!s32i>) -> !cir.ptr<!s32i> {dataClause = #acc<data_clause acc_copy>, name = "localStruct.scalarMember"}
+  // CHECK-NEXT: acc.serial combined(loop) dataOperands(%[[COPYIN1]] : !cir.ptr<!s32i>) {
+  // CHECK-NEXT: acc.loop combined(serial) {
+  // CHECK: acc.yield
+  // CHECK-NEXT: }
+  // CHECK-NEXT: acc.yield
+  // CHECK-NEXT: } loc
+  // CHECK-NEXT: acc.copyout accPtr(%[[COPYIN1]] : !cir.ptr<!s32i>) to varPtr(%[[GETMEMBER]] : !cir.ptr<!s32i>) {dataClause = #acc<data_clause acc_copy>, name = "localStruct.scalarMember"}
+
+#pragma acc kernels loop copy(localStruct.arrayMember)
+  for(int i = 0; i < 5; ++i);
+  // CHECK-NEXT: %[[GETARRAYMEMBER:.*]] = cir.get_member %[[LOCALSTRUCT]][1] {name = "arrayMember"} : !cir.ptr<!rec_StructTy> -> !cir.ptr<!cir.array<!s32i x 5>>
+  // CHECK-NEXT: %[[COPYIN1:.*]] = acc.copyin varPtr(%[[GETARRAYMEMBER]] : !cir.ptr<!cir.array<!s32i x 5>>) -> !cir.ptr<!cir.array<!s32i x 5>> {dataClause = #acc<data_clause acc_copy>, name = "localStruct.arrayMember"}
+  // CHECK-NEXT: acc.kernels combined(loop) dataOperands(%[[COPYIN1]] : !cir.ptr<!cir.array<!s32i x 5>>) {
+  // CHECK-NEXT: acc.loop combined(kernels) {
+  // CHECK: acc.yield
+  // CHECK-NEXT: }
+  // CHECK-NEXT: acc.terminator
+  // CHECK-NEXT: } loc
+  // CHECK-NEXT: acc.copyout accPtr(%[[COPYIN1]] : !cir.ptr<!cir.array<!s32i x 5>>) to varPtr(%[[GETARRAYMEMBER]] : !cir.ptr<!cir.array<!s32i x 5>>) {dataClause = #acc<data_clause acc_copy>, name = "localStruct.arrayMember"} loc
+
+#pragma acc parallel loop copy(localStruct.arrayMember[2])
+  for(int i = 0; i < 5; ++i);
+  // CHECK-NEXT: %[[TWO_CONST:.*]] = cir.const #cir.int<2> : !s32i
+  // CHECK-NEXT: %[[TWO_CAST:.*]] = builtin.unrealized_conversion_cast %[[TWO_CONST]] : !s32i to si32
+  // CHECK-NEXT: %[[ONE_CONST:.*]] = arith.constant 1 : i64
+  // CHECK-NEXT: %[[ZERO_CONST:.*]] = arith.constant 0 : i64
+  // CHECK-NEXT: %[[ONE_CONST2:.*]] = arith.constant 1 : i64
+  // CHECK-NEXT: %[[BOUNDS:.*]]  = acc.bounds lowerbound(%[[TWO_CAST]] : si32) extent(%[[ONE_CONST]] : i64) stride(%[[ONE_CONST2]] : i64) startIdx(%[[ZERO_CONST]] : i64) 
+  // CHECK-NEXT: %[[GETARRAYMEMBER:.*]] = cir.get_member %[[LOCALSTRUCT]][1] {name = "arrayMember"} : !cir.ptr<!rec_StructTy> -> !cir.ptr<!cir.array<!s32i x 5>>
+  // CHECK-NEXT: %[[COPYIN1:.*]] = acc.copyin varPtr(%[[GETARRAYMEMBER]] : !cir.ptr<!cir.array<!s32i x 5>>) bounds(%[[BOUNDS]]) -> !cir.ptr<!cir.array<!s32i x 5>> {dataClause = #acc<data_clause acc_copy>, name = "localStruct.arrayMember[2]"} 
----------------
erichkeane wrote:

None that I can think of... The OpenACC dialect defines these 'bounds', and the 'stride' of arrays/etc in C++ are always 1, based on type.

https://github.com/llvm/llvm-project/pull/142998