[flang-commits] [mlir] [flang] [flang][openacc] Lower loop directive to the new acc.loop op design (PR #65417)
Valentin Clement バレンタイン クレメン via flang-commits
flang-commits at lists.llvm.org
Wed Nov 29 14:11:22 PST 2023
https://github.com/clementval updated https://github.com/llvm/llvm-project/pull/65417
>From fd3e884f022d560453f7c1d30773429af2079c79 Mon Sep 17 00:00:00 2001
From: Valentin Clement <clementval at gmail.com>
Date: Tue, 29 Aug 2023 14:48:49 -0700
Subject: [PATCH 1/4] [mlir][openacc] Update acc.loop to be a proper loop like
operation
The initial design of the `acc.loop` was to be an operation that encapsulates
a loop like operation. This was an early design and we now want to change it
so the `acc.loop` operation becomes a real loop-like operation by implementing
the LoopLikeInterface.
Differential Revision: https://reviews.llvm.org/D159229
---
mlir/include/mlir/Dialect/OpenACC/OpenACC.h | 1 +
.../mlir/Dialect/OpenACC/OpenACCOps.td | 77 ++---
mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp | 78 ++++-
mlir/test/Dialect/OpenACC/canonicalize.mlir | 10 +-
mlir/test/Dialect/OpenACC/invalid.mlir | 88 +++---
mlir/test/Dialect/OpenACC/ops.mlir | 278 ++++++++----------
6 files changed, 300 insertions(+), 232 deletions(-)
diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACC.h b/mlir/include/mlir/Dialect/OpenACC/OpenACC.h
index 4dc94782c1c9b54..989b8ba3b78f334 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACC.h
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACC.h
@@ -25,6 +25,7 @@
#include "mlir/Dialect/OpenACC/OpenACCTypeInterfaces.h.inc"
#include "mlir/Dialect/OpenACCMPCommon/Interfaces/AtomicInterfaces.h"
#include "mlir/Interfaces/ControlFlowInterfaces.h"
+#include "mlir/Interfaces/LoopLikeInterface.h"
#include "mlir/Interfaces/SideEffectInterfaces.h"
#define GET_TYPEDEF_CLASSES
diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
index 391e77e0c4081a3..b52814c974ded12 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -14,6 +14,7 @@
#define OPENACC_OPS
include "mlir/Interfaces/ControlFlowInterfaces.td"
+include "mlir/Interfaces/LoopLikeInterface.td"
include "mlir/Interfaces/SideEffectInterfaces.td"
include "mlir/IR/BuiltinTypes.td"
include "mlir/IR/EnumAttr.td"
@@ -1161,47 +1162,53 @@ def OpenACC_HostDataOp : OpenACC_Op<"host_data", [AttrSizedOperandSegments]> {
//===----------------------------------------------------------------------===//
def OpenACC_LoopOp : OpenACC_Op<"loop",
- [AttrSizedOperandSegments, RecursiveMemoryEffects]> {
+ [AttrSizedOperandSegments, RecursiveMemoryEffects,
+ DeclareOpInterfaceMethods<LoopLikeOpInterface>]> {
let summary = "loop construct";
let description = [{
- The "acc.loop" operation represents the OpenACC loop construct.
+ The "acc.loop" operation represents the OpenACC loop construct. The lower
+ and upper bounds specify a half-open range: the range includes the lower
+ bound but does not include the upper bound. If the `inclusive` attribute is
+ set then the upper bound is included.
Example:
```mlir
- acc.loop gang vector {
- scf.for %arg3 = %c0 to %c10 step %c1 {
- scf.for %arg4 = %c0 to %c10 step %c1 {
- scf.for %arg5 = %c0 to %c10 step %c1 {
- // ... body
- }
- }
- }
+ acc.loop gang() vector() (%arg3 : index, %arg4 : index, %arg5 : index) =
+ (%c0, %c0, %c0 : index, index, index) to
+ (%c10, %c10, %c10 : index, index, index) step
+ (%c1, %c1, %c1 : index, index, index) {
+ // Loop body
acc.yield
} attributes { collapse = 3 }
```
}];
- let arguments = (ins OptionalAttr<I64Attr>:$collapse,
- Optional<IntOrIndex>:$gangNum,
- Optional<IntOrIndex>:$gangDim,
- Optional<IntOrIndex>:$gangStatic,
- Optional<IntOrIndex>:$workerNum,
- Optional<IntOrIndex>:$vectorLength,
- UnitAttr:$seq,
- UnitAttr:$independent,
- UnitAttr:$auto_,
- UnitAttr:$hasGang,
- UnitAttr:$hasWorker,
- UnitAttr:$hasVector,
- Variadic<IntOrIndex>:$tileOperands,
- Variadic<OpenACC_PointerLikeTypeInterface>:$cacheOperands,
- Variadic<OpenACC_PointerLikeTypeInterface>:$privateOperands,
- OptionalAttr<SymbolRefArrayAttr>:$privatizations,
- Variadic<AnyType>:$reductionOperands,
- OptionalAttr<SymbolRefArrayAttr>:$reductionRecipes
- );
+ let arguments = (ins
+ Variadic<IntOrIndex>:$lowerbound,
+ Variadic<IntOrIndex>:$upperbound,
+ Variadic<IntOrIndex>:$step,
+ OptionalAttr<DenseBoolArrayAttr>:$inclusiveUpperbound,
+ OptionalAttr<I64Attr>:$collapse,
+ Optional<IntOrIndex>:$gangNum,
+ Optional<IntOrIndex>:$gangDim,
+ Optional<IntOrIndex>:$gangStatic,
+ Optional<IntOrIndex>:$workerNum,
+ Optional<IntOrIndex>:$vectorLength,
+ UnitAttr:$seq,
+ UnitAttr:$independent,
+ UnitAttr:$auto_,
+ UnitAttr:$hasGang,
+ UnitAttr:$hasWorker,
+ UnitAttr:$hasVector,
+ Variadic<IntOrIndex>:$tileOperands,
+ Variadic<OpenACC_PointerLikeTypeInterface>:$cacheOperands,
+ Variadic<OpenACC_PointerLikeTypeInterface>:$privateOperands,
+ OptionalAttr<SymbolRefArrayAttr>:$privatizations,
+ Variadic<AnyType>:$reductionOperands,
+ OptionalAttr<SymbolRefArrayAttr>:$reductionRecipes
+ );
let results = (outs Variadic<AnyType>:$results);
@@ -1223,9 +1230,12 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
let hasCustomAssemblyFormat = 1;
let assemblyFormat = [{
oilist(
- `gang` `` custom<GangClause>($gangNum, type($gangNum), $gangDim, type($gangDim), $gangStatic, type($gangStatic), $hasGang)
- | `worker` `` custom<WorkerClause>($workerNum, type($workerNum), $hasWorker)
- | `vector` `` custom<VectorClause>($vectorLength, type($vectorLength), $hasVector)
+ `gang` `` custom<GangClause>($gangNum, type($gangNum), $gangDim,
+ type($gangDim), $gangStatic, type($gangStatic), $hasGang)
+ | `worker` ``
+ custom<WorkerClause>($workerNum, type($workerNum), $hasWorker)
+ | `vector` ``
+ custom<VectorClause>($vectorLength, type($vectorLength), $hasVector)
| `private` `(` custom<SymOperandList>(
$privateOperands, type($privateOperands), $privatizations)
`)`
@@ -1235,7 +1245,8 @@ def OpenACC_LoopOp : OpenACC_Op<"loop",
`)`
| `cache` `(` $cacheOperands `:` type($cacheOperands) `)`
)
- $region
+ custom<LoopControl>($region, $lowerbound, type($lowerbound), $upperbound,
+ type($upperbound), $step, type($step))
( `(` type($results)^ `)` )?
attr-dict-with-keyword
}];
diff --git a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
index 08e83cad482207d..86a4747f1e86c19 100644
--- a/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
+++ b/mlir/lib/Dialect/OpenACC/IR/OpenACC.cpp
@@ -745,6 +745,9 @@ static ParseResult parseGangClause(
// optional gang operands
if (succeeded(parser.parseOptionalLParen())) {
+ // Allow empty parenthesis.
+ if (succeeded(parser.parseOptionalRParen()))
+ return success();
while (true) {
bool newValue = false;
bool needValue = false;
@@ -791,8 +794,8 @@ static ParseResult parseGangClause(
void printGangClause(OpAsmPrinter &p, Operation *op, Value gangNum,
Type gangNumType, Value gangDim, Type gangDimType,
Value gangStatic, Type gangStaticType, UnitAttr hasGang) {
+ p << "(";
if (gangNum || gangStatic || gangDim) {
- p << "(";
if (gangNum) {
p << LoopOp::getGangNumKeyword() << "=" << gangNum << " : "
<< gangNumType;
@@ -808,8 +811,8 @@ void printGangClause(OpAsmPrinter &p, Operation *op, Value gangNum,
if (gangStatic)
p << LoopOp::getGangStaticKeyword() << "=" << gangStatic << " : "
<< gangStaticType;
- p << ")";
}
+ p << ")";
}
static ParseResult
@@ -818,6 +821,9 @@ parseWorkerClause(OpAsmParser &parser,
Type &workerNumType, UnitAttr &hasWorker) {
hasWorker = UnitAttr::get(parser.getBuilder().getContext());
if (succeeded(parser.parseOptionalLParen())) {
+ // Allow empty parenthesis.
+ if (succeeded(parser.parseOptionalRParen()))
+ return success();
workerNum = OpAsmParser::UnresolvedOperand{};
if (parser.parseOperand(*workerNum) ||
parser.parseColonType(workerNumType) || parser.parseRParen())
@@ -828,8 +834,10 @@ parseWorkerClause(OpAsmParser &parser,
void printWorkerClause(OpAsmPrinter &p, Operation *op, Value workerNum,
Type workerNumType, UnitAttr hasWorker) {
+ p << "(";
if (workerNum)
- p << "(" << workerNum << " : " << workerNumType << ")";
+ p << workerNum << " : " << workerNumType;
+ p << ")";
}
static ParseResult
@@ -838,6 +846,9 @@ parseVectorClause(OpAsmParser &parser,
Type &vectorLengthType, UnitAttr &hasVector) {
hasVector = UnitAttr::get(parser.getBuilder().getContext());
if (succeeded(parser.parseOptionalLParen())) {
+ // Allow empty parenthesis.
+ if (succeeded(parser.parseOptionalRParen()))
+ return success();
vectorLength = OpAsmParser::UnresolvedOperand{};
if (parser.parseOperand(*vectorLength) ||
parser.parseColonType(vectorLengthType) || parser.parseRParen())
@@ -848,11 +859,18 @@ parseVectorClause(OpAsmParser &parser,
void printVectorClause(OpAsmPrinter &p, Operation *op, Value vectorLength,
Type vectorLengthType, UnitAttr hasVector) {
+ p << "(";
if (vectorLength)
- p << "(" << vectorLength << " : " << vectorLengthType << ")";
+ p << vectorLength << " : " << vectorLengthType;
+ p << ")";
}
LogicalResult acc::LoopOp::verify() {
+ if (!getUpperbound().empty() &&
+ (getUpperbound().size() != getInclusiveUpperbound()->size()))
+ return emitError() << "inclusiveUpperbound size is expected to be the same"
+ << " as upperbound size";
+
// auto, independent and seq attribute are mutually exclusive.
if ((getAuto_() && (getIndependent() || getSeq())) ||
(getIndependent() && getSeq())) {
@@ -898,6 +916,58 @@ Value LoopOp::getDataOperand(unsigned i) {
return getOperand(numOptional + i);
}
+llvm::SmallVector<mlir::Region *> acc::LoopOp::getLoopRegions() {
+ return {&getRegion()};
+}
+
+/// loop-control ::= `(` ssa-id-and-type-list `)` `=` `(` ssa-id-and-type-list
+/// `)` `to` `(` ssa-id-and-type-list `)` `step` `(` ssa-id-and-type-list `)`
+ParseResult
+parseLoopControl(OpAsmParser &parser, Region ®ion,
+ SmallVectorImpl<OpAsmParser::UnresolvedOperand> &lowerbound,
+ SmallVectorImpl<Type> &lowerboundType,
+ SmallVectorImpl<OpAsmParser::UnresolvedOperand> &upperbound,
+ SmallVectorImpl<Type> &upperboundType,
+ SmallVectorImpl<OpAsmParser::UnresolvedOperand> &step,
+ SmallVectorImpl<Type> &stepType) {
+
+ SmallVector<OpAsmParser::Argument> inductionVars;
+ if (succeeded(parser.parseOptionalLParen())) {
+ if (parser.parseArgumentList(inductionVars, OpAsmParser::Delimiter::None,
+ /*allowType=*/true) ||
+ parser.parseRParen() || parser.parseEqual() || parser.parseLParen() ||
+ parser.parseOperandList(lowerbound, inductionVars.size(),
+ OpAsmParser::Delimiter::None) ||
+ parser.parseColonTypeList(lowerboundType) || parser.parseRParen() ||
+ parser.parseKeyword("to") || parser.parseLParen() ||
+ parser.parseOperandList(upperbound, inductionVars.size(),
+ OpAsmParser::Delimiter::None) ||
+ parser.parseColonTypeList(upperboundType) || parser.parseRParen() ||
+ parser.parseKeyword("step") || parser.parseLParen() ||
+ parser.parseOperandList(step, inductionVars.size(),
+ OpAsmParser::Delimiter::None) ||
+ parser.parseColonTypeList(stepType) || parser.parseRParen())
+ return failure();
+ }
+ return parser.parseRegion(region, inductionVars);
+}
+
+void printLoopControl(OpAsmPrinter &p, Operation *op, Region ®ion,
+ ValueRange lowerbound, TypeRange lowerboundType,
+ ValueRange upperbound, TypeRange upperboundType,
+ ValueRange steps, TypeRange stepType) {
+ ValueRange regionArgs = region.front().getArguments();
+ if (!regionArgs.empty()) {
+ p << "(";
+ llvm::interleaveComma(regionArgs, p,
+ [&p](Value v) { p << v << " : " << v.getType(); });
+ p << ") = (" << lowerbound << " : " << lowerboundType << ") to ("
+ << upperbound << " : " << upperboundType << ") "
+ << " step (" << steps << " : " << stepType << ") ";
+ }
+ p.printRegion(region, /*printEntryBlockArgs=*/false);
+}
+
//===----------------------------------------------------------------------===//
// DataOp
//===----------------------------------------------------------------------===//
diff --git a/mlir/test/Dialect/OpenACC/canonicalize.mlir b/mlir/test/Dialect/OpenACC/canonicalize.mlir
index 6173ab6699c6c58..4522ffb252a62e3 100644
--- a/mlir/test/Dialect/OpenACC/canonicalize.mlir
+++ b/mlir/test/Dialect/OpenACC/canonicalize.mlir
@@ -110,14 +110,16 @@ func.func @testupdateop(%a: memref<f32>, %ifCond: i1) -> () {
func.func @testhostdataop(%a: memref<f32>, %ifCond: i1) -> () {
%0 = acc.use_device varPtr(%a : memref<f32>) -> memref<f32>
+ %1 = arith.constant 1 : i32
+ %2 = arith.constant 10 : i32
%false = arith.constant false
acc.host_data dataOperands(%0 : memref<f32>) if(%false) {
- acc.loop {
+ acc.loop (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
acc.yield
- }
- acc.loop {
+ } attributes { inclusiveUpperbound = array<i1: true> }
+ acc.loop (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
acc.yield
- }
+ } attributes { inclusiveUpperbound = array<i1: true> }
acc.terminator
}
return
diff --git a/mlir/test/Dialect/OpenACC/invalid.mlir b/mlir/test/Dialect/OpenACC/invalid.mlir
index b9ac68d0592c872..a6b99a54b4c142f 100644
--- a/mlir/test/Dialect/OpenACC/invalid.mlir
+++ b/mlir/test/Dialect/OpenACC/invalid.mlir
@@ -1,71 +1,81 @@
// RUN: mlir-opt -split-input-file -verify-diagnostics %s
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{gang, worker or vector cannot appear with the seq attr}}
-acc.loop gang {
+acc.loop gang() (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
-} attributes {seq}
+} attributes {seq, inclusiveUpperbound = array<i1: true>}
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{gang, worker or vector cannot appear with the seq attr}}
-acc.loop worker {
+acc.loop worker() (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
-} attributes {seq}
+} attributes {seq, inclusiveUpperbound = array<i1: true>}
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{gang, worker or vector cannot appear with the seq attr}}
-acc.loop vector {
+acc.loop vector() (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
-} attributes {seq}
+} attributes {seq, inclusiveUpperbound = array<i1: true>}
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{gang, worker or vector cannot appear with the seq attr}}
-acc.loop gang worker {
+acc.loop gang() worker() (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
-} attributes {seq}
+} attributes {seq, inclusiveUpperbound = array<i1: true>}
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{gang, worker or vector cannot appear with the seq attr}}
-acc.loop gang vector {
+acc.loop gang() vector() (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
-} attributes {seq}
+} attributes {seq, inclusiveUpperbound = array<i1: true>}
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{gang, worker or vector cannot appear with the seq attr}}
-acc.loop worker vector {
+acc.loop worker() vector() (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
-} attributes {seq}
+} attributes {seq, inclusiveUpperbound = array<i1: true>}
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{gang, worker or vector cannot appear with the seq attr}}
-acc.loop gang worker vector {
+acc.loop gang() worker() vector() (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
-} attributes {seq}
-
-// -----
-
-// expected-error at +1 {{expected non-empty body.}}
-acc.loop {
-}
+} attributes {seq, inclusiveUpperbound = array<i1: true>}
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{only one of "auto", "independent", "seq" can be present at the same time}}
-acc.loop {
+acc.loop (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
acc.yield
-} attributes {auto_, seq}
+} attributes {auto_, seq, inclusiveUpperbound = array<i1: true>}
// -----
@@ -133,11 +143,13 @@ acc.parallel {
// -----
-acc.loop {
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
+acc.loop (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32){
// expected-error at +1 {{'acc.init' op cannot be nested in a compute operation}}
acc.init
acc.yield
-}
+} attributes {inclusiveUpperbound = array<i1: true>}
// -----
@@ -149,21 +161,25 @@ acc.parallel {
// -----
-acc.loop {
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
+acc.loop (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
// expected-error at +1 {{'acc.shutdown' op cannot be nested in a compute operation}}
acc.shutdown
acc.yield
-}
+} attributes {inclusiveUpperbound = array<i1: true>}
// -----
-acc.loop {
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
+acc.loop (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() ({
// expected-error at +1 {{'acc.shutdown' op cannot be nested in a compute operation}}
acc.shutdown
}) : () -> ()
acc.yield
-}
+} attributes {inclusiveUpperbound = array<i1: true>}
// -----
@@ -367,8 +383,10 @@ acc.firstprivate.recipe @privatization_i32 : i32 init {
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{expected ')'}}
-acc.loop gang(static=%i64Value: i64, num=%i64Value: i64 {
+acc.loop gang(static=%i64Value: i64, num=%i64Value: i64 (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
}
@@ -436,8 +454,10 @@ acc.reduction.recipe @reduction_i64 : i64 reduction_operator<add> init {
// -----
+%1 = arith.constant 1 : i32
+%2 = arith.constant 10 : i32
// expected-error at +1 {{new value expected after comma}}
-acc.loop gang(static=%i64Value: i64, ) {
+acc.loop gang(static=%i64Value: i64, ) (%iv : i32) = (%1 : i32) to (%2 : i32) step (%1 : i32) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
}
@@ -453,14 +473,6 @@ func.func @fct1(%0 : !llvm.ptr) -> () {
// -----
-// expected-error at +1 {{expect at least one of num, dim or static values}}
-acc.loop gang() {
- "test.openacc_dummy_op"() : () -> ()
- acc.yield
-}
-
-// -----
-
%i64value = arith.constant 1 : i64
// expected-error at +1 {{num_gangs expects a maximum of 3 values}}
acc.parallel num_gangs(%i64value, %i64value, %i64value, %i64value : i64, i64, i64, i64) {
diff --git a/mlir/test/Dialect/OpenACC/ops.mlir b/mlir/test/Dialect/OpenACC/ops.mlir
index c06e5a596923f94..69d11f8bb2cce13 100644
--- a/mlir/test/Dialect/OpenACC/ops.mlir
+++ b/mlir/test/Dialect/OpenACC/ops.mlir
@@ -11,52 +11,40 @@ func.func @compute1(%A: memref<10x10xf32>, %B: memref<10x10xf32>, %C: memref<10x
%async = arith.constant 1 : i64
acc.parallel async(%async: i64) {
- acc.loop gang vector {
- scf.for %arg3 = %c0 to %c10 step %c1 {
- scf.for %arg4 = %c0 to %c10 step %c1 {
- scf.for %arg5 = %c0 to %c10 step %c1 {
- %a = memref.load %A[%arg3, %arg5] : memref<10x10xf32>
- %b = memref.load %B[%arg5, %arg4] : memref<10x10xf32>
- %cij = memref.load %C[%arg3, %arg4] : memref<10x10xf32>
- %p = arith.mulf %a, %b : f32
- %co = arith.addf %cij, %p : f32
- memref.store %co, %C[%arg3, %arg4] : memref<10x10xf32>
- }
- }
- }
+ acc.loop gang() vector() (%arg3 : index, %arg4 : index, %arg5 : index) = (%c0, %c0, %c0 : index, index, index) to (%c10, %c10, %c10 : index, index, index) step (%c1, %c1, %c1 : index, index, index) {
+ %a = memref.load %A[%arg3, %arg5] : memref<10x10xf32>
+ %b = memref.load %B[%arg5, %arg4] : memref<10x10xf32>
+ %cij = memref.load %C[%arg3, %arg4] : memref<10x10xf32>
+ %p = arith.mulf %a, %b : f32
+ %co = arith.addf %cij, %p : f32
+ memref.store %co, %C[%arg3, %arg4] : memref<10x10xf32>
acc.yield
- } attributes { collapse = 3 }
+ } attributes { collapse = 3, inclusiveUpperbound = array<i1: true, true, true> }
acc.yield
}
return %C : memref<10x10xf32>
}
-// CHECK-LABEL: func @compute1(
+// CHECK-LABEL: func @compute1
// CHECK-NEXT: %{{.*}} = arith.constant 0 : index
// CHECK-NEXT: %{{.*}} = arith.constant 10 : index
// CHECK-NEXT: %{{.*}} = arith.constant 1 : index
// CHECK-NEXT: [[ASYNC:%.*]] = arith.constant 1 : i64
// CHECK-NEXT: acc.parallel async([[ASYNC]] : i64) {
-// CHECK-NEXT: acc.loop gang vector {
-// CHECK-NEXT: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} {
-// CHECK-NEXT: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} {
-// CHECK-NEXT: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} {
-// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
-// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
-// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
-// CHECK-NEXT: %{{.*}} = arith.mulf %{{.*}}, %{{.*}} : f32
-// CHECK-NEXT: %{{.*}} = arith.addf %{{.*}}, %{{.*}} : f32
-// CHECK-NEXT: memref.store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
-// CHECK-NEXT: }
-// CHECK-NEXT: }
-// CHECK-NEXT: }
+// CHECK-NEXT: acc.loop gang() vector() (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}}) {
+// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
+// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
+// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
+// CHECK-NEXT: %{{.*}} = arith.mulf %{{.*}}, %{{.*}} : f32
+// CHECK-NEXT: %{{.*}} = arith.addf %{{.*}}, %{{.*}} : f32
+// CHECK-NEXT: memref.store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: } attributes {collapse = 3 : i64}
+// CHECK-NEXT: } attributes {collapse = 3 : i64, inclusiveUpperbound = array<i1: true, true, true>}
// CHECK-NEXT: acc.yield
// CHECK-NEXT: }
// CHECK-NEXT: return %{{.*}} : memref<10x10xf32>
-// CHECK-NEXT: }
+
// -----
@@ -66,21 +54,19 @@ func.func @compute2(%A: memref<10x10xf32>, %B: memref<10x10xf32>, %C: memref<10x
%c1 = arith.constant 1 : index
acc.parallel {
- acc.loop {
- scf.for %arg3 = %c0 to %c10 step %c1 {
- scf.for %arg4 = %c0 to %c10 step %c1 {
- scf.for %arg5 = %c0 to %c10 step %c1 {
- %a = memref.load %A[%arg3, %arg5] : memref<10x10xf32>
- %b = memref.load %B[%arg5, %arg4] : memref<10x10xf32>
- %cij = memref.load %C[%arg3, %arg4] : memref<10x10xf32>
- %p = arith.mulf %a, %b : f32
- %co = arith.addf %cij, %p : f32
- memref.store %co, %C[%arg3, %arg4] : memref<10x10xf32>
- }
+ acc.loop (%arg3 : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
+ scf.for %arg4 = %c0 to %c10 step %c1 {
+ scf.for %arg5 = %c0 to %c10 step %c1 {
+ %a = memref.load %A[%arg3, %arg5] : memref<10x10xf32>
+ %b = memref.load %B[%arg5, %arg4] : memref<10x10xf32>
+ %cij = memref.load %C[%arg3, %arg4] : memref<10x10xf32>
+ %p = arith.mulf %a, %b : f32
+ %co = arith.addf %cij, %p : f32
+ memref.store %co, %C[%arg3, %arg4] : memref<10x10xf32>
}
}
acc.yield
- } attributes {seq}
+ } attributes { seq, inclusiveUpperbound = array<i1: true> }
acc.yield
}
@@ -92,8 +78,7 @@ func.func @compute2(%A: memref<10x10xf32>, %B: memref<10x10xf32>, %C: memref<10x
// CHECK-NEXT: %{{.*}} = arith.constant 10 : index
// CHECK-NEXT: %{{.*}} = arith.constant 1 : index
// CHECK-NEXT: acc.parallel {
-// CHECK-NEXT: acc.loop {
-// CHECK-NEXT: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} {
+// CHECK-NEXT: acc.loop (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
// CHECK-NEXT: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} {
// CHECK-NEXT: scf.for %{{.*}} = %{{.*}} to %{{.*}} step %{{.*}} {
// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
@@ -104,9 +89,8 @@ func.func @compute2(%A: memref<10x10xf32>, %B: memref<10x10xf32>, %C: memref<10x
// CHECK-NEXT: memref.store %{{.*}}, %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
// CHECK-NEXT: }
// CHECK-NEXT: }
-// CHECK-NEXT: }
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: } attributes {seq}
+// CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, seq}
// CHECK-NEXT: acc.yield
// CHECK-NEXT: }
// CHECK-NEXT: return %{{.*}} : memref<10x10xf32>
@@ -138,32 +122,25 @@ func.func @compute3(%a: memref<10x10xf32>, %b: memref<10x10xf32>, %c: memref<10x
acc.data dataOperands(%pa, %pb, %pc, %pd: memref<10x10xf32>, memref<10x10xf32>, memref<10xf32>, memref<10xf32>) {
%private = acc.private varPtr(%c : memref<10xf32>) -> memref<10xf32>
acc.parallel num_gangs(%numGangs: i64) num_workers(%numWorkers: i64) private(@privatization_memref_10_f32 -> %private : memref<10xf32>) {
- acc.loop gang {
- scf.for %x = %lb to %c10 step %st {
- acc.loop worker {
- scf.for %y = %lb to %c10 step %st {
- %axy = memref.load %a[%x, %y] : memref<10x10xf32>
- %bxy = memref.load %b[%x, %y] : memref<10x10xf32>
- %tmp = arith.addf %axy, %bxy : f32
- memref.store %tmp, %c[%y] : memref<10xf32>
- }
- acc.yield
- }
-
- acc.loop {
- // for i = 0 to 10 step 1
- // d[x] += c[i]
- scf.for %i = %lb to %c10 step %st {
- %ci = memref.load %c[%i] : memref<10xf32>
- %dx = memref.load %d[%x] : memref<10xf32>
- %z = arith.addf %ci, %dx : f32
- memref.store %z, %d[%x] : memref<10xf32>
- }
- acc.yield
- } attributes {seq}
- }
+ acc.loop gang() (%x : index) = (%lb : index) to (%c10 : index) step (%st : index) {
+ acc.loop worker() (%y : index) = (%lb : index) to (%c10 : index) step (%st : index) {
+ %axy = memref.load %a[%x, %y] : memref<10x10xf32>
+ %bxy = memref.load %b[%x, %y] : memref<10x10xf32>
+ %tmp = arith.addf %axy, %bxy : f32
+ memref.store %tmp, %c[%y] : memref<10xf32>
+ acc.yield
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop (%i : index) = (%lb : index) to (%c10 : index) step (%st : index) {
+ // for i = 0 to 10 step 1
+ // d[x] += c[i]
+ %ci = memref.load %c[%i] : memref<10xf32>
+ %dx = memref.load %d[%x] : memref<10xf32>
+ %z = arith.addf %ci, %dx : f32
+ memref.store %z, %d[%x] : memref<10xf32>
+ acc.yield
+ } attributes {seq, inclusiveUpperbound = array<i1: true>}
acc.yield
- }
+ } attributes {inclusiveUpperbound = array<i1: true>}
acc.yield
}
acc.terminator
@@ -181,27 +158,21 @@ func.func @compute3(%a: memref<10x10xf32>, %b: memref<10x10xf32>, %c: memref<10x
// CHECK: acc.data dataOperands(%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}} : memref<10x10xf32>, memref<10x10xf32>, memref<10xf32>, memref<10xf32>) {
// CHECK-NEXT: %[[P_ARG2:.*]] = acc.private varPtr([[ARG2]] : memref<10xf32>) -> memref<10xf32>
// CHECK-NEXT: acc.parallel num_gangs([[NUMGANG]] : i64) num_workers([[NUMWORKERS]] : i64) private(@privatization_memref_10_f32 -> %[[P_ARG2]] : memref<10xf32>) {
-// CHECK-NEXT: acc.loop gang {
-// CHECK-NEXT: scf.for %{{.*}} = [[C0]] to [[C10]] step [[C1]] {
-// CHECK-NEXT: acc.loop worker {
-// CHECK-NEXT: scf.for %{{.*}} = [[C0]] to [[C10]] step [[C1]] {
+// CHECK-NEXT: acc.loop gang() (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
+// CHECK-NEXT: acc.loop worker() (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}, %{{.*}}] : memref<10x10xf32>
// CHECK-NEXT: %{{.*}} = arith.addf %{{.*}}, %{{.*}} : f32
// CHECK-NEXT: memref.store %{{.*}}, %{{.*}}[%{{.*}}] : memref<10xf32>
-// CHECK-NEXT: }
// CHECK-NEXT: acc.yield
// CHECK-NEXT: }
-// CHECK-NEXT: acc.loop {
-// CHECK-NEXT: scf.for %{{.*}} = [[C0]] to [[C10]] step [[C1]] {
+// CHECK-NEXT: acc.loop (%{{.*}}) = (%{{.*}}) to (%{{.*}}) step (%{{.*}})
// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}] : memref<10xf32>
// CHECK-NEXT: %{{.*}} = memref.load %{{.*}}[%{{.*}}] : memref<10xf32>
// CHECK-NEXT: %{{.*}} = arith.addf %{{.*}}, %{{.*}} : f32
// CHECK-NEXT: memref.store %{{.*}}, %{{.*}}[%{{.*}}] : memref<10xf32>
-// CHECK-NEXT: }
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: } attributes {seq}
-// CHECK-NEXT: }
+// CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, seq}
// CHECK-NEXT: acc.yield
// CHECK-NEXT: }
// CHECK-NEXT: acc.yield
@@ -217,68 +188,75 @@ func.func @testloopop(%a : memref<10xf32>) -> () {
%i64Value = arith.constant 1 : i64
%i32Value = arith.constant 128 : i32
%idxValue = arith.constant 8 : index
+ %c0 = arith.constant 0 : index
+ %c10 = arith.constant 10 : index
+ %c1 = arith.constant 1 : index
- acc.loop gang worker vector {
+ acc.loop gang() worker() vector() (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop gang(num=%i64Value: i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop gang(num=%i64Value: i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop gang(static=%i64Value: i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop gang(static=%i64Value: i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop worker(%i64Value: i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop worker(%i64Value: i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop worker(%i32Value: i32) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop worker(%i32Value: i32) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop worker(%idxValue: index) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop worker(%idxValue: index) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop vector(%i64Value: i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop vector(%i64Value: i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop vector(%i32Value: i32) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop vector(%i32Value: i32) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop vector(%idxValue: index) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop vector(%idxValue: index) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop gang(num=%i64Value: i64) worker vector {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop gang(num=%i64Value: i64) worker() vector() (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop gang(num=%i64Value: i64, static=%i64Value: i64) worker(%i64Value: i64) vector(%i64Value: i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop gang(num=%i64Value: i64, static=%i64Value: i64) worker(%i64Value: i64) vector(%i64Value: i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop gang(num=%i32Value: i32, static=%idxValue: index) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop gang(num=%i32Value: i32, static=%idxValue: index) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop tile(%i64Value, %i64Value : i64, i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop tile(%i64Value, %i64Value : i64, i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop tile(%i32Value, %i32Value : i32, i32) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop tile(%i32Value, %i32Value : i32, i32) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop gang(static=%i64Value: i64, num=%i64Value: i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop gang(static=%i64Value: i64, num=%i64Value: i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
"test.openacc_dummy_op"() : () -> ()
acc.yield
- }
- acc.loop gang(dim=%i64Value : i64, static=%i64Value: i64) {
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop gang(dim=%i64Value : i64, static=%i64Value: i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
+ "test.openacc_dummy_op"() : () -> ()
+ acc.yield
+ } attributes {inclusiveUpperbound = array<i1: true>}
+ acc.loop {
"test.openacc_dummy_op"() : () -> ()
acc.yield
}
@@ -293,67 +271,55 @@ func.func @testloopop(%a : memref<10xf32>) -> () {
// CHECK: [[I64VALUE:%.*]] = arith.constant 1 : i64
// CHECK-NEXT: [[I32VALUE:%.*]] = arith.constant 128 : i32
// CHECK-NEXT: [[IDXVALUE:%.*]] = arith.constant 8 : index
-// CHECK: acc.loop gang worker vector {
+// CHECK: acc.loop gang() worker() vector()
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop gang(num=[[I64VALUE]] : i64) {
+// CHECK: acc.loop gang(num=[[I64VALUE]] : i64)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop gang(static=[[I64VALUE]] : i64) {
+// CHECK: acc.loop gang(static=[[I64VALUE]] : i64)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop worker([[I64VALUE]] : i64) {
+// CHECK: acc.loop worker([[I64VALUE]] : i64)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop worker([[I32VALUE]] : i32) {
+// CHECK: acc.loop worker([[I32VALUE]] : i32)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop worker([[IDXVALUE]] : index) {
+// CHECK: acc.loop worker([[IDXVALUE]] : index)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop vector([[I64VALUE]] : i64) {
+// CHECK: acc.loop vector([[I64VALUE]] : i64)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop vector([[I32VALUE]] : i32) {
+// CHECK: acc.loop vector([[I32VALUE]] : i32)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop vector([[IDXVALUE]] : index) {
+// CHECK: acc.loop vector([[IDXVALUE]] : index)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop gang(num=[[I64VALUE]] : i64) worker vector {
+// CHECK: acc.loop gang(num=[[I64VALUE]] : i64) worker() vector()
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop gang(num=[[I64VALUE]] : i64, static=[[I64VALUE]] : i64) worker([[I64VALUE]] : i64) vector([[I64VALUE]] : i64) {
+// CHECK: acc.loop gang(num=[[I64VALUE]] : i64, static=[[I64VALUE]] : i64) worker([[I64VALUE]] : i64) vector([[I64VALUE]] : i64)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop gang(num=[[I32VALUE]] : i32, static=[[IDXVALUE]] : index) {
+// CHECK: acc.loop gang(num=[[I32VALUE]] : i32, static=[[IDXVALUE]] : index)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop tile([[I64VALUE]], [[I64VALUE]] : i64, i64) {
+// CHECK: acc.loop tile([[I64VALUE]], [[I64VALUE]] : i64, i64)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop tile([[I32VALUE]], [[I32VALUE]] : i32, i32) {
+// CHECK: acc.loop tile([[I32VALUE]], [[I32VALUE]] : i32, i32)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop gang(num=[[I64VALUE]] : i64, static=[[I64VALUE]] : i64) {
+// CHECK: acc.loop gang(num=[[I64VALUE]] : i64, static=[[I64VALUE]] : i64)
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
-// CHECK-NEXT: }
-// CHECK: acc.loop gang(dim=[[I64VALUE]] : i64, static=[[I64VALUE]] : i64) {
+// CHECK: acc.loop gang(dim=[[I64VALUE]] : i64, static=[[I64VALUE]] : i64)
+// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
+// CHECK-NEXT: acc.yield
+// CHECK: acc.loop {
// CHECK-NEXT: "test.openacc_dummy_op"() : () -> ()
// CHECK-NEXT: acc.yield
// CHECK-NEXT: }
@@ -366,22 +332,25 @@ func.func @testloopop(%a : memref<10xf32>) -> () {
// -----
func.func @acc_loop_multiple_block() {
+ %c0 = arith.constant 0 : index
+ %c10 = arith.constant 10 : index
+ %c1 = arith.constant 1 : index
acc.parallel {
- acc.loop {
- %c1 = arith.constant 1 : index
- cf.br ^bb1(%c1 : index)
+ acc.loop (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
+ %c1_1 = arith.constant 1 : index
+ cf.br ^bb1(%c1_1 : index)
^bb1(%9: index):
- %c0 = arith.constant 0 : index
- %12 = arith.cmpi sgt, %9, %c0 : index
+ %c0_1 = arith.constant 0 : index
+ %12 = arith.cmpi sgt, %9, %c0_1 : index
cf.cond_br %12, ^bb2, ^bb3
^bb2:
%c1_0 = arith.constant 1 : index
- %c10 = arith.constant 10 : index
- %22 = arith.subi %c10, %c1_0 : index
+ %c10_1 = arith.constant 10 : index
+ %22 = arith.subi %c10_1, %c1_0 : index
cf.br ^bb1(%22 : index)
^bb3:
acc.yield
- }
+ } attributes {inclusiveUpperbound = array<i1: true>}
acc.yield
}
return
@@ -1509,10 +1478,13 @@ acc.reduction.recipe @reduction_add_i64 : i64 reduction_operator<add> init {
// CHECK: }
func.func @acc_reduc_test(%a : i64) -> () {
+ %c0 = arith.constant 0 : index
+ %c10 = arith.constant 10 : index
+ %c1 = arith.constant 1 : index
acc.parallel reduction(@reduction_add_i64 -> %a : i64) {
- acc.loop reduction(@reduction_add_i64 -> %a : i64) {
+ acc.loop reduction(@reduction_add_i64 -> %a : i64) (%iv : index) = (%c0 : index) to (%c10 : index) step (%c1 : index) {
acc.yield
- }
+ } attributes { inclusiveUpperbound = array<i1: true> }
acc.yield
}
return
>From bc123ca98eeaeada4225327d524d6a7ddb700133 Mon Sep 17 00:00:00 2001
From: Valentin Clement <clementval at gmail.com>
Date: Fri, 1 Sep 2023 11:10:59 -0700
Subject: [PATCH 2/4] [flang][openacc] Lower loop directive to the new acc.loop
op design
acc.loop was redesigned in https://reviews.llvm.org/D159229. This patch
updates the lowering to match the new op.
DO CONCURRENT construct will be added in a follow up patch.
---
flang/include/flang/Lower/OpenACC.h | 5 +
flang/lib/Lower/Bridge.cpp | 19 +-
flang/lib/Lower/OpenACC.cpp | 147 ++++++++++--
flang/test/Lower/OpenACC/acc-kernels-loop.f90 | 154 +++++--------
flang/test/Lower/OpenACC/acc-loop.f90 | 213 ++++++++----------
.../test/Lower/OpenACC/acc-parallel-loop.f90 | 180 ++++++---------
flang/test/Lower/OpenACC/acc-private.f90 | 14 +-
flang/test/Lower/OpenACC/acc-reduction.f90 | 45 ++--
flang/test/Lower/OpenACC/acc-serial-loop.f90 | 138 ++++--------
9 files changed, 437 insertions(+), 478 deletions(-)
diff --git a/flang/include/flang/Lower/OpenACC.h b/flang/include/flang/Lower/OpenACC.h
index 409956f0ecb309f..da601c456a1fb2e 100644
--- a/flang/include/flang/Lower/OpenACC.h
+++ b/flang/include/flang/Lower/OpenACC.h
@@ -35,6 +35,7 @@ class FirOpBuilder;
namespace Fortran {
namespace parser {
+struct AccClauseList;
struct OpenACCConstruct;
struct OpenACCDeclarativeConstruct;
struct OpenACCRoutineConstruct;
@@ -64,6 +65,8 @@ static constexpr llvm::StringRef declarePreDeallocSuffix =
static constexpr llvm::StringRef declarePostDeallocSuffix =
"_acc_declare_update_desc_post_dealloc";
+static constexpr llvm::StringRef privatizationRecipePrefix = "privatization";
+
void genOpenACCConstruct(AbstractConverter &,
Fortran::semantics::SemanticsContext &,
pft::Evaluation &, const parser::OpenACCConstruct &);
@@ -112,6 +115,8 @@ void attachDeclarePostDeallocAction(AbstractConverter &, fir::FirOpBuilder &,
void genOpenACCTerminator(fir::FirOpBuilder &, mlir::Operation *,
mlir::Location);
+int64_t getCollapseValue(const Fortran::parser::AccClauseList &);
+
} // namespace lower
} // namespace Fortran
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index 0090c2f1212164a..fcb5f275c8ffa02 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -2383,7 +2383,24 @@ class FirConverter : public Fortran::lower::AbstractConverter {
mlir::OpBuilder::InsertPoint insertPt = builder->saveInsertionPoint();
localSymbols.pushScope();
genOpenACCConstruct(*this, bridge.getSemanticsContext(), getEval(), acc);
- for (Fortran::lower::pft::Evaluation &e : getEval().getNestedEvaluations())
+
+ const Fortran::parser::OpenACCLoopConstruct *accLoop =
+ std::get_if<Fortran::parser::OpenACCLoopConstruct>(&acc.u);
+
+ Fortran::lower::pft::Evaluation *curEval = &getEval();
+
+ if (accLoop) {
+ const Fortran::parser::AccBeginLoopDirective &beginLoopDir =
+ std::get<Fortran::parser::AccBeginLoopDirective>(accLoop->t);
+ const Fortran::parser::AccClauseList &clauseList =
+ std::get<Fortran::parser::AccClauseList>(beginLoopDir.t);
+ int64_t collapseValue = Fortran::lower::getCollapseValue(clauseList);
+ curEval = &curEval->getFirstNestedEvaluation();
+ for (int64_t i = 1; i < collapseValue; i++)
+ curEval = &*std::next(curEval->getNestedEvaluations().begin());
+ }
+
+ for (Fortran::lower::pft::Evaluation &e : curEval->getNestedEvaluations())
genFIR(e);
localSymbols.popScope();
builder->restoreInsertionPoint(insertPt);
diff --git a/flang/lib/Lower/OpenACC.cpp b/flang/lib/Lower/OpenACC.cpp
index 8c6c22210cf0894..e3451688f9af51c 100644
--- a/flang/lib/Lower/OpenACC.cpp
+++ b/flang/lib/Lower/OpenACC.cpp
@@ -754,7 +754,8 @@ genPrivatizations(const Fortran::parser::AccObjectList &objectList,
mlir::Type retTy = getTypeFromBounds(bounds, baseAddr.getType());
if constexpr (std::is_same_v<RecipeOp, mlir::acc::PrivateRecipeOp>) {
std::string recipeName =
- fir::getTypeAsString(retTy, converter.getKindMap(), "privatization");
+ fir::getTypeAsString(retTy, converter.getKindMap(),
+ Fortran::lower::privatizationRecipePrefix);
recipe = Fortran::lower::createOrGetPrivateRecipe(builder, recipeName,
operandLocation, retTy);
auto op = createDataEntryOp<mlir::acc::PrivateOp>(
@@ -1381,10 +1382,12 @@ static Op createRegionOp(fir::FirOpBuilder &builder, mlir::Location loc,
Fortran::lower::pft::Evaluation &eval,
const llvm::SmallVectorImpl<mlir::Value> &operands,
const llvm::SmallVectorImpl<int32_t> &operandSegments,
- bool outerCombined = false) {
- llvm::ArrayRef<mlir::Type> argTy;
- Op op = builder.create<Op>(loc, argTy, operands);
- builder.createBlock(&op.getRegion());
+ bool outerCombined = false,
+ mlir::TypeRange argsTy = {},
+ llvm::SmallVector<mlir::Location> locs = {}) {
+ llvm::ArrayRef<mlir::Type> retTy;
+ Op op = builder.create<Op>(loc, retTy, operands);
+ builder.createBlock(&op.getRegion(), op.getRegion().end(), argsTy, locs);
mlir::Block &block = op.getRegion().back();
builder.setInsertionPointToStart(&block);
@@ -1488,12 +1491,22 @@ static void genWaitClause(Fortran::lower::AbstractConverter &converter,
}
}
+mlir::Type getTypeFromIvTypeSize(fir::FirOpBuilder &builder,
+ const Fortran::semantics::Symbol &ivSym) {
+ std::size_t ivTypeSize = ivSym.size();
+ if (ivTypeSize == 0)
+ llvm::report_fatal_error("unexpected induction variable size");
+ // ivTypeSize is in bytes and IntegerType needs to be in bits.
+ return builder.getIntegerType(ivTypeSize * 8);
+}
+
static mlir::acc::LoopOp
createLoopOp(Fortran::lower::AbstractConverter &converter,
mlir::Location currentLocation,
- Fortran::lower::pft::Evaluation &eval,
Fortran::semantics::SemanticsContext &semanticsContext,
Fortran::lower::StatementContext &stmtCtx,
+ const Fortran::parser::DoConstruct &outerDoConstruct,
+ Fortran::lower::pft::Evaluation &eval,
const Fortran::parser::AccClauseList &accClauseList) {
fir::FirOpBuilder &builder = converter.getFirOpBuilder();
@@ -1502,11 +1515,73 @@ createLoopOp(Fortran::lower::AbstractConverter &converter,
mlir::Value gangNum;
mlir::Value gangDim;
mlir::Value gangStatic;
- llvm::SmallVector<mlir::Value> tileOperands, privateOperands,
- reductionOperands, cacheOperands;
+ llvm::SmallVector<mlir::Value> tileOperands, privateOperands, ivPrivate,
+ reductionOperands, cacheOperands, lowerbounds, upperbounds, steps;
llvm::SmallVector<mlir::Attribute> privatizations, reductionRecipes;
+
bool hasGang = false, hasVector = false, hasWorker = false;
+ llvm::SmallVector<mlir::Type> ivTypes;
+ llvm::SmallVector<mlir::Location> ivLocs;
+ llvm::SmallVector<bool> inclusiveBounds;
+
+ if (outerDoConstruct.IsDoConcurrent())
+ TODO(currentLocation, "OpenACC loop with DO CONCURRENT");
+
+ int64_t collapseValue = Fortran::lower::getCollapseValue(accClauseList);
+ Fortran::lower::pft::Evaluation *crtEval = &eval.getFirstNestedEvaluation();
+ for (unsigned i = 0; i < collapseValue; ++i) {
+ const Fortran::parser::LoopControl *loopControl;
+ if (i == 0) {
+ loopControl = &*outerDoConstruct.GetLoopControl();
+ } else {
+ auto *doCons = crtEval->getIf<Fortran::parser::DoConstruct>();
+ assert(doCons && "expect do construct");
+ loopControl = &*doCons->GetLoopControl();
+ }
+
+ const Fortran::parser::LoopControl::Bounds *bounds =
+ std::get_if<Fortran::parser::LoopControl::Bounds>(&loopControl->u);
+ assert(bounds && "Expected bounds on the loop construct");
+ lowerbounds.push_back(fir::getBase(converter.genExprValue(
+ *Fortran::semantics::GetExpr(bounds->lower), stmtCtx)));
+ upperbounds.push_back(fir::getBase(converter.genExprValue(
+ *Fortran::semantics::GetExpr(bounds->upper), stmtCtx)));
+ if (bounds->step)
+ steps.push_back(fir::getBase(converter.genExprValue(
+ *Fortran::semantics::GetExpr(bounds->step), stmtCtx)));
+ else // If `step` is not present, assume it as `1`.
+ steps.push_back(builder.createIntegerConstant(
+ currentLocation, upperbounds[upperbounds.size() - 1].getType(), 1));
+
+ Fortran::semantics::Symbol &ivSym =
+ bounds->name.thing.symbol->GetUltimate();
+
+ mlir::Type ivTy = getTypeFromIvTypeSize(builder, ivSym);
+ mlir::Value ivValue = converter.getSymbolAddress(ivSym);
+ ivTypes.push_back(ivTy);
+ ivLocs.push_back(currentLocation);
+ std::string recipeName =
+ fir::getTypeAsString(ivValue.getType(), converter.getKindMap(),
+ Fortran::lower::privatizationRecipePrefix);
+ auto recipe = Fortran::lower::createOrGetPrivateRecipe(
+ builder, recipeName, currentLocation, ivValue.getType());
+ std::stringstream asFortran;
+ auto op = createDataEntryOp<mlir::acc::PrivateOp>(
+ builder, currentLocation, ivValue, asFortran, {}, true,
+ /*implicit=*/true, mlir::acc::DataClause::acc_private,
+ ivValue.getType());
+
+ privateOperands.push_back(op.getAccPtr());
+ ivPrivate.push_back(op.getAccPtr());
+ privatizations.push_back(mlir::SymbolRefAttr::get(
+ builder.getContext(), recipe.getSymName().str()));
+ inclusiveBounds.push_back(true);
+ converter.bindSymbol(ivSym, op.getAccPtr());
+ if (i < collapseValue - 1)
+ crtEval = &*std::next(crtEval->getNestedEvaluations().begin());
+ }
+
for (const Fortran::parser::AccClause &clause : accClauseList.v) {
mlir::Location clauseLocation = converter.genLocation(clause.source);
if (const auto *gangClause =
@@ -1589,6 +1664,9 @@ createLoopOp(Fortran::lower::AbstractConverter &converter,
// Prepare the operand segment size attribute and the operands value range.
llvm::SmallVector<mlir::Value> operands;
llvm::SmallVector<int32_t> operandSegments;
+ addOperands(operands, operandSegments, lowerbounds);
+ addOperands(operands, operandSegments, upperbounds);
+ addOperands(operands, operandSegments, steps);
addOperand(operands, operandSegments, gangNum);
addOperand(operands, operandSegments, gangDim);
addOperand(operands, operandSegments, gangStatic);
@@ -1600,7 +1678,14 @@ createLoopOp(Fortran::lower::AbstractConverter &converter,
addOperands(operands, operandSegments, reductionOperands);
auto loopOp = createRegionOp<mlir::acc::LoopOp, mlir::acc::YieldOp>(
- builder, currentLocation, eval, operands, operandSegments);
+ builder, currentLocation, eval, operands, operandSegments,
+ /*outerCombined=*/false, ivTypes, ivLocs);
+
+ for (auto [arg, value] : llvm::zip(
+ loopOp.getLoopRegions().front()->front().getArguments(), ivPrivate))
+ builder.create<fir::StoreOp>(currentLocation, arg, value);
+
+ loopOp.setInclusiveUpperbound(inclusiveBounds);
if (hasGang)
loopOp.setHasGangAttr(builder.getUnitAttr());
@@ -1661,12 +1746,16 @@ static void genACC(Fortran::lower::AbstractConverter &converter,
converter.genLocation(beginLoopDirective.source);
Fortran::lower::StatementContext stmtCtx;
- if (loopDirective.v == llvm::acc::ACCD_loop) {
- const auto &accClauseList =
- std::get<Fortran::parser::AccClauseList>(beginLoopDirective.t);
- createLoopOp(converter, currentLocation, eval, semanticsContext, stmtCtx,
- accClauseList);
- }
+ assert(loopDirective.v == llvm::acc::ACCD_loop &&
+ "Unsupported OpenACC loop construct");
+
+ const auto &accClauseList =
+ std::get<Fortran::parser::AccClauseList>(beginLoopDirective.t);
+ const auto &outerDoConstruct =
+ std::get<std::optional<Fortran::parser::DoConstruct>>(loopConstruct.t);
+
+ createLoopOp(converter, currentLocation, semanticsContext, stmtCtx,
+ *outerDoConstruct, eval, accClauseList);
}
template <typename Op, typename Clause>
@@ -2242,6 +2331,9 @@ genACC(Fortran::lower::AbstractConverter &converter,
std::get<Fortran::parser::AccCombinedDirective>(beginCombinedDirective.t);
const auto &accClauseList =
std::get<Fortran::parser::AccClauseList>(beginCombinedDirective.t);
+ const auto &outerDoConstruct =
+ std::get<std::optional<Fortran::parser::DoConstruct>>(
+ combinedConstruct.t);
mlir::Location currentLocation =
converter.genLocation(beginCombinedDirective.source);
@@ -2251,20 +2343,20 @@ genACC(Fortran::lower::AbstractConverter &converter,
createComputeOp<mlir::acc::KernelsOp>(
converter, currentLocation, eval, semanticsContext, stmtCtx,
accClauseList, /*outerCombined=*/true);
- createLoopOp(converter, currentLocation, eval, semanticsContext, stmtCtx,
- accClauseList);
+ createLoopOp(converter, currentLocation, semanticsContext, stmtCtx,
+ *outerDoConstruct, eval, accClauseList);
} else if (combinedDirective.v == llvm::acc::ACCD_parallel_loop) {
createComputeOp<mlir::acc::ParallelOp>(
converter, currentLocation, eval, semanticsContext, stmtCtx,
accClauseList, /*outerCombined=*/true);
- createLoopOp(converter, currentLocation, eval, semanticsContext, stmtCtx,
- accClauseList);
+ createLoopOp(converter, currentLocation, semanticsContext, stmtCtx,
+ *outerDoConstruct, eval, accClauseList);
} else if (combinedDirective.v == llvm::acc::ACCD_serial_loop) {
createComputeOp<mlir::acc::SerialOp>(converter, currentLocation, eval,
semanticsContext, stmtCtx,
accClauseList, /*outerCombined=*/true);
- createLoopOp(converter, currentLocation, eval, semanticsContext, stmtCtx,
- accClauseList);
+ createLoopOp(converter, currentLocation, semanticsContext, stmtCtx,
+ *outerDoConstruct, eval, accClauseList);
} else {
llvm::report_fatal_error("Unknown combined construct encountered");
}
@@ -3560,3 +3652,16 @@ void Fortran::lower::genOpenACCTerminator(fir::FirOpBuilder &builder,
else
builder.create<mlir::acc::TerminatorOp>(loc);
}
+
+int64_t Fortran::lower::getCollapseValue(
+ const Fortran::parser::AccClauseList &clauseList) {
+ for (const Fortran::parser::AccClause &clause : clauseList.v) {
+ if (const auto *collapseClause =
+ std::get_if<Fortran::parser::AccClause::Collapse>(&clause.u)) {
+ const parser::AccCollapseArg &arg = collapseClause->v;
+ const auto &collapseValue{std::get<parser::ScalarIntConstantExpr>(arg.t)};
+ return *Fortran::semantics::GetIntValue(collapseValue);
+ }
+ }
+ return 1;
+}
diff --git a/flang/test/Lower/OpenACC/acc-kernels-loop.f90 b/flang/test/Lower/OpenACC/acc-kernels-loop.f90
index 34e723269724173..2469f251df970f3 100644
--- a/flang/test/Lower/OpenACC/acc-kernels-loop.f90
+++ b/flang/test/Lower/OpenACC/acc-kernels-loop.f90
@@ -43,8 +43,7 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -57,8 +56,7 @@ subroutine acc_kernels_loop
!$acc end kernels loop
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -71,8 +69,7 @@ subroutine acc_kernels_loop
! CHECK: [[ASYNC1:%.*]] = arith.constant 1 : i32
! CHECK: acc.kernels async([[ASYNC1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -85,8 +82,7 @@ subroutine acc_kernels_loop
! CHECK: [[ASYNC2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.kernels async([[ASYNC2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -98,8 +94,7 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -112,8 +107,7 @@ subroutine acc_kernels_loop
! CHECK: [[WAIT1:%.*]] = arith.constant 1 : i32
! CHECK: acc.kernels wait([[WAIT1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -127,8 +121,7 @@ subroutine acc_kernels_loop
! CHECK: [[WAIT2:%.*]] = arith.constant 1 : i32
! CHECK: [[WAIT3:%.*]] = arith.constant 2 : i32
! CHECK: acc.kernels wait([[WAIT2]], [[WAIT3]] : i32, i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -142,8 +135,7 @@ subroutine acc_kernels_loop
! CHECK: [[WAIT4:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: [[WAIT5:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.kernels wait([[WAIT4]], [[WAIT5]] : i32, i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -156,8 +148,7 @@ subroutine acc_kernels_loop
! CHECK: [[NUMGANGS1:%.*]] = arith.constant 1 : i32
! CHECK: acc.kernels num_gangs([[NUMGANGS1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -170,8 +161,7 @@ subroutine acc_kernels_loop
! CHECK: [[NUMGANGS2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.kernels num_gangs([[NUMGANGS2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -184,8 +174,7 @@ subroutine acc_kernels_loop
! CHECK: [[NUMWORKERS1:%.*]] = arith.constant 10 : i32
! CHECK: acc.kernels num_workers([[NUMWORKERS1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -198,8 +187,7 @@ subroutine acc_kernels_loop
! CHECK: [[NUMWORKERS2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.kernels num_workers([[NUMWORKERS2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -212,8 +200,7 @@ subroutine acc_kernels_loop
! CHECK: [[VECTORLENGTH1:%.*]] = arith.constant 128 : i32
! CHECK: acc.kernels vector_length([[VECTORLENGTH1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -226,8 +213,7 @@ subroutine acc_kernels_loop
! CHECK: [[VECTORLENGTH2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.kernels vector_length([[VECTORLENGTH2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -240,8 +226,7 @@ subroutine acc_kernels_loop
! CHECK: [[IF1:%.*]] = arith.constant true
! CHECK: acc.kernels if([[IF1]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -255,8 +240,7 @@ subroutine acc_kernels_loop
! CHECK: [[IFCOND:%.*]] = fir.load %{{.*}} : !fir.ref<!fir.logical<4>>
! CHECK: [[IF2:%.*]] = fir.convert [[IFCOND]] : (!fir.logical<4>) -> i1
! CHECK: acc.kernels if([[IF2]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -269,8 +253,7 @@ subroutine acc_kernels_loop
! CHECK: [[SELF1:%.*]] = arith.constant true
! CHECK: acc.kernels self([[SELF1]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -282,8 +265,7 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -297,8 +279,7 @@ subroutine acc_kernels_loop
! CHECK: %[[SELF2:.*]] = fir.convert %[[DECLIFCONDITION]]#1 : (!fir.ref<!fir.logical<4>>) -> i1
! CHECK: acc.kernels self(%[[SELF2]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -313,8 +294,7 @@ subroutine acc_kernels_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "b"}
! CHECK: acc.kernels dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -330,8 +310,7 @@ subroutine acc_kernels_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "b"}
! CHECK: acc.kernels dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -347,8 +326,7 @@ subroutine acc_kernels_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyin_readonly>, name = "b"}
! CHECK: acc.kernels dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -362,8 +340,7 @@ subroutine acc_kernels_loop
! CHECK: %[[CREATE_A:.*]] = acc.create varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyout>, name = "a"}
! CHECK: %[[CREATE_B:.*]] = acc.create varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyout>, name = "b"}
! CHECK: acc.kernels dataOperands(%[[CREATE_A]], %[[CREATE_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -379,8 +356,7 @@ subroutine acc_kernels_loop
! CHECK: %[[CREATE_B:.*]] = acc.create varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: %[[CREATE_A:.*]] = acc.create varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_create_zero>, name = "a"}
! CHECK: acc.kernels dataOperands(%[[CREATE_B]], %[[CREATE_A]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -396,8 +372,7 @@ subroutine acc_kernels_loop
! CHECK: %[[NOCREATE_A:.*]] = acc.nocreate varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[NOCREATE_B:.*]] = acc.nocreate varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.kernels dataOperands(%[[NOCREATE_A]], %[[NOCREATE_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -411,8 +386,7 @@ subroutine acc_kernels_loop
! CHECK: %[[PRESENT_A:.*]] = acc.present varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[PRESENT_B:.*]] = acc.present varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.kernels dataOperands(%[[PRESENT_A]], %[[PRESENT_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -426,8 +400,7 @@ subroutine acc_kernels_loop
! CHECK: %[[DEVICEPTR_A:.*]] = acc.deviceptr varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[DEVICEPTR_B:.*]] = acc.deviceptr varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.kernels dataOperands(%[[DEVICEPTR_A]], %[[DEVICEPTR_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -445,8 +418,7 @@ subroutine acc_kernels_loop
! CHECK: %[[BOX_ADDR_G:.*]] = fir.box_addr %[[BOX_G]] : (!fir.box<!fir.ptr<f32>>) -> !fir.ptr<f32>
! CHECK: %[[ATTACH_G:.*]] = acc.attach varPtr(%[[BOX_ADDR_G]] : !fir.ptr<f32>) -> !fir.ptr<f32> {name = "g"}
! CHECK: acc.kernels dataOperands(%[[ATTACH_F]], %[[ATTACH_G]] : !fir.ptr<f32>, !fir.ptr<f32>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -458,10 +430,9 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {seq}
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, seq}
! CHECK: acc.terminator
! CHECK-NEXT: }{{$}}
@@ -471,10 +442,9 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {auto}
+! CHECK-NEXT: } attributes {auto, inclusiveUpperbound = array<i1: true>}
! CHECK: acc.terminator
! CHECK-NEXT: }{{$}}
@@ -484,10 +454,9 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {independent}
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, independent}
! CHECK: acc.terminator
! CHECK-NEXT: }{{$}}
@@ -497,8 +466,7 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop gang {
-! CHECK: fir.do_loop
+! CHECK: acc.loop gang() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -511,8 +479,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[GANGNUM1:%.*]] = arith.constant 8 : i32
-! CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) {
-! CHECK: fir.do_loop
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -525,8 +492,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[GANGNUM2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-! CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) {
-! CHECK: fir.do_loop
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -539,7 +505,6 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: acc.loop gang(num=%{{.*}} : i32, static=%{{.*}} : i32) {
-! CHECK: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -550,8 +515,7 @@ subroutine acc_kernels_loop
a(i) = b(i)
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop vector {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -564,8 +528,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[CONSTANT128:%.*]] = arith.constant 128 : i32
-! CHECK: acc.loop vector([[CONSTANT128]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector([[CONSTANT128]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -578,8 +541,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[VECTORLENGTH:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -591,8 +553,7 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop worker {
-! CHECK: fir.do_loop
+! CHECK: acc.loop worker() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -605,8 +566,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[WORKER128:%.*]] = arith.constant 128 : i32
-! CHECK: acc.loop worker([[WORKER128]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop worker([[WORKER128]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -620,11 +580,9 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {collapse = 2 : i64}
+! CHECK-NEXT: } attributes {collapse = 2 : i64, inclusiveUpperbound = array<i1: true, true>}
! CHECK: acc.terminator
! CHECK-NEXT: }{{$}}
@@ -637,10 +595,8 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -655,8 +611,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[TILESIZE:%.*]] = arith.constant 2 : i32
-! CHECK: acc.loop tile([[TILESIZE]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZE]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -669,8 +624,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[TILESIZEM1:%.*]] = arith.constant -1 : i32
-! CHECK: acc.loop tile([[TILESIZEM1]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZEM1]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -686,8 +640,7 @@ subroutine acc_kernels_loop
! CHECK: acc.kernels {
! CHECK: [[TILESIZE1:%.*]] = arith.constant 2 : i32
! CHECK: [[TILESIZE2:%.*]] = arith.constant 2 : i32
-! CHECK: acc.loop tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -699,8 +652,7 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop tile(%{{.*}} : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile(%{{.*}} : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -714,8 +666,7 @@ subroutine acc_kernels_loop
END DO
! CHECK: acc.kernels {
-! CHECK: acc.loop tile(%{{.*}}, %{{.*}} : i32, i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile(%{{.*}}, %{{.*}} : i32, i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
@@ -730,8 +681,7 @@ subroutine acc_kernels_loop
! CHECK: %[[COPYINREDR:.*]] = acc.copyin varPtr(%{{.*}} : !fir.ref<f32>) -> !fir.ref<f32> {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_r"}
! CHECK: %[[COPYINREDI:.*]] = acc.copyin varPtr(%{{.*}} : !fir.ref<i32>) -> !fir.ref<i32> {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_i"}
! CHECK: acc.kernels dataOperands(%[[COPYINREDR]], %[[COPYINREDI]] : !fir.ref<f32>, !fir.ref<i32>) {
-! CHECK: acc.loop reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.terminator
diff --git a/flang/test/Lower/OpenACC/acc-loop.f90 b/flang/test/Lower/OpenACC/acc-loop.f90
index 924574512da4c73..224f7f983bccf87 100644
--- a/flang/test/Lower/OpenACC/acc-loop.f90
+++ b/flang/test/Lower/OpenACC/acc-loop.f90
@@ -28,185 +28,168 @@ program acc_loop
a(i) = b(i)
END DO
-!CHECK: acc.loop {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}{{$}}
!$acc loop seq
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: } attributes {seq}
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, seq}
!$acc loop auto
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: } attributes {auto}
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {auto, inclusiveUpperbound = array<i1: true>}
!$acc loop independent
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: } attributes {independent}
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, independent}
!$acc loop gang
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop gang {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop gang() private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop gang(num: 8)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[GANGNUM1:%.*]] = arith.constant 8 : i32
-!CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: [[GANGNUM1:%.*]] = arith.constant 8 : i32
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop gang(num: gangNum)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[GANGNUM2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-!CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: [[GANGNUM2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop gang(num: gangNum, static: gangStatic)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop gang(num=%{{.*}} : i32, static=%{{.*}} : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop gang(num=%{{.*}} : i32, static=%{{.*}} : i32) private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop vector
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop vector {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop vector() private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop vector(128)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[CONSTANT128:%.*]] = arith.constant 128 : i32
-!CHECK: acc.loop vector([[CONSTANT128]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: [[CONSTANT128:%.*]] = arith.constant 128 : i32
+! CHECK: acc.loop vector([[CONSTANT128]] : i32) private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: }{{$}}
!$acc loop vector(vectorLength)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[VECTORLENGTH:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-!CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: [[VECTORLENGTH:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
+! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop worker
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop worker {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop worker() private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop worker(128)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[WORKER128:%.*]] = arith.constant 128 : i32
-!CHECK: acc.loop worker([[WORKER128]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: [[WORKER128:%.*]] = arith.constant 128 : i32
+! CHECK: acc.loop worker([[WORKER128]] : i32) private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop private(c)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop private(@privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>, @privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop private(c, d)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop private(@privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>, @privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>, @privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>, @privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop private(c) private(d)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop private(@privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>, @privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>, @privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>, @privatization_ref_10x10xf32 -> %{{.*}} : !fir.ref<!fir.array<10x10xf32>>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop tile(2)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[TILESIZE:%.*]] = arith.constant 2 : i32
-!CHECK: acc.loop tile([[TILESIZE]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+
+! CHECK: [[TILESIZE:%.*]] = arith.constant 2 : i32
+! CHECK: acc.loop {{.*}} tile([[TILESIZE]] : i32) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop tile(*)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[TILESIZEM1:%.*]] = arith.constant -1 : i32
-!CHECK: acc.loop tile([[TILESIZEM1]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: [[TILESIZEM1:%.*]] = arith.constant -1 : i32
+! CHECK: acc.loop {{.*}} tile([[TILESIZEM1]] : i32) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop tile(2, 2)
DO i = 1, n
@@ -215,22 +198,20 @@ program acc_loop
END DO
END DO
-!CHECK: [[TILESIZE1:%.*]] = arith.constant 2 : i32
-!CHECK: [[TILESIZE2:%.*]] = arith.constant 2 : i32
-!CHECK: acc.loop tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: [[TILESIZE1:%.*]] = arith.constant 2 : i32
+! CHECK: [[TILESIZE2:%.*]] = arith.constant 2 : i32
+! CHECK: acc.loop {{.*}} tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop tile(tileSize)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop tile(%{{.*}} : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop {{.*}} tile(%{{.*}} : i32) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop tile(tileSize, tileSize)
DO i = 1, n
@@ -239,10 +220,9 @@ program acc_loop
END DO
END DO
-!CHECK: acc.loop tile(%{{.*}}, %{{.*}} : i32, i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop {{.*}} tile(%{{.*}}, %{{.*}} : i32, i32) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop collapse(2)
DO i = 1, n
@@ -251,11 +231,11 @@ program acc_loop
END DO
END DO
-!CHECK: acc.loop {
-!CHECK: fir.do_loop
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: } attributes {collapse = 2 : i64}
+! CHECK: acc.loop {{.*}} (%arg0 : i32, %arg1 : i32) = (%{{.*}} : i32, i32) to (%{{.*}} : i32, i32) step (%{{.*}} : i32, i32) {
+! CHECK: fir.store %arg0 to %{{.*}} : !fir.ref<i32>
+! CHECK: fir.store %arg1 to %{{.*}} : !fir.ref<i32>
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {collapse = 2 : i64, inclusiveUpperbound = array<i1: true, true>}
!$acc loop
DO i = 1, n
@@ -265,14 +245,12 @@ program acc_loop
END DO
END DO
-!CHECK: acc.loop {
-!CHECK: fir.do_loop
-!CHECK: acc.loop {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop {{.*}} (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.loop {{.*}} (%arg1 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop reduction(+:reduction_r) reduction(*:reduction_i)
do i = 1, n
@@ -280,31 +258,27 @@ program acc_loop
reduction_i = 1
end do
-! CHECK: acc.loop reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop private(@privatization_ref_i32 -> %{{.*}} : !fir.ref<i32>) reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>) (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
! CHECK: acc.yield
-! CHECK-NEXT: }{{$}}
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop gang(dim: gangDim, static: gangStatic)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: acc.loop gang(dim=%{{.*}}, static=%{{.*}} : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop gang(dim=%{{.*}}, static=%{{.*}} : i32) {{.*}} (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop gang(dim: 1)
DO i = 1, n
a(i) = b(i)
END DO
-!CHECK: [[GANGDIM1:%.*]] = arith.constant 1 : i32
-!CHECK-NEXT: acc.loop gang(dim=[[GANGDIM1]] : i32) {
-!CHECK: fir.do_loop
-!CHECK: acc.yield
-!CHECK-NEXT: }{{$}}
+! CHECK: acc.loop gang(dim=%{{.*}} : i32) {{.*}} (%arg0 : i32) = (%{{.*}} : i32) to (%{{.*}} : i32) step (%{{.*}} : i32) {
+! CHECK: acc.yield
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>}
!$acc loop
DO i = 1, n
@@ -313,12 +287,11 @@ program acc_loop
END DO
! CHECK: %[[CACHE:.*]] = acc.cache varPtr(%{{.*}} : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
-! CHECK: acc.loop cache(%[[CACHE]] : !fir.ref<!fir.array<10xf32>>)
+! CHECK: acc.loop {{.*}} cache(%[[CACHE]] : !fir.ref<!fir.array<10xf32>>)
!$acc loop
do 100 i=0, n
100 continue
! CHECK: acc.loop
-! CHECK: fir.do_loop
end program
diff --git a/flang/test/Lower/OpenACC/acc-parallel-loop.f90 b/flang/test/Lower/OpenACC/acc-parallel-loop.f90
index 1856215ce59d134..712e00918169534 100644
--- a/flang/test/Lower/OpenACC/acc-parallel-loop.f90
+++ b/flang/test/Lower/OpenACC/acc-parallel-loop.f90
@@ -45,8 +45,7 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -59,8 +58,7 @@ subroutine acc_parallel_loop
!$acc end parallel loop
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -73,8 +71,7 @@ subroutine acc_parallel_loop
! CHECK: [[ASYNC1:%.*]] = arith.constant 1 : i32
! CHECK: acc.parallel async([[ASYNC1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -87,8 +84,7 @@ subroutine acc_parallel_loop
! CHECK: [[ASYNC2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.parallel async([[ASYNC2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -100,8 +96,7 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -114,8 +109,7 @@ subroutine acc_parallel_loop
! CHECK: [[WAIT1:%.*]] = arith.constant 1 : i32
! CHECK: acc.parallel wait([[WAIT1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -129,8 +123,7 @@ subroutine acc_parallel_loop
! CHECK: [[WAIT2:%.*]] = arith.constant 1 : i32
! CHECK: [[WAIT3:%.*]] = arith.constant 2 : i32
! CHECK: acc.parallel wait([[WAIT2]], [[WAIT3]] : i32, i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -144,8 +137,7 @@ subroutine acc_parallel_loop
! CHECK: [[WAIT4:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: [[WAIT5:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.parallel wait([[WAIT4]], [[WAIT5]] : i32, i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -158,8 +150,7 @@ subroutine acc_parallel_loop
! CHECK: [[NUMGANGS1:%.*]] = arith.constant 1 : i32
! CHECK: acc.parallel num_gangs([[NUMGANGS1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -172,8 +163,7 @@ subroutine acc_parallel_loop
! CHECK: [[NUMGANGS2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.parallel num_gangs([[NUMGANGS2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -186,8 +176,7 @@ subroutine acc_parallel_loop
! CHECK: [[NUMWORKERS1:%.*]] = arith.constant 10 : i32
! CHECK: acc.parallel num_workers([[NUMWORKERS1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -200,8 +189,7 @@ subroutine acc_parallel_loop
! CHECK: [[NUMWORKERS2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.parallel num_workers([[NUMWORKERS2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -214,8 +202,7 @@ subroutine acc_parallel_loop
! CHECK: [[VECTORLENGTH1:%.*]] = arith.constant 128 : i32
! CHECK: acc.parallel vector_length([[VECTORLENGTH1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -228,8 +215,7 @@ subroutine acc_parallel_loop
! CHECK: [[VECTORLENGTH2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.parallel vector_length([[VECTORLENGTH2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -242,8 +228,7 @@ subroutine acc_parallel_loop
! CHECK: [[IF1:%.*]] = arith.constant true
! CHECK: acc.parallel if([[IF1]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -257,8 +242,7 @@ subroutine acc_parallel_loop
! CHECK: [[IFCOND:%.*]] = fir.load %{{.*}} : !fir.ref<!fir.logical<4>>
! CHECK: [[IF2:%.*]] = fir.convert [[IFCOND]] : (!fir.logical<4>) -> i1
! CHECK: acc.parallel if([[IF2]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -271,8 +255,7 @@ subroutine acc_parallel_loop
! CHECK: [[SELF1:%.*]] = arith.constant true
! CHECK: acc.parallel self([[SELF1]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -284,8 +267,7 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -298,8 +280,7 @@ subroutine acc_parallel_loop
! CHECK: %[[SELF2:.*]] = fir.convert %[[DECLIFCONDITION]]#1 : (!fir.ref<!fir.logical<4>>) -> i1
! CHECK: acc.parallel self(%[[SELF2]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -313,8 +294,7 @@ subroutine acc_parallel_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "b"}
! CHECK: acc.parallel dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -330,8 +310,7 @@ subroutine acc_parallel_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "b"}
! CHECK: acc.parallel dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -347,8 +326,7 @@ subroutine acc_parallel_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyin_readonly>, name = "b"}
! CHECK: acc.parallel dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -362,8 +340,7 @@ subroutine acc_parallel_loop
! CHECK: %[[CREATE_A:.*]] = acc.create varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyout>, name = "a"}
! CHECK: %[[CREATE_B:.*]] = acc.create varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyout>, name = "b"}
! CHECK: acc.parallel dataOperands(%[[CREATE_A]], %[[CREATE_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -379,8 +356,7 @@ subroutine acc_parallel_loop
! CHECK: %[[CREATE_B:.*]] = acc.create varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: %[[CREATE_A:.*]] = acc.create varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_create_zero>, name = "a"}
! CHECK: acc.parallel dataOperands(%[[CREATE_B]], %[[CREATE_A]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -396,8 +372,7 @@ subroutine acc_parallel_loop
! CHECK: %[[NOCREATE_A:.*]] = acc.nocreate varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[NOCREATE_B:.*]] = acc.nocreate varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.parallel dataOperands(%[[NOCREATE_A]], %[[NOCREATE_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -411,8 +386,7 @@ subroutine acc_parallel_loop
! CHECK: %[[PRESENT_A:.*]] = acc.present varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[PRESENT_B:.*]] = acc.present varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.parallel dataOperands(%[[PRESENT_A]], %[[PRESENT_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -426,8 +400,7 @@ subroutine acc_parallel_loop
! CHECK: %[[DEVICEPTR_A:.*]] = acc.deviceptr varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[DEVICEPTR_B:.*]] = acc.deviceptr varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.parallel dataOperands(%[[DEVICEPTR_A]], %[[DEVICEPTR_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -445,8 +418,7 @@ subroutine acc_parallel_loop
! CHECK: %[[BOX_ADDR_G:.*]] = fir.box_addr %[[BOX_G]] : (!fir.box<!fir.ptr<f32>>) -> !fir.ptr<f32>
! CHECK: %[[ATTACH_G:.*]] = acc.attach varPtr(%[[BOX_ADDR_G]] : !fir.ptr<f32>) -> !fir.ptr<f32> {name = "g"}
! CHECK: acc.parallel dataOperands(%[[ATTACH_F]], %[[ATTACH_G]] : !fir.ptr<f32>, !fir.ptr<f32>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -460,7 +432,7 @@ subroutine acc_parallel_loop
! CHECK: %[[ACC_PRIVATE_B:.*]] = acc.firstprivate varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.parallel firstprivate(@firstprivatization_section_ext10_ref_10xf32 -> %[[ACC_PRIVATE_B]] : !fir.ref<!fir.array<10xf32>>) {
! CHECK: %[[ACC_PRIVATE_A:.*]] = acc.private varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
-! CHECK: acc.loop private(@privatization_ref_10xf32 -> %[[ACC_PRIVATE_A]] : !fir.ref<!fir.array<10xf32>>) {
+! CHECK: acc.loop private({{.*}}@privatization_ref_10xf32 -> %[[ACC_PRIVATE_A]] : !fir.ref<!fir.array<10xf32>>)
! CHECK: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -473,10 +445,9 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {seq}
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, seq}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -486,10 +457,9 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {auto}
+! CHECK-NEXT: } attributes {auto, inclusiveUpperbound = array<i1: true>}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -499,10 +469,9 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {independent}
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, independent}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -512,8 +481,7 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop gang {
-! CHECK: fir.do_loop
+! CHECK: acc.loop gang() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -526,8 +494,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[GANGNUM1:%.*]] = arith.constant 8 : i32
-! CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) {
-! CHECK: fir.do_loop
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -540,8 +507,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[GANGNUM2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-! CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) {
-! CHECK: fir.do_loop
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -553,8 +519,7 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop gang(num=%{{.*}} : i32, static=%{{.*}} : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop gang(num=%{{.*}} : i32, static=%{{.*}} : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -565,8 +530,7 @@ subroutine acc_parallel_loop
a(i) = b(i)
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop vector {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -579,8 +543,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[CONSTANT128:%.*]] = arith.constant 128 : i32
-! CHECK: acc.loop vector([[CONSTANT128]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector([[CONSTANT128]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -593,7 +556,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[VECTORLENGTH:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {
+! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {{.*}} {
! CHECK: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -606,7 +569,7 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop worker {
+! CHECK: acc.loop worker() {{.*}} {
! CHECK: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -620,7 +583,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[WORKER128:%.*]] = arith.constant 128 : i32
-! CHECK: acc.loop worker([[WORKER128]] : i32) {
+! CHECK: acc.loop worker([[WORKER128]] : i32) {{.*}} {
! CHECK: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -635,11 +598,9 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {collapse = 2 : i64}
+! CHECK-NEXT: } attributes {collapse = 2 : i64, inclusiveUpperbound = array<i1: true, true>}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -652,10 +613,8 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -670,22 +629,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[TILESIZE:%.*]] = arith.constant 2 : i32
-! CHECK: acc.loop tile([[TILESIZE]] : i32) {
-! CHECK: fir.do_loop
-! CHECK: acc.yield
-! CHECK-NEXT: }{{$}}
-! CHECK: acc.yield
-! CHECK-NEXT: }{{$}}
-
- !$acc parallel loop tile(*)
- DO i = 1, n
- a(i) = b(i)
- END DO
-
-! CHECK: acc.parallel {
-! CHECK: [[TILESIZEM1:%.*]] = arith.constant -1 : i32
-! CHECK: acc.loop tile([[TILESIZEM1]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZE]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -701,8 +645,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[TILESIZE1:%.*]] = arith.constant 2 : i32
! CHECK: [[TILESIZE2:%.*]] = arith.constant 2 : i32
-! CHECK: acc.loop tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -714,8 +657,7 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop tile(%{{.*}} : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile(%{{.*}} : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -729,8 +671,20 @@ subroutine acc_parallel_loop
END DO
! CHECK: acc.parallel {
-! CHECK: acc.loop tile(%{{.*}}, %{{.*}} : i32, i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile(%{{.*}}, %{{.*}} : i32, i32) {{.*}} {
+! CHECK: acc.yield
+! CHECK-NEXT: }{{$}}
+! CHECK: acc.yield
+! CHECK-NEXT: }{{$}}
+
+ !$acc parallel loop tile(*)
+ DO i = 1, n
+ a(i) = b(i)
+ END DO
+
+! CHECK: acc.parallel {
+! CHECK: [[TILESIZEM1:%.*]] = arith.constant -1 : i32
+! CHECK: acc.loop {{.*}} tile([[TILESIZEM1]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -745,8 +699,7 @@ subroutine acc_parallel_loop
! CHECK: %[[COPYINREDR:.*]] = acc.copyin varPtr(%{{.*}} : !fir.ref<f32>) -> !fir.ref<f32> {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_r"}
! CHECK: %[[COPYINREDI:.*]] = acc.copyin varPtr(%{{.*}} : !fir.ref<i32>) -> !fir.ref<i32> {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_i"}
! CHECK: acc.parallel dataOperands(%[[COPYINREDR]], %[[COPYINREDI]] : !fir.ref<f32>, !fir.ref<i32>) {
-! CHECK: acc.loop reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>) {{.*}}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -754,6 +707,7 @@ subroutine acc_parallel_loop
! CHECK: acc.copyout accPtr(%[[COPYINREDR]] : !fir.ref<f32>) to varPtr(%{{.*}} : !fir.ref<f32>) {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_r"}
! CHECK: acc.copyout accPtr(%[[COPYINREDI]] : !fir.ref<i32>) to varPtr(%{{.*}} : !fir.ref<i32>) {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_i"}
+
!$acc parallel loop
do 10 i=0, n
10 continue
diff --git a/flang/test/Lower/OpenACC/acc-private.f90 b/flang/test/Lower/OpenACC/acc-private.f90
index 0caaefc82225424..69820b1cde49619 100644
--- a/flang/test/Lower/OpenACC/acc-private.f90
+++ b/flang/test/Lower/OpenACC/acc-private.f90
@@ -189,7 +189,7 @@ program acc_private
END DO
! CHECK: %[[C_PRIVATE:.*]] = acc.private varPtr(%[[DECLC]]#1 : !fir.ref<i32>) -> !fir.ref<i32> {name = "c"}
-! CHECK: acc.loop private(@privatization_ref_i32 -> %[[C_PRIVATE]] : !fir.ref<i32>)
+! CHECK: acc.loop private({{.*}}@privatization_ref_i32 -> %[[C_PRIVATE]] : !fir.ref<i32>)
! CHECK: acc.yield
!$acc loop private(b)
@@ -203,7 +203,7 @@ program acc_private
! CHECK: %[[UB:.*]] = arith.subi %{{.*}}, %[[C1]] : index
! CHECK: %[[BOUND:.*]] = acc.bounds lowerbound(%[[LB]] : index) upperbound(%[[UB]] : index) extent(%{{.*}} : index) stride(%[[C1]] : index) startIdx(%[[C1]] : index)
! CHECK: %[[B_PRIVATE:.*]] = acc.private varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<100xf32>>) bounds(%[[BOUND]]) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
-! CHECK: acc.loop private(@privatization_ref_100xf32 -> %[[B_PRIVATE]] : !fir.ref<!fir.array<100xf32>>) {
+! CHECK: acc.loop private({{.*}}@privatization_ref_100xf32 -> %[[B_PRIVATE]] : !fir.ref<!fir.array<100xf32>>)
! CHECK: acc.yield
!$acc loop private(b(1:50))
@@ -217,7 +217,7 @@ program acc_private
! CHECK: %[[UB:.*]] = arith.constant 49 : index
! CHECK: %[[BOUND:.*]] = acc.bounds lowerbound(%[[LB]] : index) upperbound(%[[UB]] : index) stride(%[[C1]] : index) startIdx(%[[C1]] : index)
! CHECK: %[[B_PRIVATE:.*]] = acc.private varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<100xf32>>) bounds(%[[BOUND]]) -> !fir.ref<!fir.array<50xf32>> {name = "b(1:50)"}
-! CHECK: acc.loop private(@privatization_ref_50xf32 -> %[[B_PRIVATE]] : !fir.ref<!fir.array<50xf32>>)
+! CHECK: acc.loop private({{.*}}@privatization_ref_50xf32 -> %[[B_PRIVATE]] : !fir.ref<!fir.array<50xf32>>)
!$acc parallel loop firstprivate(c)
DO i = 1, n
@@ -273,7 +273,7 @@ subroutine acc_private_assumed_shape(a, n)
! CHECK: acc.parallel {
! CHECK: %[[ADDR:.*]] = fir.box_addr %[[DECL_A]]#1 : (!fir.box<!fir.array<?xi32>>) -> !fir.ref<!fir.array<?xi32>>
! CHECK: %[[PRIVATE:.*]] = acc.private varPtr(%[[ADDR]] : !fir.ref<!fir.array<?xi32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<?xi32>> {name = "a"}
-! CHECK: acc.loop private(@privatization_box_Uxi32 -> %[[PRIVATE]] : !fir.ref<!fir.array<?xi32>>) {
+! CHECK: acc.loop private({{.*}}@privatization_box_Uxi32 -> %[[PRIVATE]] : !fir.ref<!fir.array<?xi32>>)
subroutine acc_private_allocatable_array(a, n)
integer, allocatable :: a(:)
@@ -296,7 +296,7 @@ subroutine acc_private_allocatable_array(a, n)
! CHECK: %[[BOX:.*]] = fir.load %[[DECLA_A]]#1 : !fir.ref<!fir.box<!fir.heap<!fir.array<?xi32>>>>
! CHECK: %[[BOX_ADDR:.*]] = fir.box_addr %[[BOX]] : (!fir.box<!fir.heap<!fir.array<?xi32>>>) -> !fir.heap<!fir.array<?xi32>>
! CHECK: %[[PRIVATE:.*]] = acc.private varPtr(%[[BOX_ADDR]] : !fir.heap<!fir.array<?xi32>>) bounds(%{{.*}}) -> !fir.heap<!fir.array<?xi32>> {name = "a"}
-! CHECK: acc.loop private(@privatization_box_heap_Uxi32 -> %[[PRIVATE]] : !fir.heap<!fir.array<?xi32>>)
+! CHECK: acc.loop private({{.*}}@privatization_box_heap_Uxi32 -> %[[PRIVATE]] : !fir.heap<!fir.array<?xi32>>)
! CHECK: acc.serial private(@privatization_box_heap_Uxi32 -> %{{.*}} : !fir.heap<!fir.array<?xi32>>)
subroutine acc_private_pointer_array(a, n)
@@ -316,7 +316,7 @@ subroutine acc_private_pointer_array(a, n)
! CHECK: %[[BOX:.*]] = fir.load %[[DECLA_A]]#1 : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>
! CHECK: %[[BOX_ADDR:.*]] = fir.box_addr %[[BOX]] : (!fir.box<!fir.ptr<!fir.array<?xi32>>>) -> !fir.ptr<!fir.array<?xi32>>
! CHECK: %[[PRIVATE:.*]] = acc.private varPtr(%[[BOX_ADDR]] : !fir.ptr<!fir.array<?xi32>>) bounds(%{{.*}}) -> !fir.ptr<!fir.array<?xi32>> {name = "a"}
-! CHECK: acc.loop private(@privatization_box_ptr_Uxi32 -> %[[PRIVATE]] : !fir.ptr<!fir.array<?xi32>>)
+! CHECK: acc.loop private({{.*}}@privatization_box_ptr_Uxi32 -> %[[PRIVATE]] : !fir.ptr<!fir.array<?xi32>>)
subroutine acc_private_dynamic_extent(a, n)
integer :: n, i
@@ -334,7 +334,7 @@ subroutine acc_private_dynamic_extent(a, n)
! CHECK: %[[DECL_A:.*]]:2 = hlfir.declare %[[ARG0]](%16) {uniq_name = "_QFacc_private_dynamic_extentEa"} : (!fir.ref<!fir.array<?x?x2xi32>>, !fir.shape<3>) -> (!fir.box<!fir.array<?x?x2xi32>>, !fir.ref<!fir.array<?x?x2xi32>>)
! CHECK: acc.parallel {
! CHECK: %[[PRIV:.*]] = acc.private varPtr(%[[DECL_A]]#1 : !fir.ref<!fir.array<?x?x2xi32>>) bounds(%{{.*}}, %{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<?x?x2xi32>> {name = "a"}
-! CHECK: acc.loop private(@privatization_ref_UxUx2xi32 -> %[[PRIV]] : !fir.ref<!fir.array<?x?x2xi32>>)
+! CHECK: acc.loop private({{.*}}@privatization_ref_UxUx2xi32 -> %[[PRIV]] : !fir.ref<!fir.array<?x?x2xi32>>)
subroutine acc_firstprivate_assumed_shape(a, n)
integer :: a(:), i, n
diff --git a/flang/test/Lower/OpenACC/acc-reduction.f90 b/flang/test/Lower/OpenACC/acc-reduction.f90
index 98fef6a2d2f29a3..1b61c055001ce73 100644
--- a/flang/test/Lower/OpenACC/acc-reduction.f90
+++ b/flang/test/Lower/OpenACC/acc-reduction.f90
@@ -706,7 +706,7 @@ subroutine acc_reduction_add_int(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_add_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
subroutine acc_reduction_add_int_array_1d(a, b)
integer :: a(100)
@@ -723,7 +723,7 @@ subroutine acc_reduction_add_int_array_1d(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<!fir.array<100xi32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xi32>> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<100xi32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xi32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_add_section_ext100_ref_100xi32 -> %[[RED_B]] : !fir.ref<!fir.array<100xi32>>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_section_ext100_ref_100xi32 -> %[[RED_B]] : !fir.ref<!fir.array<100xi32>>)
subroutine acc_reduction_add_int_array_2d(a, b)
integer :: a(100, 10), b(100, 10)
@@ -742,8 +742,7 @@ subroutine acc_reduction_add_int_array_2d(a, b)
! HLFIR: %[[DECLARG1:.*]]:2 = hlfir.declare %[[ARG1]]
! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100x10xi32>>) bounds(%{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10xi32>> {name = "b"}
! HLFIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[DECLARG1]]#1 : !fir.ref<!fir.array<100x10xi32>>) bounds(%{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10xi32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_add_section_ext100xext10_ref_100x10xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10xi32>>) {
-! CHECK: } attributes {collapse = 2 : i64}
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_section_ext100xext10_ref_100x10xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10xi32>>)
subroutine acc_reduction_add_int_array_3d(a, b)
integer :: a(100, 10, 2), b(100, 10, 2)
@@ -764,8 +763,7 @@ subroutine acc_reduction_add_int_array_3d(a, b)
! HLFIR: %[[DECLARG1:.*]]:2 = hlfir.declare %[[ARG1]]
! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100x10x2xi32>>) bounds(%{{.*}}, %{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10x2xi32>> {name = "b"}
! HLFIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[DECLARG1]]#1 : !fir.ref<!fir.array<100x10x2xi32>>) bounds(%{{.*}}, %{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10x2xi32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_add_section_ext100xext10xext2_ref_100x10x2xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10x2xi32>>)
-! CHECK: } attributes {collapse = 3 : i64}
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_section_ext100xext10xext2_ref_100x10x2xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10x2xi32>>)
subroutine acc_reduction_add_float(a, b)
real :: a(100), b
@@ -782,7 +780,7 @@ subroutine acc_reduction_add_float(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_add_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
subroutine acc_reduction_add_float_array_1d(a, b)
real :: a(100), b(100)
@@ -799,7 +797,7 @@ subroutine acc_reduction_add_float_array_1d(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<!fir.array<100xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<100xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_add_section_ext100_ref_100xf32 -> %[[RED_B]] : !fir.ref<!fir.array<100xf32>>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_section_ext100_ref_100xf32 -> %[[RED_B]] : !fir.ref<!fir.array<100xf32>>)
subroutine acc_reduction_mul_int(a, b)
integer :: a(100)
@@ -816,7 +814,7 @@ subroutine acc_reduction_mul_int(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_mul_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_mul_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
subroutine acc_reduction_mul_int_array_1d(a, b)
integer :: a(100)
@@ -833,7 +831,7 @@ subroutine acc_reduction_mul_int_array_1d(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<!fir.array<100xi32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xi32>> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<100xi32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xi32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_mul_section_ext100_ref_100xi32 -> %[[RED_B]] : !fir.ref<!fir.array<100xi32>>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_mul_section_ext100_ref_100xi32 -> %[[RED_B]] : !fir.ref<!fir.array<100xi32>>)
subroutine acc_reduction_mul_float(a, b)
real :: a(100), b
@@ -850,7 +848,7 @@ subroutine acc_reduction_mul_float(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_mul_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_mul_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
subroutine acc_reduction_mul_float_array_1d(a, b)
real :: a(100), b(100)
@@ -865,9 +863,9 @@ subroutine acc_reduction_mul_float_array_1d(a, b)
! CHECK-LABEL: func.func @_QPacc_reduction_mul_float_array_1d(
! CHECK-SAME: %{{.*}}: !fir.ref<!fir.array<100xf32>> {fir.bindc_name = "a"}, %[[B:.*]]: !fir.ref<!fir.array<100xf32>> {fir.bindc_name = "b"})
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
-! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<!fir.array<100xf32>>) bounds(%2) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
+! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<!fir.array<100xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<100xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_mul_section_ext100_ref_100xf32 -> %[[RED_B]] : !fir.ref<!fir.array<100xf32>>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_mul_section_ext100_ref_100xf32 -> %[[RED_B]] : !fir.ref<!fir.array<100xf32>>)
subroutine acc_reduction_min_int(a, b)
integer :: a(100)
@@ -884,7 +882,7 @@ subroutine acc_reduction_min_int(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_min_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_min_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
subroutine acc_reduction_min_int_array_1d(a, b)
integer :: a(100), b(100)
@@ -899,9 +897,9 @@ subroutine acc_reduction_min_int_array_1d(a, b)
! CHECK-LABEL: func.func @_QPacc_reduction_min_int_array_1d(
! CHECK-SAME: %{{.*}}: !fir.ref<!fir.array<100xi32>> {fir.bindc_name = "a"}, %[[ARG1:.*]]: !fir.ref<!fir.array<100xi32>> {fir.bindc_name = "b"})
! HLFIR: %[[DECLARG1:.*]]:2 = hlfir.declare %[[ARG1]]
-! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100xi32>>) bounds(%2) -> !fir.ref<!fir.array<100xi32>> {name = "b"}
+! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100xi32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xi32>> {name = "b"}
! HLFIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[DECLARG1]]#1 : !fir.ref<!fir.array<100xi32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xi32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_min_section_ext100_ref_100xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100xi32>>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_min_section_ext100_ref_100xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100xi32>>)
subroutine acc_reduction_min_float(a, b)
real :: a(100), b
@@ -918,7 +916,7 @@ subroutine acc_reduction_min_float(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_min_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_min_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
subroutine acc_reduction_min_float_array2d(a, b)
real :: a(100, 10), b(100, 10)
@@ -935,10 +933,9 @@ subroutine acc_reduction_min_float_array2d(a, b)
! CHECK-LABEL: func.func @_QPacc_reduction_min_float_array2d(
! CHECK-SAME: %{{.*}}: !fir.ref<!fir.array<100x10xf32>> {fir.bindc_name = "a"}, %[[ARG1:.*]]: !fir.ref<!fir.array<100x10xf32>> {fir.bindc_name = "b"})
! HLFIR: %[[DECLARG1:.*]]:2 = hlfir.declare %[[ARG1]]
-! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100x10xf32>>) bounds(%3, %5) -> !fir.ref<!fir.array<100x10xf32>> {name = "b"}
+! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100x10xf32>>) bounds(%{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10xf32>> {name = "b"}
! HLFIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[DECLARG1]]#1 : !fir.ref<!fir.array<100x10xf32>>) bounds(%{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10xf32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_min_section_ext100xext10_ref_100x10xf32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10xf32>>)
-! CHECK: attributes {collapse = 2 : i64}
+! CHECK: acc.loop {{.*}} reduction(@reduction_min_section_ext100xext10_ref_100x10xf32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10xf32>>)
subroutine acc_reduction_max_int(a, b)
integer :: a(100)
@@ -955,7 +952,7 @@ subroutine acc_reduction_max_int(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<i32>) -> !fir.ref<i32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_max_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_max_ref_i32 -> %[[RED_B]] : !fir.ref<i32>)
subroutine acc_reduction_max_int_array2d(a, b)
integer :: a(100, 10), b(100, 10)
@@ -974,7 +971,7 @@ subroutine acc_reduction_max_int_array2d(a, b)
! HLFIR: %[[DECLARG1:.*]]:2 = hlfir.declare %[[ARG1]]
! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100x10xi32>>) bounds(%{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10xi32>> {name = "b"}
! HLFIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[DECLARG1]]#1 : !fir.ref<!fir.array<100x10xi32>>) bounds(%{{.*}}, %{{.*}}) -> !fir.ref<!fir.array<100x10xi32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_max_section_ext100xext10_ref_100x10xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10xi32>>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_max_section_ext100xext10_ref_100x10xi32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100x10xi32>>)
subroutine acc_reduction_max_float(a, b)
real :: a(100), b
@@ -991,7 +988,7 @@ subroutine acc_reduction_max_float(a, b)
! HLFIR: %[[DECLB:.*]]:2 = hlfir.declare %[[B]]
! FIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[B]] : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
! HLFIR: %[[RED_B:.*]] = acc.reduction varPtr(%[[DECLB]]#1 : !fir.ref<f32>) -> !fir.ref<f32> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_max_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
+! CHECK: acc.loop {{.*}} reduction(@reduction_max_ref_f32 -> %[[RED_B]] : !fir.ref<f32>)
subroutine acc_reduction_max_float_array1d(a, b)
real :: a(100), b(100)
@@ -1008,7 +1005,7 @@ subroutine acc_reduction_max_float_array1d(a, b)
! HLFIR: %[[DECLARG1:.*]]:2 = hlfir.declare %[[ARG1]]
! FIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[ARG1]] : !fir.ref<!fir.array<100xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
! HLFIR: %[[RED_ARG1:.*]] = acc.reduction varPtr(%[[DECLARG1]]#1 : !fir.ref<!fir.array<100xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<100xf32>> {name = "b"}
-! CHECK: acc.loop reduction(@reduction_max_section_ext100_ref_100xf32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100xf32>>) {
+! CHECK: acc.loop {{.*}} reduction(@reduction_max_section_ext100_ref_100xf32 -> %[[RED_ARG1]] : !fir.ref<!fir.array<100xf32>>)
subroutine acc_reduction_iand()
integer :: i
diff --git a/flang/test/Lower/OpenACC/acc-serial-loop.f90 b/flang/test/Lower/OpenACC/acc-serial-loop.f90
index 4ed7bb8da29a1a0..ca504f36c368353 100644
--- a/flang/test/Lower/OpenACC/acc-serial-loop.f90
+++ b/flang/test/Lower/OpenACC/acc-serial-loop.f90
@@ -64,8 +64,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -78,8 +77,7 @@ subroutine acc_serial_loop
!$acc end serial loop
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -92,8 +90,7 @@ subroutine acc_serial_loop
! CHECK: [[ASYNC1:%.*]] = arith.constant 1 : i32
! CHECK: acc.serial async([[ASYNC1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -106,8 +103,7 @@ subroutine acc_serial_loop
! CHECK: [[ASYNC2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.serial async([[ASYNC2]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -119,8 +115,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -133,8 +128,7 @@ subroutine acc_serial_loop
! CHECK: [[WAIT1:%.*]] = arith.constant 1 : i32
! CHECK: acc.serial wait([[WAIT1]] : i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -148,8 +142,7 @@ subroutine acc_serial_loop
! CHECK: [[WAIT2:%.*]] = arith.constant 1 : i32
! CHECK: [[WAIT3:%.*]] = arith.constant 2 : i32
! CHECK: acc.serial wait([[WAIT2]], [[WAIT3]] : i32, i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -163,8 +156,7 @@ subroutine acc_serial_loop
! CHECK: [[WAIT4:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: [[WAIT5:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.serial wait([[WAIT4]], [[WAIT5]] : i32, i32) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -177,8 +169,7 @@ subroutine acc_serial_loop
! CHECK: [[IF1:%.*]] = arith.constant true
! CHECK: acc.serial if([[IF1]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -192,8 +183,7 @@ subroutine acc_serial_loop
! CHECK: [[IFCOND:%.*]] = fir.load %{{.*}} : !fir.ref<!fir.logical<4>>
! CHECK: [[IF2:%.*]] = fir.convert [[IFCOND]] : (!fir.logical<4>) -> i1
! CHECK: acc.serial if([[IF2]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -206,8 +196,7 @@ subroutine acc_serial_loop
! CHECK: [[SELF1:%.*]] = arith.constant true
! CHECK: acc.serial self([[SELF1]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -219,8 +208,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -233,8 +221,7 @@ subroutine acc_serial_loop
! CHECK: %[[SELF2:.*]] = fir.convert %[[DECLIFCONDITION]]#1 : (!fir.ref<!fir.logical<4>>) -> i1
! CHECK: acc.serial self(%[[SELF2]]) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -248,8 +235,7 @@ subroutine acc_serial_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "b"}
! CHECK: acc.serial dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -265,8 +251,7 @@ subroutine acc_serial_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copy>, name = "b"}
! CHECK: acc.serial dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -282,8 +267,7 @@ subroutine acc_serial_loop
! CHECK: %[[COPYIN_A:.*]] = acc.copyin varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[COPYIN_B:.*]] = acc.copyin varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyin_readonly>, name = "b"}
! CHECK: acc.serial dataOperands(%[[COPYIN_A]], %[[COPYIN_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -297,8 +281,7 @@ subroutine acc_serial_loop
! CHECK: %[[CREATE_A:.*]] = acc.create varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyout>, name = "a"}
! CHECK: %[[CREATE_B:.*]] = acc.create varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_copyout>, name = "b"}
! CHECK: acc.serial dataOperands(%[[CREATE_A]], %[[CREATE_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -314,8 +297,7 @@ subroutine acc_serial_loop
! CHECK: %[[CREATE_B:.*]] = acc.create varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: %[[CREATE_A:.*]] = acc.create varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {dataClause = #acc<data_clause acc_create_zero>, name = "a"}
! CHECK: acc.serial dataOperands(%[[CREATE_B]], %[[CREATE_A]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -331,8 +313,7 @@ subroutine acc_serial_loop
! CHECK: %[[NOCREATE_A:.*]] = acc.nocreate varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[NOCREATE_B:.*]] = acc.nocreate varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.serial dataOperands(%[[NOCREATE_A]], %[[NOCREATE_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -346,8 +327,7 @@ subroutine acc_serial_loop
! CHECK: %[[PRESENT_A:.*]] = acc.present varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[PRESENT_B:.*]] = acc.present varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.serial dataOperands(%[[PRESENT_A]], %[[PRESENT_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -361,8 +341,7 @@ subroutine acc_serial_loop
! CHECK: %[[DEVICEPTR_A:.*]] = acc.deviceptr varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: %[[DEVICEPTR_B:.*]] = acc.deviceptr varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.serial dataOperands(%[[DEVICEPTR_A]], %[[DEVICEPTR_B]] : !fir.ref<!fir.array<10xf32>>, !fir.ref<!fir.array<10xf32>>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -380,8 +359,7 @@ subroutine acc_serial_loop
! CHECK: %[[BOX_ADDR_G:.*]] = fir.box_addr %[[BOX_G]] : (!fir.box<!fir.ptr<f32>>) -> !fir.ptr<f32>
! CHECK: %[[ATTACH_G:.*]] = acc.attach varPtr(%[[BOX_ADDR_G]] : !fir.ptr<f32>) -> !fir.ptr<f32> {name = "g"}
! CHECK: acc.serial dataOperands(%[[ATTACH_F]], %[[ATTACH_G]] : !fir.ptr<f32>, !fir.ptr<f32>) {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -395,7 +373,7 @@ subroutine acc_serial_loop
! CHECK: %[[ACC_FPRIVATE_B:.*]] = acc.firstprivate varPtr(%[[DECLB]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "b"}
! CHECK: acc.serial firstprivate(@firstprivatization_section_ext10_ref_10xf32 -> %[[ACC_FPRIVATE_B]] : !fir.ref<!fir.array<10xf32>>) {
! CHECK: %[[ACC_PRIVATE_A:.*]] = acc.private varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
-! CHECK: acc.loop private(@privatization_ref_10xf32 -> %[[ACC_PRIVATE_A]] : !fir.ref<!fir.array<10xf32>>) {
+! CHECK: acc.loop private({{.*}}@privatization_ref_10xf32 -> %[[ACC_PRIVATE_A]] : !fir.ref<!fir.array<10xf32>>)
! CHECK: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -408,10 +386,9 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {seq}
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, seq}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -421,10 +398,9 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {auto}
+! CHECK-NEXT: } attributes {auto, inclusiveUpperbound = array<i1: true>}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -434,10 +410,9 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {independent}
+! CHECK-NEXT: } attributes {inclusiveUpperbound = array<i1: true>, independent}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -447,8 +422,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop gang {
-! CHECK: fir.do_loop
+! CHECK: acc.loop gang() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -461,8 +435,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[GANGNUM1:%.*]] = arith.constant 8 : i32
-! CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) {
-! CHECK: fir.do_loop
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM1]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -475,8 +448,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[GANGNUM2:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-! CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) {
-! CHECK: fir.do_loop
+! CHECK-NEXT: acc.loop gang(num=[[GANGNUM2]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -488,8 +460,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop gang(num=%{{.*}} : i32, static=%{{.*}} : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop gang(num=%{{.*}} : i32, static=%{{.*}} : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -500,8 +471,7 @@ subroutine acc_serial_loop
a(i) = b(i)
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop vector {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -514,8 +484,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[CONSTANT128:%.*]] = arith.constant 128 : i32
-! CHECK: acc.loop vector([[CONSTANT128]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector([[CONSTANT128]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -528,8 +497,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[VECTORLENGTH:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
-! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -541,8 +509,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop worker {
-! CHECK: fir.do_loop
+! CHECK: acc.loop worker() {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -555,8 +522,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[WORKER128:%.*]] = arith.constant 128 : i32
-! CHECK: acc.loop worker([[WORKER128]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop worker([[WORKER128]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -570,11 +536,10 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
! CHECK: fir.do_loop
! CHECK: acc.yield
-! CHECK-NEXT: } attributes {collapse = 2 : i64}
+! CHECK-NEXT: } attributes {collapse = 2 : i64, inclusiveUpperbound = array<i1: true, true>}
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
@@ -587,10 +552,8 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
-! CHECK: acc.loop {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} {
+! CHECK: acc.loop {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -605,8 +568,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[TILESIZE:%.*]] = arith.constant 2 : i32
-! CHECK: acc.loop tile([[TILESIZE]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZE]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -619,8 +581,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[TILESIZEM1:%.*]] = arith.constant -1 : i32
-! CHECK: acc.loop tile([[TILESIZEM1]] : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZEM1]] : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -636,8 +597,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: [[TILESIZE1:%.*]] = arith.constant 2 : i32
! CHECK: [[TILESIZE2:%.*]] = arith.constant 2 : i32
-! CHECK: acc.loop tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile([[TILESIZE1]], [[TILESIZE2]] : i32, i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -649,8 +609,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop tile(%{{.*}} : i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile(%{{.*}} : i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -664,8 +623,7 @@ subroutine acc_serial_loop
END DO
! CHECK: acc.serial {
-! CHECK: acc.loop tile(%{{.*}}, %{{.*}} : i32, i32) {
-! CHECK: fir.do_loop
+! CHECK: acc.loop {{.*}} tile(%{{.*}}, %{{.*}} : i32, i32) {{.*}} {
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -680,7 +638,7 @@ subroutine acc_serial_loop
! CHECK: %[[COPYINREDR:.*]] = acc.copyin varPtr(%{{.*}} : !fir.ref<f32>) -> !fir.ref<f32> {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_r"}
! CHECK: %[[COPYINREDI:.*]] = acc.copyin varPtr(%{{.*}} : !fir.ref<i32>) -> !fir.ref<i32> {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_i"}
! CHECK: acc.serial dataOperands(%[[COPYINREDR]], %[[COPYINREDI]] : !fir.ref<f32>, !fir.ref<i32>) {
-! CHECK: acc.loop reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>) {
+! CHECK: acc.loop {{.*}} reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>)
! CHECK: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
>From b0fff85a5d6950c8048b3b962caa470acccdb4f1 Mon Sep 17 00:00:00 2001
From: Valentin Clement <clementval at gmail.com>
Date: Wed, 29 Nov 2023 14:09:48 -0800
Subject: [PATCH 3/4] Fix lowering of combined construct
---
flang/lib/Lower/Bridge.cpp | 32 +++++++++++++------
.../test/Lower/OpenACC/acc-parallel-loop.f90 | 10 +++---
flang/test/Lower/OpenACC/acc-serial-loop.f90 | 6 ++--
3 files changed, 31 insertions(+), 17 deletions(-)
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index fcb5f275c8ffa02..d7e4c40ae454a53 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -2386,18 +2386,32 @@ class FirConverter : public Fortran::lower::AbstractConverter {
const Fortran::parser::OpenACCLoopConstruct *accLoop =
std::get_if<Fortran::parser::OpenACCLoopConstruct>(&acc.u);
+ const Fortran::parser::OpenACCCombinedConstruct *accCombined =
+ std::get_if<Fortran::parser::OpenACCCombinedConstruct>(&acc.u);
Fortran::lower::pft::Evaluation *curEval = &getEval();
- if (accLoop) {
- const Fortran::parser::AccBeginLoopDirective &beginLoopDir =
- std::get<Fortran::parser::AccBeginLoopDirective>(accLoop->t);
- const Fortran::parser::AccClauseList &clauseList =
- std::get<Fortran::parser::AccClauseList>(beginLoopDir.t);
- int64_t collapseValue = Fortran::lower::getCollapseValue(clauseList);
- curEval = &curEval->getFirstNestedEvaluation();
- for (int64_t i = 1; i < collapseValue; i++)
- curEval = &*std::next(curEval->getNestedEvaluations().begin());
+ if (accLoop || accCombined) {
+ int64_t collapseValue;
+ if (accLoop) {
+ const Fortran::parser::AccBeginLoopDirective &beginLoopDir =
+ std::get<Fortran::parser::AccBeginLoopDirective>(accLoop->t);
+ const Fortran::parser::AccClauseList &clauseList =
+ std::get<Fortran::parser::AccClauseList>(beginLoopDir.t);
+ collapseValue = Fortran::lower::getCollapseValue(clauseList);
+ } else if (accCombined) {
+ const Fortran::parser::AccBeginCombinedDirective &beginCombinedDir =
+ std::get<Fortran::parser::AccBeginCombinedDirective>(accCombined->t);
+ const Fortran::parser::AccClauseList &clauseList =
+ std::get<Fortran::parser::AccClauseList>(beginCombinedDir.t);
+ collapseValue = Fortran::lower::getCollapseValue(clauseList);
+ }
+
+ if (curEval->lowerAsStructured()) {
+ curEval = &curEval->getFirstNestedEvaluation();
+ for (int64_t i = 1; i < collapseValue; i++)
+ curEval = &*std::next(curEval->getNestedEvaluations().begin());
+ }
}
for (Fortran::lower::pft::Evaluation &e : curEval->getNestedEvaluations())
diff --git a/flang/test/Lower/OpenACC/acc-parallel-loop.f90 b/flang/test/Lower/OpenACC/acc-parallel-loop.f90
index 712e00918169534..ef9a74260a07a6b 100644
--- a/flang/test/Lower/OpenACC/acc-parallel-loop.f90
+++ b/flang/test/Lower/OpenACC/acc-parallel-loop.f90
@@ -433,7 +433,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel firstprivate(@firstprivatization_section_ext10_ref_10xf32 -> %[[ACC_PRIVATE_B]] : !fir.ref<!fir.array<10xf32>>) {
! CHECK: %[[ACC_PRIVATE_A:.*]] = acc.private varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: acc.loop private({{.*}}@privatization_ref_10xf32 -> %[[ACC_PRIVATE_A]] : !fir.ref<!fir.array<10xf32>>)
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -557,7 +557,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[VECTORLENGTH:%.*]] = fir.load %{{.*}} : !fir.ref<i32>
! CHECK: acc.loop vector([[VECTORLENGTH]] : i32) {{.*}} {
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -570,7 +570,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: acc.loop worker() {{.*}} {
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -584,7 +584,7 @@ subroutine acc_parallel_loop
! CHECK: acc.parallel {
! CHECK: [[WORKER128:%.*]] = arith.constant 128 : i32
! CHECK: acc.loop worker([[WORKER128]] : i32) {{.*}} {
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -713,6 +713,6 @@ subroutine acc_parallel_loop
10 continue
! CHECK: acc.parallel
! CHECK: acc.loop
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
end subroutine acc_parallel_loop
diff --git a/flang/test/Lower/OpenACC/acc-serial-loop.f90 b/flang/test/Lower/OpenACC/acc-serial-loop.f90
index ca504f36c368353..b5028c82380d27d 100644
--- a/flang/test/Lower/OpenACC/acc-serial-loop.f90
+++ b/flang/test/Lower/OpenACC/acc-serial-loop.f90
@@ -374,7 +374,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial firstprivate(@firstprivatization_section_ext10_ref_10xf32 -> %[[ACC_FPRIVATE_B]] : !fir.ref<!fir.array<10xf32>>) {
! CHECK: %[[ACC_PRIVATE_A:.*]] = acc.private varPtr(%[[DECLA]]#1 : !fir.ref<!fir.array<10xf32>>) bounds(%{{.*}}) -> !fir.ref<!fir.array<10xf32>> {name = "a"}
! CHECK: acc.loop private({{.*}}@privatization_ref_10xf32 -> %[[ACC_PRIVATE_A]] : !fir.ref<!fir.array<10xf32>>)
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
@@ -537,7 +537,7 @@ subroutine acc_serial_loop
! CHECK: acc.serial {
! CHECK: acc.loop {{.*}} {
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: } attributes {collapse = 2 : i64, inclusiveUpperbound = array<i1: true, true>}
! CHECK: acc.yield
@@ -639,7 +639,7 @@ subroutine acc_serial_loop
! CHECK: %[[COPYINREDI:.*]] = acc.copyin varPtr(%{{.*}} : !fir.ref<i32>) -> !fir.ref<i32> {dataClause = #acc<data_clause acc_reduction>, implicit = true, name = "reduction_i"}
! CHECK: acc.serial dataOperands(%[[COPYINREDR]], %[[COPYINREDI]] : !fir.ref<f32>, !fir.ref<i32>) {
! CHECK: acc.loop {{.*}} reduction(@reduction_add_ref_f32 -> %{{.*}} : !fir.ref<f32>, @reduction_mul_ref_i32 -> %{{.*}} : !fir.ref<i32>)
-! CHECK: fir.do_loop
+! CHECK-NOT: fir.do_loop
! CHECK: acc.yield
! CHECK-NEXT: }{{$}}
! CHECK: acc.yield
>From 13b61f877314b1a7134827985c8946c942a47f99 Mon Sep 17 00:00:00 2001
From: Valentin Clement <clementval at gmail.com>
Date: Wed, 29 Nov 2023 14:10:46 -0800
Subject: [PATCH 4/4] clang-format
---
flang/lib/Lower/Bridge.cpp | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/flang/lib/Lower/Bridge.cpp b/flang/lib/Lower/Bridge.cpp
index d7e4c40ae454a53..78a5ac75a74c4fd 100644
--- a/flang/lib/Lower/Bridge.cpp
+++ b/flang/lib/Lower/Bridge.cpp
@@ -2401,7 +2401,8 @@ class FirConverter : public Fortran::lower::AbstractConverter {
collapseValue = Fortran::lower::getCollapseValue(clauseList);
} else if (accCombined) {
const Fortran::parser::AccBeginCombinedDirective &beginCombinedDir =
- std::get<Fortran::parser::AccBeginCombinedDirective>(accCombined->t);
+ std::get<Fortran::parser::AccBeginCombinedDirective>(
+ accCombined->t);
const Fortran::parser::AccClauseList &clauseList =
std::get<Fortran::parser::AccClauseList>(beginCombinedDir.t);
collapseValue = Fortran::lower::getCollapseValue(clauseList);
More information about the flang-commits
mailing list