[flang-commits] [flang] [Flang][OpenMP] Add trivial conversion pattern for omp.ordered_region (PR #66085)

[Kiran Chandramohan via flang-commits]

https://github.com/kiranchandramohan updated https://github.com/llvm/llvm-project/pull/66085

>From 5f9174b2a52091d806d0003fc11e97ab47c06c0f Mon Sep 17 00:00:00 2001
From: Kiran Chandramohan <kiran.chandramohan at arm.com>
Date: Tue, 12 Sep 2023 12:48:50 +0000
Subject: [PATCH] [Flang][OpenMP] Add trivial conversion pattern for
 omp.ordered_region

Fixes #65570
---
 .../Fir/convert-to-llvm-openmp-and-fir.fir    | 33 +++++++++++++++++++
 .../Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp  |  9 ++---
 .../OpenMPToLLVM/convert-to-llvmir.mlir       | 31 +++++++++++++++++
 3 files changed, 69 insertions(+), 4 deletions(-)

diff --git a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
index 8134fb8792d76bd..28b89838a68046f 100644
--- a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
+++ b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
@@ -657,6 +657,7 @@ func.func @_QPs(%arg0: !fir.ref<!fir.complex<4>> {fir.bindc_name = "x"}) {
 }
 
 // -----
+
 // Test if llvm.alloca is properly inserted in the omp section
 
 //CHECK:  %[[CONST:.*]] = llvm.mlir.constant(1 : i64) : i64
@@ -693,3 +694,35 @@ fir.global internal @_QFEx target : i32 {
   %0 = fir.zero_bits i32
   fir.has_value %0 : i32
 }
+
+// -----
+
+// Test if llvm.alloca is properly inserted in the omp ordered region
+
+// CHECK: llvm.func @sub_
+func.func @sub_() {
+  %c0 = arith.constant 0 : index
+  %c1 = arith.constant 1 : index
+  %0 = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFsubEi"}
+// CHECK: omp.ordered_region
+  omp.ordered_region {
+    %1 = fir.convert %c1 : (index) -> i32
+    cf.br ^bb1(%1, %c1 : i32, index)
+  ^bb1(%2: i32, %3: index):  // 2 preds: ^bb0, ^bb2
+    %4 = arith.cmpi sgt, %3, %c0 : index
+    cf.cond_br %4, ^bb2, ^bb3
+  ^bb2:  // pred: ^bb1
+    fir.store %2 to %0 : !fir.ref<i32>
+    %5 = fir.load %0 : !fir.ref<i32>
+// CHECK: llvm.add
+    %6 = arith.addi %5, %1 : i32
+// CHECK: llvm.sub
+    %7 = arith.subi %3, %c1 : index
+    cf.br ^bb1(%6, %7 : i32, index)
+  ^bb3:  // pred: ^bb1
+    fir.store %2 to %0 : !fir.ref<i32>
+// CHECK: omp.terminator
+    omp.terminator
+  }
+  return
+} 
diff --git a/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp b/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
index d06b7033257196a..adcbbc3f0abb249 100644
--- a/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
+++ b/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
@@ -199,10 +199,10 @@ void mlir::configureOpenMPToLLVMConversionLegality(
     ConversionTarget &target, LLVMTypeConverter &typeConverter) {
   target.addDynamicallyLegalOp<
       mlir::omp::AtomicUpdateOp, mlir::omp::CriticalOp, mlir::omp::TargetOp,
-      mlir::omp::DataOp, mlir::omp::ParallelOp, mlir::omp::WsLoopOp,
-      mlir::omp::SimdLoopOp, mlir::omp::MasterOp, mlir::omp::SectionOp,
-      mlir::omp::SectionsOp, mlir::omp::SingleOp, mlir::omp::TaskGroupOp,
-      mlir::omp::TaskOp>([&](Operation *op) {
+      mlir::omp::DataOp, mlir::omp::OrderedRegionOp, mlir::omp::ParallelOp,
+      mlir::omp::WsLoopOp, mlir::omp::SimdLoopOp, mlir::omp::MasterOp,
+      mlir::omp::SectionOp, mlir::omp::SectionsOp, mlir::omp::SingleOp,
+      mlir::omp::TaskGroupOp, mlir::omp::TaskOp>([&](Operation *op) {
     return typeConverter.isLegal(&op->getRegion(0)) &&
            typeConverter.isLegal(op->getOperandTypes()) &&
            typeConverter.isLegal(op->getResultTypes());
@@ -234,6 +234,7 @@ void mlir::populateOpenMPToLLVMConversionPatterns(LLVMTypeConverter &converter,
       AtomicReadOpConversion, ReductionOpConversion,
       ReductionDeclareOpConversion, RegionOpConversion<omp::CriticalOp>,
       RegionOpConversion<omp::MasterOp>, ReductionOpConversion,
+      RegionOpConversion<omp::OrderedRegionOp>,
       RegionOpConversion<omp::ParallelOp>, RegionOpConversion<omp::WsLoopOp>,
       RegionOpConversion<omp::SectionsOp>, RegionOpConversion<omp::SectionOp>,
       RegionOpConversion<omp::SimdLoopOp>, RegionOpConversion<omp::SingleOp>,
diff --git a/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir b/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
index b83b122f75e4b08..fedbcd401d44c94 100644
--- a/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
+++ b/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
@@ -384,3 +384,34 @@ llvm.func @_QQmain() {
 llvm.func @_QFPdo_work(%arg0: !llvm.ptr<i32> {fir.bindc_name = "i"}) {
   llvm.return
 }
+
+// -----
+
+// CHECK-LABEL:  @sub_
+llvm.func @sub_() {
+  %0 = llvm.mlir.constant(0 : index) : i64
+  %1 = llvm.mlir.constant(1 : index) : i64
+  %2 = llvm.mlir.constant(1 : i64) : i64
+  %3 = llvm.alloca %2 x i32 {bindc_name = "i", in_type = i32, operandSegmentSizes = array<i32: 0, 0>, uniq_name = "_QFsubEi"} : (i64) -> !llvm.ptr<i32>
+// CHECK: omp.ordered_region
+  omp.ordered_region {
+    %4 = llvm.trunc %1 : i64 to i32
+    llvm.br ^bb1(%4, %1 : i32, i64)
+  ^bb1(%5: i32, %6: i64):  // 2 preds: ^bb0, ^bb2
+    %7 = llvm.icmp "sgt" %6, %0 : i64
+    llvm.cond_br %7, ^bb2, ^bb3
+  ^bb2:  // pred: ^bb1
+    llvm.store %5, %3 : !llvm.ptr<i32>
+    %8 = llvm.load %3 : !llvm.ptr<i32>
+// CHECK: llvm.add
+    %9 = arith.addi %8, %4 : i32
+// CHECK: llvm.sub
+    %10 = arith.subi %6, %1 : i64
+    llvm.br ^bb1(%9, %10 : i32, i64)
+  ^bb3:  // pred: ^bb1
+    llvm.store %5, %3 : !llvm.ptr<i32>
+// CHECK: omp.terminator
+    omp.terminator
+  }
+  llvm.return
+}