[llvm-branch-commits] [flang] [WIP][flang] Introduce HLFIR lowerings to omp.workshare_loop_nest (PR #104748)
Ivan R. Ivanov via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Mon Sep 23 00:04:32 PDT 2024
https://github.com/ivanradanov updated https://github.com/llvm/llvm-project/pull/104748
>From b710a580ab1732b3b41f0e3fb0684b45108d2c09 Mon Sep 17 00:00:00 2001
From: Ivan Radanov Ivanov <ivanov.i.aa at m.titech.ac.jp>
Date: Mon, 23 Sep 2024 16:04:07 +0900
Subject: [PATCH 1/3] clang-format
---
flang/include/flang/Tools/CLOptions.inc | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/flang/include/flang/Tools/CLOptions.inc b/flang/include/flang/Tools/CLOptions.inc
index bb00e079008a0b..81ce69b4ec7e1b 100644
--- a/flang/include/flang/Tools/CLOptions.inc
+++ b/flang/include/flang/Tools/CLOptions.inc
@@ -336,8 +336,8 @@ inline void createDefaultFIROptimizerPassPipeline(
/// \param pm - MLIR pass manager that will hold the pipeline definition
/// \param optLevel - optimization level used for creating FIR optimization
/// passes pipeline
-inline void createHLFIRToFIRPassPipeline(
- mlir::PassManager &pm, bool enableOpenMP, llvm::OptimizationLevel optLevel = defaultOptLevel) {
+inline void createHLFIRToFIRPassPipeline(mlir::PassManager &pm,
+ bool enableOpenMP, llvm::OptimizationLevel optLevel = defaultOptLevel) {
if (optLevel.isOptimizingForSpeed()) {
addCanonicalizerPassWithoutRegionSimplification(pm);
addNestedPassToAllTopLevelOperations(
>From 93ce40863c66da660e44f89fe8a984132077843c Mon Sep 17 00:00:00 2001
From: Ivan Radanov Ivanov <ivanov.i.aa at m.titech.ac.jp>
Date: Sun, 4 Aug 2024 17:33:52 +0900
Subject: [PATCH 2/3] Add workshare loop wrapper lowerings
Bufferize test
Bufferize test
Bufferize test
Add test for should use workshare lowering
---
.../HLFIR/Transforms/BufferizeHLFIR.cpp | 4 +-
.../Transforms/OptimizedBufferization.cpp | 10 +-
flang/test/HLFIR/bufferize-workshare.fir | 58 ++++++++
.../OpenMP/should-use-workshare-lowering.mlir | 140 ++++++++++++++++++
4 files changed, 208 insertions(+), 4 deletions(-)
create mode 100644 flang/test/HLFIR/bufferize-workshare.fir
create mode 100644 flang/test/Transforms/OpenMP/should-use-workshare-lowering.mlir
diff --git a/flang/lib/Optimizer/HLFIR/Transforms/BufferizeHLFIR.cpp b/flang/lib/Optimizer/HLFIR/Transforms/BufferizeHLFIR.cpp
index 07794828fce267..1848dbe2c7a2c2 100644
--- a/flang/lib/Optimizer/HLFIR/Transforms/BufferizeHLFIR.cpp
+++ b/flang/lib/Optimizer/HLFIR/Transforms/BufferizeHLFIR.cpp
@@ -26,6 +26,7 @@
#include "flang/Optimizer/HLFIR/HLFIRDialect.h"
#include "flang/Optimizer/HLFIR/HLFIROps.h"
#include "flang/Optimizer/HLFIR/Passes.h"
+#include "flang/Optimizer/OpenMP/Passes.h"
#include "mlir/Dialect/OpenMP/OpenMPDialect.h"
#include "mlir/IR/Dominance.h"
#include "mlir/IR/PatternMatch.h"
@@ -792,7 +793,8 @@ struct ElementalOpConversion
// Generate a loop nest looping around the fir.elemental shape and clone
// fir.elemental region inside the inner loop.
hlfir::LoopNest loopNest =
- hlfir::genLoopNest(loc, builder, extents, !elemental.isOrdered());
+ hlfir::genLoopNest(loc, builder, extents, !elemental.isOrdered(),
+ flangomp::shouldUseWorkshareLowering(elemental));
auto insPt = builder.saveInsertionPoint();
builder.setInsertionPointToStart(loopNest.body);
auto yield = hlfir::inlineElementalOp(loc, builder, elemental,
diff --git a/flang/lib/Optimizer/HLFIR/Transforms/OptimizedBufferization.cpp b/flang/lib/Optimizer/HLFIR/Transforms/OptimizedBufferization.cpp
index 3a0a98dc594463..f014724861e333 100644
--- a/flang/lib/Optimizer/HLFIR/Transforms/OptimizedBufferization.cpp
+++ b/flang/lib/Optimizer/HLFIR/Transforms/OptimizedBufferization.cpp
@@ -20,6 +20,7 @@
#include "flang/Optimizer/HLFIR/HLFIRDialect.h"
#include "flang/Optimizer/HLFIR/HLFIROps.h"
#include "flang/Optimizer/HLFIR/Passes.h"
+#include "flang/Optimizer/OpenMP/Passes.h"
#include "flang/Optimizer/Transforms/Utils.h"
#include "mlir/Dialect/Func/IR/FuncOps.h"
#include "mlir/IR/Dominance.h"
@@ -482,7 +483,8 @@ llvm::LogicalResult ElementalAssignBufferization::matchAndRewrite(
// Generate a loop nest looping around the hlfir.elemental shape and clone
// hlfir.elemental region inside the inner loop
hlfir::LoopNest loopNest =
- hlfir::genLoopNest(loc, builder, extents, !elemental.isOrdered());
+ hlfir::genLoopNest(loc, builder, extents, !elemental.isOrdered(),
+ flangomp::shouldUseWorkshareLowering(elemental));
builder.setInsertionPointToStart(loopNest.body);
auto yield = hlfir::inlineElementalOp(loc, builder, elemental,
loopNest.oneBasedIndices);
@@ -553,7 +555,8 @@ llvm::LogicalResult BroadcastAssignBufferization::matchAndRewrite(
llvm::SmallVector<mlir::Value> extents =
hlfir::getIndexExtents(loc, builder, shape);
hlfir::LoopNest loopNest =
- hlfir::genLoopNest(loc, builder, extents, /*isUnordered=*/true);
+ hlfir::genLoopNest(loc, builder, extents, /*isUnordered=*/true,
+ flangomp::shouldUseWorkshareLowering(assign));
builder.setInsertionPointToStart(loopNest.body);
auto arrayElement =
hlfir::getElementAt(loc, builder, lhs, loopNest.oneBasedIndices);
@@ -648,7 +651,8 @@ llvm::LogicalResult VariableAssignBufferization::matchAndRewrite(
llvm::SmallVector<mlir::Value> extents =
hlfir::getIndexExtents(loc, builder, shape);
hlfir::LoopNest loopNest =
- hlfir::genLoopNest(loc, builder, extents, /*isUnordered=*/true);
+ hlfir::genLoopNest(loc, builder, extents, /*isUnordered=*/true,
+ flangomp::shouldUseWorkshareLowering(assign));
builder.setInsertionPointToStart(loopNest.body);
auto rhsArrayElement =
hlfir::getElementAt(loc, builder, rhs, loopNest.oneBasedIndices);
diff --git a/flang/test/HLFIR/bufferize-workshare.fir b/flang/test/HLFIR/bufferize-workshare.fir
new file mode 100644
index 00000000000000..9b7341ae43398a
--- /dev/null
+++ b/flang/test/HLFIR/bufferize-workshare.fir
@@ -0,0 +1,58 @@
+// RUN: fir-opt --bufferize-hlfir %s | FileCheck %s
+
+// CHECK-LABEL: func.func @simple(
+// CHECK-SAME: %[[VAL_0:.*]]: !fir.ref<!fir.array<42xi32>>) {
+// CHECK: omp.parallel {
+// CHECK: omp.workshare {
+// CHECK: %[[VAL_1:.*]] = arith.constant 42 : index
+// CHECK: %[[VAL_2:.*]] = arith.constant 1 : i32
+// CHECK: %[[VAL_3:.*]] = fir.shape %[[VAL_1]] : (index) -> !fir.shape<1>
+// CHECK: %[[VAL_4:.*]]:2 = hlfir.declare %[[VAL_0]](%[[VAL_3]]) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+// CHECK: %[[VAL_5:.*]] = fir.allocmem !fir.array<42xi32> {bindc_name = ".tmp.array", uniq_name = ""}
+// CHECK: %[[VAL_6:.*]]:2 = hlfir.declare %[[VAL_5]](%[[VAL_3]]) {uniq_name = ".tmp.array"} : (!fir.heap<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.heap<!fir.array<42xi32>>, !fir.heap<!fir.array<42xi32>>)
+// CHECK: %[[VAL_7:.*]] = arith.constant true
+// CHECK: %[[VAL_8:.*]] = arith.constant 1 : index
+// CHECK: omp.workshare.loop_wrapper {
+// CHECK: omp.loop_nest (%[[VAL_9:.*]]) : index = (%[[VAL_8]]) to (%[[VAL_1]]) inclusive step (%[[VAL_8]]) {
+// CHECK: %[[VAL_10:.*]] = hlfir.designate %[[VAL_4]]#0 (%[[VAL_9]]) : (!fir.ref<!fir.array<42xi32>>, index) -> !fir.ref<i32>
+// CHECK: %[[VAL_11:.*]] = fir.load %[[VAL_10]] : !fir.ref<i32>
+// CHECK: %[[VAL_12:.*]] = arith.subi %[[VAL_11]], %[[VAL_2]] : i32
+// CHECK: %[[VAL_13:.*]] = hlfir.designate %[[VAL_6]]#0 (%[[VAL_9]]) : (!fir.heap<!fir.array<42xi32>>, index) -> !fir.ref<i32>
+// CHECK: hlfir.assign %[[VAL_12]] to %[[VAL_13]] temporary_lhs : i32, !fir.ref<i32>
+// CHECK: omp.yield
+// CHECK: }
+// CHECK: omp.terminator
+// CHECK: }
+// CHECK: %[[VAL_14:.*]] = fir.undefined tuple<!fir.heap<!fir.array<42xi32>>, i1>
+// CHECK: %[[VAL_15:.*]] = fir.insert_value %[[VAL_14]], %[[VAL_7]], [1 : index] : (tuple<!fir.heap<!fir.array<42xi32>>, i1>, i1) -> tuple<!fir.heap<!fir.array<42xi32>>, i1>
+// CHECK: %[[VAL_16:.*]] = fir.insert_value %[[VAL_15]], %[[VAL_6]]#0, [0 : index] : (tuple<!fir.heap<!fir.array<42xi32>>, i1>, !fir.heap<!fir.array<42xi32>>) -> tuple<!fir.heap<!fir.array<42xi32>>, i1>
+// CHECK: hlfir.assign %[[VAL_6]]#0 to %[[VAL_4]]#0 : !fir.heap<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>
+// CHECK: fir.freemem %[[VAL_6]]#0 : !fir.heap<!fir.array<42xi32>>
+// CHECK: omp.terminator
+// CHECK: }
+// CHECK: omp.terminator
+// CHECK: }
+// CHECK: return
+// CHECK: }
+func.func @simple(%arg: !fir.ref<!fir.array<42xi32>>) {
+ omp.parallel {
+ omp.workshare {
+ %c42 = arith.constant 42 : index
+ %c1_i32 = arith.constant 1 : i32
+ %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+ %array:2 = hlfir.declare %arg(%shape) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+ %elemental = hlfir.elemental %shape unordered : (!fir.shape<1>) -> !hlfir.expr<42xi32> {
+ ^bb0(%i: index):
+ %ref = hlfir.designate %array#0 (%i) : (!fir.ref<!fir.array<42xi32>>, index) -> !fir.ref<i32>
+ %val = fir.load %ref : !fir.ref<i32>
+ %sub = arith.subi %val, %c1_i32 : i32
+ hlfir.yield_element %sub : i32
+ }
+ hlfir.assign %elemental to %array#0 : !hlfir.expr<42xi32>, !fir.ref<!fir.array<42xi32>>
+ hlfir.destroy %elemental : !hlfir.expr<42xi32>
+ omp.terminator
+ }
+ omp.terminator
+ }
+ return
+}
diff --git a/flang/test/Transforms/OpenMP/should-use-workshare-lowering.mlir b/flang/test/Transforms/OpenMP/should-use-workshare-lowering.mlir
new file mode 100644
index 00000000000000..229fe592a02b9b
--- /dev/null
+++ b/flang/test/Transforms/OpenMP/should-use-workshare-lowering.mlir
@@ -0,0 +1,140 @@
+// RUN: fir-opt --bufferize-hlfir %s | FileCheck %s
+
+// Checks that we correctly identify when to use the lowering to
+// omp.workshare.loop_wrapper
+
+// CHECK-LABEL: @should_parallelize_0
+// CHECK: omp.workshare.loop_wrapper
+func.func @should_parallelize_0(%arg: !fir.ref<!fir.array<42xi32>>, %idx : index) {
+ omp.workshare {
+ %c42 = arith.constant 42 : index
+ %c1_i32 = arith.constant 1 : i32
+ %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+ %array:2 = hlfir.declare %arg(%shape) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+ %elemental = hlfir.elemental %shape unordered : (!fir.shape<1>) -> !hlfir.expr<42xi32> {
+ ^bb0(%i: index):
+ hlfir.yield_element %c1_i32 : i32
+ }
+ hlfir.assign %elemental to %array#0 : !hlfir.expr<42xi32>, !fir.ref<!fir.array<42xi32>>
+ hlfir.destroy %elemental : !hlfir.expr<42xi32>
+ omp.terminator
+ }
+ return
+}
+
+// CHECK-LABEL: @should_parallelize_1
+// CHECK: omp.workshare.loop_wrapper
+func.func @should_parallelize_1(%arg: !fir.ref<!fir.array<42xi32>>, %idx : index) {
+ omp.parallel {
+ omp.workshare {
+ %c42 = arith.constant 42 : index
+ %c1_i32 = arith.constant 1 : i32
+ %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+ %array:2 = hlfir.declare %arg(%shape) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+ %elemental = hlfir.elemental %shape unordered : (!fir.shape<1>) -> !hlfir.expr<42xi32> {
+ ^bb0(%i: index):
+ hlfir.yield_element %c1_i32 : i32
+ }
+ hlfir.assign %elemental to %array#0 : !hlfir.expr<42xi32>, !fir.ref<!fir.array<42xi32>>
+ hlfir.destroy %elemental : !hlfir.expr<42xi32>
+ omp.terminator
+ }
+ omp.terminator
+ }
+ return
+}
+
+
+// CHECK-LABEL: @should_not_parallelize_0
+// CHECK-NOT: omp.workshare.loop_wrapper
+func.func @should_not_parallelize_0(%arg: !fir.ref<!fir.array<42xi32>>, %idx : index) {
+ omp.workshare {
+ omp.single {
+ %c42 = arith.constant 42 : index
+ %c1_i32 = arith.constant 1 : i32
+ %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+ %array:2 = hlfir.declare %arg(%shape) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+ %elemental = hlfir.elemental %shape unordered : (!fir.shape<1>) -> !hlfir.expr<42xi32> {
+ ^bb0(%i: index):
+ hlfir.yield_element %c1_i32 : i32
+ }
+ hlfir.assign %elemental to %array#0 : !hlfir.expr<42xi32>, !fir.ref<!fir.array<42xi32>>
+ hlfir.destroy %elemental : !hlfir.expr<42xi32>
+ omp.terminator
+ }
+ omp.terminator
+ }
+ return
+}
+
+// CHECK-LABEL: @should_not_parallelize_1
+// CHECK-NOT: omp.workshare.loop_wrapper
+func.func @should_not_parallelize_1(%arg: !fir.ref<!fir.array<42xi32>>, %idx : index) {
+ omp.workshare {
+ omp.critical {
+ %c42 = arith.constant 42 : index
+ %c1_i32 = arith.constant 1 : i32
+ %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+ %array:2 = hlfir.declare %arg(%shape) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+ %elemental = hlfir.elemental %shape unordered : (!fir.shape<1>) -> !hlfir.expr<42xi32> {
+ ^bb0(%i: index):
+ hlfir.yield_element %c1_i32 : i32
+ }
+ hlfir.assign %elemental to %array#0 : !hlfir.expr<42xi32>, !fir.ref<!fir.array<42xi32>>
+ hlfir.destroy %elemental : !hlfir.expr<42xi32>
+ omp.terminator
+ }
+ omp.terminator
+ }
+ return
+}
+
+// CHECK-LABEL: @should_not_parallelize_2
+// CHECK-NOT: omp.workshare.loop_wrapper
+func.func @should_not_parallelize_2(%arg: !fir.ref<!fir.array<42xi32>>, %idx : index) {
+ omp.workshare {
+ omp.parallel {
+ %c42 = arith.constant 42 : index
+ %c1_i32 = arith.constant 1 : i32
+ %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+ %array:2 = hlfir.declare %arg(%shape) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+ %elemental = hlfir.elemental %shape unordered : (!fir.shape<1>) -> !hlfir.expr<42xi32> {
+ ^bb0(%i: index):
+ hlfir.yield_element %c1_i32 : i32
+ }
+ hlfir.assign %elemental to %array#0 : !hlfir.expr<42xi32>, !fir.ref<!fir.array<42xi32>>
+ hlfir.destroy %elemental : !hlfir.expr<42xi32>
+ omp.terminator
+ }
+ omp.terminator
+ }
+ return
+}
+
+// CHECK-LABEL: @should_not_parallelize_3
+// CHECK-NOT: omp.workshare.loop_wrapper
+func.func @should_not_parallelize_3(%arg: !fir.ref<!fir.array<42xi32>>, %idx : index) {
+ omp.workshare {
+ omp.parallel {
+ omp.workshare {
+ omp.parallel {
+ %c42 = arith.constant 42 : index
+ %c1_i32 = arith.constant 1 : i32
+ %shape = fir.shape %c42 : (index) -> !fir.shape<1>
+ %array:2 = hlfir.declare %arg(%shape) {uniq_name = "array"} : (!fir.ref<!fir.array<42xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<42xi32>>, !fir.ref<!fir.array<42xi32>>)
+ %elemental = hlfir.elemental %shape unordered : (!fir.shape<1>) -> !hlfir.expr<42xi32> {
+ ^bb0(%i: index):
+ hlfir.yield_element %c1_i32 : i32
+ }
+ hlfir.assign %elemental to %array#0 : !hlfir.expr<42xi32>, !fir.ref<!fir.array<42xi32>>
+ hlfir.destroy %elemental : !hlfir.expr<42xi32>
+ omp.terminator
+ }
+ omp.terminator
+ }
+ omp.terminator
+ }
+ omp.terminator
+ }
+ return
+}
>From 7f69c407f8aa8090b72c83f3780a511f1530bc8c Mon Sep 17 00:00:00 2001
From: Ivan Radanov Ivanov <ivanov.i.aa at m.titech.ac.jp>
Date: Mon, 23 Sep 2024 12:56:11 +0900
Subject: [PATCH 3/3] Add integration test for workshare
---
flang/test/Integration/OpenMP/workshare.f90 | 57 +++++++++++++++++++++
1 file changed, 57 insertions(+)
create mode 100644 flang/test/Integration/OpenMP/workshare.f90
diff --git a/flang/test/Integration/OpenMP/workshare.f90 b/flang/test/Integration/OpenMP/workshare.f90
new file mode 100644
index 00000000000000..0c4524f8552906
--- /dev/null
+++ b/flang/test/Integration/OpenMP/workshare.f90
@@ -0,0 +1,57 @@
+!===----------------------------------------------------------------------===!
+! This directory can be used to add Integration tests involving multiple
+! stages of the compiler (for eg. from Fortran to LLVM IR). It should not
+! contain executable tests. We should only add tests here sparingly and only
+! if there is no other way to test. Repeat this message in each test that is
+! added to this directory and sub-directories.
+!===----------------------------------------------------------------------===!
+
+!RUN: %flang_fc1 -emit-hlfir -fopenmp -O3 %s -o - | FileCheck %s --check-prefix HLFIR
+!RUN: %flang_fc1 -emit-fir -fopenmp -O3 %s -o - | FileCheck %s --check-prefix FIR
+
+subroutine sb1(a, x, y, z)
+ integer :: a
+ integer :: x(:)
+ integer :: y(:)
+ integer :: z(:)
+ !$omp parallel workshare
+ z = a * x + y
+ !$omp end parallel workshare
+end subroutine
+
+! HLFIR: func.func @_QPsb1
+! HLFIR: omp.parallel {
+! HLFIR: omp.workshare {
+! HLFIR: hlfir.elemental {{.*}} unordered : (!fir.shape<1>) -> !hlfir.expr<?xi32> {
+! HLFIR: hlfir.elemental {{.*}} unordered : (!fir.shape<1>) -> !hlfir.expr<?xi32> {
+! HLFIR: hlfir.assign
+! HLFIR: hlfir.destroy
+! HLFIR: hlfir.destroy
+! HLFIR-NOT: omp.barrier
+! HLFIR: omp.terminator
+! HLFIR: }
+! HLFIR-NOT: omp.barrier
+! HLFIR: omp.terminator
+! HLFIR: }
+! HLFIR: return
+! HLFIR: }
+! HLFIR:}
+
+
+! FIR: func.func private @_workshare_copy_heap_Uxi32(%{{[a-z0-9]+}}: !fir.ref<!fir.heap<!fir.array<?xi32>>>, %{{[a-z0-9]+}}: !fir.ref<!fir.heap<!fir.array<?xi32>>>
+! FIR: func.func private @_workshare_copy_i32(%{{[a-z0-9]+}}: !fir.ref<i32>, %{{[a-z0-9]+}}: !fir.ref<i32>
+
+! FIR: func.func @_QPsb1
+! FIR: omp.parallel {
+! FIR: omp.single copyprivate(%9 -> @_workshare_copy_i32 : !fir.ref<i32>, %10 -> @_workshare_copy_heap_Uxi32 : !fir.ref<!fir.heap<!fir.array<?xi32>>>) {
+! FIR: fir.allocmem
+! FIR: omp.wsloop {
+! FIR: omp.loop_nest
+! FIR: omp.single nowait {
+! FIR: fir.call @_FortranAAssign
+! FIR: fir.freemem
+! FIR: omp.terminator
+! FIR: }
+! FIR: omp.barrier
+! FIR: omp.terminator
+! FIR: }
More information about the llvm-branch-commits
mailing list