[flang][OpenMP] Privatize "loop-local" values in do concurent on de…

…vice Extends ROCm#112. This PR extends support for `do concurrent` mapping to the device a bit more. In particular, it handles localization of loop-local values on the deive. Previously, this was only supported and tested on the host. See docs for `looputils::collectLoopLocalValues` for the definition of "loop-local" values.
ergawy · Aug 19, 2024 · ad3ac5d · ad3ac5d
1 parent d2b3307
commit ad3ac5d
Show file tree

Hide file tree

Showing 3 changed files with 41 additions and 24 deletions.
diff --git a/flang/lib/Optimizer/Transforms/DoConcurrentConversion.cpp b/flang/lib/Optimizer/Transforms/DoConcurrentConversion.cpp
@@ -565,9 +565,9 @@ class DoConcurrentConversion : public mlir::OpConversionPattern<fir::DoLoopOp> {
                   "defining operation.");
     }
 
-    llvm::SmallVector<mlir::Value> outermostLoopLives;
-    looputils::collectLoopLiveIns(doLoop, outermostLoopLives);
-    assert(!outermostLoopLives.empty());
+    llvm::SmallVector<mlir::Value> outermostLoopLiveIns;
+    looputils::collectLoopLiveIns(doLoop, outermostLoopLiveIns);
+    assert(!outermostLoopLiveIns.empty());
 
     looputils::LoopNestToIndVarMap loopNest;
     bool hasRemainingNestedLoops =
@@ -577,28 +577,33 @@ class DoConcurrentConversion : public mlir::OpConversionPattern<fir::DoLoopOp> {
                         "Some `do concurent` loops are not perfectly-nested. "
                         "These will be serialzied.");
 
-    mlir::IRMapping mapper;
-
     llvm::SetVector<mlir::Value> locals;
     looputils::collectLoopLocalValues(loopNest.back().first, locals);
+    outermostLoopLiveIns.erase(llvm::remove_if(outermostLoopLiveIns,
+                                               [&](mlir::Value liveIn) {
+                                                 return locals.contains(liveIn);
+                                               }),
+                               outermostLoopLiveIns.end());
 
     looputils::sinkLoopIVArgs(rewriter, loopNest);
 
     mlir::omp::TargetOp targetOp;
     mlir::omp::LoopNestOperands loopNestClauseOps;
 
+    mlir::IRMapping mapper;
+
     if (mapToDevice) {
       mlir::omp::TargetOperands targetClauseOps;
 
       // The outermost loop will contain all the live-in values in all nested
       // loops since live-in values are collected recursively for all nested
       // ops.
-      for (mlir::Value liveIn : outermostLoopLives)
+      for (mlir::Value liveIn : outermostLoopLiveIns)
         targetClauseOps.mapVars.push_back(
             genMapInfoOpForLiveIn(rewriter, liveIn));
 
       targetOp = genTargetOp(doLoop.getLoc(), rewriter, mapper,
-                             outermostLoopLives, targetClauseOps);
+                             outermostLoopLiveIns, targetClauseOps);
       genTeamsOp(doLoop.getLoc(), rewriter);
     }
 

diff --git a/flang/test/Transforms/DoConcurrent/locally_destroyed_temp.f90 b/flang/test/Transforms/DoConcurrent/locally_destroyed_temp.f90
@@ -5,7 +5,10 @@
 ! occur due to multiple teams trying to access the same allocation.
 
 ! RUN: %flang_fc1 -emit-hlfir -fopenmp -fdo-concurrent-parallel=host %s -o - \
-! RUN:   | FileCheck %s
+! RUN:   | FileCheck %s --check-prefixes=COMMON
+
+! RUN: %flang_fc1 -emit-hlfir -fopenmp -fdo-concurrent-parallel=device %s -o - \
+! RUN:   | FileCheck %s --check-prefixes=COMMON,DEVICE
 
 module struct_mod
     type test_struct
@@ -49,18 +52,27 @@ program main
     print *, "total =", total
 end program main
 
-! CHECK: omp.parallel {
-! CHECK:   %[[LOCAL_TEMP:.*]] = fir.alloca !fir.type<_QMstruct_modTtest_struct{x_:!fir.box<!fir.heap<i32>>}> {bindc_name = ".result"}
-! CHECK:   omp.wsloop {
-! CHECK:     omp.loop_nest {{.*}} {
-! CHECK:       %[[TEMP_VAL:.*]] = fir.call @_QMstruct_modPconstruct_from_components
-! CHECK:       fir.save_result %[[TEMP_VAL]] to %[[LOCAL_TEMP]]
-! CHECK:       %[[EMBOXED_LOCAL:.*]] = fir.embox %[[LOCAL_TEMP]]
-! CHECK:       %[[CONVERTED_LOCAL:.*]] = fir.convert %[[EMBOXED_LOCAL]]
-! CHECK:       fir.call @_FortranADestroy(%[[CONVERTED_LOCAL]])
-! CHECK:       omp.yield
-! CHECK:     }
-! CHECK:     omp.terminator
-! CHECK:   }
-! CHECK:   omp.terminator
-! CHECK: }
+! DEVICE: omp.target {{.*}} {
+! DEVICE: omp.teams {
+! COMMON: omp.parallel {
+! COMMON:   %[[LOCAL_TEMP:.*]] = fir.alloca !fir.type<_QMstruct_modTtest_struct{x_:!fir.box<!fir.heap<i32>>}> {bindc_name = ".result"}
+! DEVICE:   omp.distribute {
+! COMMON:   omp.wsloop {
+! COMMON:     omp.loop_nest {{.*}} {
+! COMMON:       %[[TEMP_VAL:.*]] = fir.call @_QMstruct_modPconstruct_from_components
+! COMMON:       fir.save_result %[[TEMP_VAL]] to %[[LOCAL_TEMP]]
+! COMMON:       %[[EMBOXED_LOCAL:.*]] = fir.embox %[[LOCAL_TEMP]]
+! COMMON:       %[[CONVERTED_LOCAL:.*]] = fir.convert %[[EMBOXED_LOCAL]]
+! COMMON:       fir.call @_FortranADestroy(%[[CONVERTED_LOCAL]])
+! COMMON:       omp.yield
+! COMMON:     }
+! COMMON:     omp.terminator
+! COMMON:   }
+! DEVICE:   omp.terminator
+! DEVICE:   }
+! COMMON:   omp.terminator
+! COMMON: }
+! DEVICE: omp.terminator
+! DEVICE: }
+! DEVICE: omp.terminator
+! DEVICE: }
diff --git a/offload/test/offloading/fortran/target_private.f90 b/offload/test/offloading/fortran/target_private.f90
@@ -24,6 +24,6 @@ program target_update
     print *, "y =", y(1)
 
 end program target_update
-! CHECK: "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}}
+! CHECK: "PluginInterface" dvice {{[0-9]+}} info: Launching kernel {{.*}}
 ! CHECK: x = 42
 ! CHECK: y = 84