[Polly][IslAst] Fix minimal dependence distance.

Summary:
When checking the parallelism of a scheduling dimension, we first check if excluding reduction dependences the loop is parallel or not.
If the loop is not parallel, then we need to return the minimal dependence distance of all data dependences, including the previously subtracted reduction dependences.

Reviewers: grosser, Meinersbur, efriedma, eli.friedman, jdoerfert, bollu

Reviewed By: Meinersbur

Subscribers: llvm-commits, pollydev

Tags: #polly

Differential Revision: https://reviews.llvm.org/D45236

llvm-svn: 329214

diff --git a/polly/lib/Analysis/DependenceInfo.cpp b/polly/lib/Analysis/DependenceInfo.cpp
index 1b09e4f..7720712 100644 (file)
--- a/polly/lib/Analysis/DependenceInfo.cpp
+++ b/polly/lib/Analysis/DependenceInfo.cpp
@@ -646,7 +646,8 @@ void Dependences::calculateDependences(Scop &S) {
 
     // Step 4)
     addPrivatizationDependences();
-  }
+  } else
+    TC_RED = isl_union_map_empty(isl_union_map_get_space(RED));
 
   DEBUG({
     dbgs() << "Final Wrapped Dependences:\n";

diff --git a/polly/lib/CodeGen/IslAst.cpp b/polly/lib/CodeGen/IslAst.cpp
index 23767cf..7e8739d 100644 (file)
--- a/polly/lib/CodeGen/IslAst.cpp
+++ b/polly/lib/CodeGen/IslAst.cpp
@@ -219,9 +219,14 @@ static bool astScheduleDimIsParallel(__isl_keep isl_ast_build *Build,
   isl_union_map *Deps = D->getDependences(
       Dependences::TYPE_RAW | Dependences::TYPE_WAW | Dependences::TYPE_WAR);
 
-  if (!D->isParallel(Schedule, Deps, &NodeInfo->MinimalDependenceDistance) &&
-      !isl_union_map_free(Schedule))
+  if (!D->isParallel(Schedule, Deps)) {
+    isl_union_map *DepsAll =
+        D->getDependences(Dependences::TYPE_RAW | Dependences::TYPE_WAW |
+                          Dependences::TYPE_WAR | Dependences::TYPE_TC_RED);
+    D->isParallel(Schedule, DepsAll, &NodeInfo->MinimalDependenceDistance);
+    isl_union_map_free(Schedule);
     return false;
+  }
 
   isl_union_map *RedDeps = D->getDependences(Dependences::TYPE_TC_RED);
   if (!D->isParallel(Schedule, RedDeps))

diff --git a/polly/test/Isl/Ast/dependence_distance_minimal.ll b/polly/test/Isl/Ast/dependence_distance_minimal.ll
new file mode 100644 (file)
index 0000000..2785ae7
--- /dev/null
+++ b/polly/test/Isl/Ast/dependence_distance_minimal.ll
@@ -0,0 +1,57 @@
+; RUN: opt %loadPolly -polly-ast -polly-ast-detect-parallel -analyze < %s | FileCheck %s
+;
+; The minimal dependence distance of the innermost loop should be 1 instead of 250.
+; CHECK:    #pragma minimal dependence distance: 1
+; CHECK:    for (int c0 = 0; c0 <= 499; c0 += 1)
+; CHECK:      #pragma minimal dependence distance: 1
+; CHECK:      for (int c1 = 0; c1 <= 998; c1 += 1) {
+; CHECK:        Stmt_bb9(c0, c1);
+; CHECK:        Stmt_bb9_b(c0, c1);
+;
+;    void foo (int *A, int *B) {
+;      for (int i=0; i < 500; i++) {
+;        for (int j=0; j < 1000; j++) {
+;          B[i] = B[i] + 1;
+;          A[j] += A[j % 250];
+;        }
+;      }
+;    }
+;
+target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
+
+define void @foo(i32* nocapture %arg, i32* nocapture %arg1) {
+bb:
+  br label %bb2
+
+bb2:                                              ; preds = %bb
+  br label %bb4
+
+bb3:                                              ; preds = %bb6
+  ret void
+
+bb4:                                              ; preds = %bb6, %bb2
+  %tmp = phi i32 [ 0, %bb2 ], [ %tmp7, %bb6 ]
+  %tmp5 = getelementptr inbounds i32, i32* %arg1, i32 %tmp
+  br label %bb9
+
+bb6:                                              ; preds = %bb9
+  %tmp7 = add nuw nsw i32 %tmp, 1
+  %tmp8 = icmp eq i32 %tmp7, 500
+  br i1 %tmp8, label %bb3, label %bb4
+
+bb9:                                              ; preds = %bb9, %bb4
+  %tmp10 = phi i32 [ 1, %bb4 ], [ %tmp19, %bb9 ]
+  %tmp11 = load i32, i32* %tmp5, align 4
+  %tmp12 = add nsw i32 %tmp11, 1
+  store i32 %tmp12, i32* %tmp5, align 4
+  %tmp13 = urem i32 %tmp10, 250
+  %tmp14 = getelementptr inbounds i32, i32* %arg, i32 %tmp13
+  %tmp15 = load i32, i32* %tmp14, align 4
+  %tmp16 = getelementptr inbounds i32, i32* %arg, i32 %tmp10
+  %tmp17 = load i32, i32* %tmp16, align 4
+  %tmp18 = add nsw i32 %tmp17, %tmp15
+  store i32 %tmp18, i32* %tmp16, align 4
+  %tmp19 = add nuw nsw i32 %tmp10, 1
+  %tmp20 = icmp eq i32 %tmp19, 1000
+  br i1 %tmp20, label %bb6, label %bb9
+}