73 lines · plain
1; RUN: opt %loadNPMPolly '-passes=polly-custom<import-jscop;ast>' -polly-print-ast -polly-ast-detect-parallel -disable-output < %s | FileCheck %s2;3; CHECK: #pragma known-parallel reduction (+ : MemRef_A)4; CHECK-NEXT: for (int c0 = 0; c0 <= 2; c0 += 1) {5; CHECK-NEXT: if (c0 == 1) {6; CHECK-NEXT: #pragma simd7; CHECK-NEXT: for (int c1 = 0; c1 < 2 * n; c1 += 1)8; CHECK-NEXT: Stmt_S1(c1);9; CHECK-NEXT: } else10; CHECK-NEXT: #pragma simd reduction (+ : MemRef_A)11; CHECK-NEXT: for (int c1 = (-c0 / 2) + 2; c1 <= 2 * n; c1 += 2)12; CHECK-NEXT: Stmt_S0(2 * n - c1);13; CHECK-NEXT: }14;15; void rmalrs(int *A, long n) {16; for (long i = 0; i < 2 * n; i++)17; S0: A[0] += i;18; for (long i = 0; i < 2 * n; i++)19; S1: A[i + 1] = 1;20; }21;22target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-n32-S64"23 24define void @rmalrs(ptr %A, i32 %n) {25entry:26 br label %for.cond27 28for.cond: ; preds = %for.inc, %entry29 %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]30 %mul = shl nsw i32 %n, 131 %cmp = icmp slt i32 %i.0, %mul32 br i1 %cmp, label %for.body, label %for.end33 34for.body: ; preds = %for.cond35 br label %S036 37S0: ; preds = %for.body38 %tmp = load i32, ptr %A, align 439 %add = add nsw i32 %tmp, %i.040 store i32 %add, ptr %A, align 441 br label %for.inc42 43for.inc: ; preds = %S044 %inc = add nsw i32 %i.0, 145 br label %for.cond46 47for.end: ; preds = %for.cond48 br label %for.cond249 50for.cond2: ; preds = %for.inc8, %for.end51 %i1.0 = phi i32 [ 0, %for.end ], [ %inc9, %for.inc8 ]52 %mul3 = shl nsw i32 %n, 153 %cmp4 = icmp slt i32 %i1.0, %mul354 br i1 %cmp4, label %for.body5, label %for.end1055 56for.body5: ; preds = %for.cond257 br label %S158 59S1: ; preds = %for.body560 %add6 = add nsw i32 %i1.0, 161 %arrayidx7 = getelementptr inbounds i32, ptr %A, i32 %add662 store i32 1, ptr %arrayidx7, align 463 br label %for.inc864 65for.inc8: ; preds = %S166 %inc9 = add nsw i32 %i1.0, 167 br label %for.cond268 69for.end10: ; preds = %for.cond270 ret void71}72 73