50 lines · plain
1; RUN: opt -S -passes=loop-vectorize,dce,instcombine -force-vector-interleave=1 -force-vector-width=4 < %s | FileCheck %s2 3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"4 5@uf = common global [100 x i32] zeroinitializer, align 166@xi = common global [100 x i32] zeroinitializer, align 167@q = common global [100 x i32] zeroinitializer, align 168 9; PR1645510 11 12; Due to a bug in the way we handled reverse induction stores we would generate13; a shuffle too many.14 15define void @t() {16entry:17 br label %for.body18 19; CHECK-LABEL: @t(20; CHECK: vector.body:21; CHECK: [[VAR1:%[a-zA-Z0-9.]+]] = load <4 x i32>22; CHECK: [[VAR2:%[a-zA-Z0-9.]+]] = load <4 x i32>23; CHECK: [[VAR3:%[a-zA-Z0-9]+]] = add nsw <4 x i32> [[VAR2]], [[VAR1]]24; CHECK: store <4 x i32> [[VAR3]]25; CHECK: [[VAR4:%[a-zA-Z0-9.]+]] = load <4 x i32>26; CHECK: add nsw <4 x i32> [[VAR3]], [[VAR4]]27; CHECK-NOT: shufflevector28 29for.body:30 %indvars.iv = phi i64 [ 93, %entry ], [ %indvars.iv.next, %for.body ]31 %0 = add i64 %indvars.iv, 132 %arrayidx = getelementptr inbounds [100 x i32], ptr @uf, i64 0, i64 %033 %arrayidx3 = getelementptr inbounds [100 x i32], ptr @xi, i64 0, i64 %034 %1 = load i32, ptr %arrayidx3, align 435 %2 = load i32, ptr %arrayidx, align 436 %add4 = add nsw i32 %2, %137 store i32 %add4, ptr %arrayidx, align 438 %arrayidx7 = getelementptr inbounds [100 x i32], ptr @q, i64 0, i64 %039 %3 = load i32, ptr %arrayidx7, align 440 %add8 = add nsw i32 %add4, %341 store i32 %add8, ptr %arrayidx, align 442 %indvars.iv.next = add i64 %indvars.iv, -143 %4 = trunc i64 %indvars.iv.next to i3244 %cmp = icmp ugt i32 %4, 245 br i1 %cmp, label %for.body, label %for.end46 47for.end:48 ret void49}50