1 ; RUN: opt -mtriple=x86_64-unknown-linux-gnu -loop-reduce -lsr-insns-cost=false -S < %s | FileCheck %s
2 ; Check LSR formula canonicalization will put loop invariant regs before
3 ; induction variable of current loop, so exprs involving loop invariant regs
4 ; can be promoted outside of current loop.
6 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
8 define void @foo(i32 %size, i32 %nsteps, i8* nocapture %maxarray, i8* nocapture readnone %buffer, i32 %init) local_unnamed_addr #0 {
10 %cmp25 = icmp sgt i32 %nsteps, 0
11 br i1 %cmp25, label %for.cond1.preheader.lr.ph, label %for.end12
13 for.cond1.preheader.lr.ph: ; preds = %entry
14 %cmp223 = icmp sgt i32 %size, 1
15 %t0 = sext i32 %init to i64
16 %wide.trip.count = zext i32 %size to i64
17 %wide.trip.count31 = zext i32 %nsteps to i64
18 br label %for.cond1.preheader
20 for.cond1.preheader: ; preds = %for.inc10, %for.cond1.preheader.lr.ph
21 %indvars.iv28 = phi i64 [ 0, %for.cond1.preheader.lr.ph ], [ %indvars.iv.next29, %for.inc10 ]
22 br i1 %cmp223, label %for.body3.lr.ph, label %for.inc10
24 for.body3.lr.ph: ; preds = %for.cond1.preheader
25 %t1 = add nsw i64 %indvars.iv28, %t0
26 %t2 = trunc i64 %indvars.iv28 to i8
29 ; Make sure loop invariant items are grouped together so that load address can
30 ; be represented in one getelementptr.
31 ; CHECK-LABEL: for.body3:
32 ; CHECK-NEXT: [[LSR:%[^,]+]] = phi i64 [ 1, %for.body3.lr.ph ], [ {{.*}}, %for.body3 ]
33 ; CHECK-NOT: = phi i64
34 ; CHECK-NEXT: [[LOADADDR:%[^,]+]] = getelementptr i8, i8* {{.*}}, i64 [[LSR]]
35 ; CHECK-NEXT: = load i8, i8* [[LOADADDR]], align 1
36 ; CHECK: br i1 %exitcond, label %for.inc10.loopexit, label %for.body3
38 for.body3: ; preds = %for.body3, %for.body3.lr.ph
39 %indvars.iv = phi i64 [ 1, %for.body3.lr.ph ], [ %indvars.iv.next, %for.body3 ]
40 %t5 = trunc i64 %indvars.iv to i8
41 %t3 = add nsw i64 %t1, %indvars.iv
42 %arrayidx = getelementptr inbounds i8, i8* %maxarray, i64 %t3
43 %t4 = load i8, i8* %arrayidx, align 1
44 %add5 = add i8 %t4, %t5
45 %add6 = add i8 %add5, %t2
46 %arrayidx9 = getelementptr inbounds i8, i8* %maxarray, i64 %indvars.iv
47 store i8 %add6, i8* %arrayidx9, align 1
48 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
49 %exitcond = icmp eq i64 %indvars.iv.next, %wide.trip.count
50 br i1 %exitcond, label %for.inc10.loopexit, label %for.body3
52 for.inc10.loopexit: ; preds = %for.body3
55 for.inc10: ; preds = %for.inc10.loopexit, %for.cond1.preheader
56 %indvars.iv.next29 = add nuw nsw i64 %indvars.iv28, 1
57 %exitcond32 = icmp eq i64 %indvars.iv.next29, %wide.trip.count31
58 br i1 %exitcond32, label %for.end12.loopexit, label %for.cond1.preheader
60 for.end12.loopexit: ; preds = %for.inc10
63 for.end12: ; preds = %for.end12.loopexit, %entry