1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt < %s -passes=loop-vectorize -force-target-supports-scalable-vectors=true -scalable-vectorization=on -S | FileCheck %s
4 target 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"
6 define i8 @reduction_add_trunc(ptr noalias nocapture %A) {
7 ; CHECK-LABEL: @reduction_add_trunc(
8 ; CHECK: call i32 @llvm.vscale.i32()
9 ; CHECK: call i32 @llvm.vscale.i32()
10 ; CHECK: [[TMP30:%.*]] = call i32 @llvm.vscale.i32()
11 ; CHECK-NEXT: [[TMP31:%.*]] = mul i32 [[TMP30]], 16
13 ; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
14 ; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <vscale x 8 x i32> [ insertelement (<vscale x 8 x i32> zeroinitializer, i32 255, i32 0), %vector.ph ], [ [[TMP34:%.*]], %vector.body ]
15 ; CHECK-NEXT: [[VEC_PHI1:%.*]] = phi <vscale x 8 x i32> [ zeroinitializer, %vector.ph ], [ [[TMP36:%.*]], %vector.body ]
16 ; CHECK: [[TMP14:%.*]] = and <vscale x 8 x i32> [[VEC_PHI]], splat (i32 255)
17 ; CHECK-NEXT: [[TMP15:%.*]] = and <vscale x 8 x i32> [[VEC_PHI1]], splat (i32 255)
18 ; CHECK: [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr
19 ; CHECK: [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr
20 ; CHECK-NEXT: [[TMP26:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
21 ; CHECK-NEXT: [[TMP27:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
22 ; CHECK-NEXT: [[TMP28:%.*]] = add <vscale x 8 x i32> [[TMP14]], [[TMP26]]
23 ; CHECK-NEXT: [[TMP29:%.*]] = add <vscale x 8 x i32> [[TMP15]], [[TMP27]]
24 ; CHECK-NEXT: [[TMP33:%.*]] = trunc <vscale x 8 x i32> [[TMP28]] to <vscale x 8 x i8>
25 ; CHECK-NEXT: [[TMP35:%.*]] = trunc <vscale x 8 x i32> [[TMP29]] to <vscale x 8 x i8>
26 ; CHECK-NEXT: [[TMP34]] = zext <vscale x 8 x i8> [[TMP33]] to <vscale x 8 x i32>
27 ; CHECK-NEXT: [[TMP36]] = zext <vscale x 8 x i8> [[TMP35]] to <vscale x 8 x i32>
28 ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP31]]
29 ; CHECK-NEXT: [[TMP32:%.*]] = icmp eq i32 [[INDEX_NEXT]], {{%.*}}
30 ; CHECK: middle.block:
31 ; CHECK-NEXT: [[TMP37:%.*]] = trunc <vscale x 8 x i32> [[TMP34]] to <vscale x 8 x i8>
32 ; CHECK-NEXT: [[TMP38:%.*]] = trunc <vscale x 8 x i32> [[TMP36]] to <vscale x 8 x i8>
33 ; CHECK-NEXT: [[BIN_RDX:%.*]] = add <vscale x 8 x i8> [[TMP38]], [[TMP37]]
34 ; CHECK-NEXT: [[TMP39:%.*]] = call i8 @llvm.vector.reduce.add.nxv8i8(<vscale x 8 x i8> [[BIN_RDX]])
35 ; CHECK-NEXT: [[TMP40:%.*]] = zext i8 [[TMP39]] to i32
36 ; CHECK-NEXT: %cmp.n = icmp eq i32 256, %n.vec
41 loop: ; preds = %entry, %loop
42 %indvars.iv = phi i32 [ %indvars.iv.next, %loop ], [ 0, %entry ]
43 %sum.02p = phi i32 [ %l9, %loop ], [ 255, %entry ]
44 %sum.02 = and i32 %sum.02p, 255
45 %l2 = getelementptr inbounds i8, ptr %A, i32 %indvars.iv
46 %l3 = load i8, ptr %l2, align 4
47 %l3e = zext i8 %l3 to i32
48 %l9 = add i32 %sum.02, %l3e
49 %indvars.iv.next = add i32 %indvars.iv, 1
50 %exitcond = icmp eq i32 %indvars.iv.next, 256
51 br i1 %exitcond, label %exit, label %loop, !llvm.loop !0
54 %sum.0.lcssa = phi i32 [ %l9, %loop ]
55 %ret = trunc i32 %sum.0.lcssa to i8
59 !0 = distinct !{!0, !1, !2, !3, !4}
60 !1 = !{!"llvm.loop.vectorize.width", i32 8}
61 !2 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}
62 !3 = !{!"llvm.loop.interleave.count", i32 2}
63 !4 = !{!"llvm.loop.vectorize.enable", i1 true}