1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2 ; RUN: opt -passes=loop-vectorize \
3 ; RUN: -force-tail-folding-style=data-with-evl \
4 ; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
5 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=IF-EVL %s
7 ; RUN: opt -passes=loop-vectorize \
8 ; RUN: -force-tail-folding-style=none \
9 ; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
10 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=NO-VP %s
12 define void @iv32(ptr noalias %a, ptr noalias %b, i32 %N) {
13 ; IF-EVL-LABEL: @iv32(
15 ; IF-EVL-NEXT: [[TMP19:%.*]] = sub i32 -1, [[N:%.*]]
16 ; IF-EVL-NEXT: [[TMP1:%.*]] = call i32 @llvm.vscale.i32()
17 ; IF-EVL-NEXT: [[TMP2:%.*]] = mul i32 [[TMP1]], 4
18 ; IF-EVL-NEXT: [[TMP3:%.*]] = icmp ult i32 [[TMP19]], [[TMP2]]
19 ; IF-EVL-NEXT: br i1 [[TMP3]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
21 ; IF-EVL-NEXT: [[TMP4:%.*]] = call i32 @llvm.vscale.i32()
22 ; IF-EVL-NEXT: [[TMP5:%.*]] = mul i32 [[TMP4]], 4
23 ; IF-EVL-NEXT: [[TMP8:%.*]] = sub i32 [[TMP5]], 1
24 ; IF-EVL-NEXT: [[N_RND_UP:%.*]] = add i32 [[N]], [[TMP8]]
25 ; IF-EVL-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[N_RND_UP]], [[TMP5]]
26 ; IF-EVL-NEXT: [[N_VEC:%.*]] = sub i32 [[N_RND_UP]], [[N_MOD_VF]]
27 ; IF-EVL-NEXT: [[TMP9:%.*]] = call i32 @llvm.vscale.i32()
28 ; IF-EVL-NEXT: [[TMP10:%.*]] = mul i32 [[TMP9]], 4
29 ; IF-EVL-NEXT: br label [[FOR_BODY:%.*]]
30 ; IF-EVL: vector.body:
31 ; IF-EVL-NEXT: [[IV:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
32 ; IF-EVL-NEXT: [[EVL_BASED_IV:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INDEX_EVL_NEXT:%.*]], [[FOR_BODY]] ]
33 ; IF-EVL-NEXT: [[TMP11:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
34 ; IF-EVL-NEXT: [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[TMP11]], i32 4, i1 true)
35 ; IF-EVL-NEXT: [[TMP13:%.*]] = add i32 [[EVL_BASED_IV]], 0
36 ; IF-EVL-NEXT: [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[TMP13]]
37 ; IF-EVL-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
38 ; IF-EVL-NEXT: [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer), i32 [[TMP12]])
39 ; IF-EVL-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[TMP13]]
40 ; IF-EVL-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
41 ; IF-EVL-NEXT: call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP_LOAD]], ptr align 4 [[TMP17]], <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer), i32 [[TMP12]])
42 ; IF-EVL-NEXT: [[INDEX_EVL_NEXT]] = add i32 [[TMP12]], [[EVL_BASED_IV]]
43 ; IF-EVL-NEXT: [[IV_NEXT]] = add i32 [[IV]], [[TMP10]]
44 ; IF-EVL-NEXT: [[TMP18:%.*]] = icmp eq i32 [[IV_NEXT]], [[N_VEC]]
45 ; IF-EVL-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
46 ; IF-EVL: middle.block:
47 ; IF-EVL-NEXT: br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
49 ; IF-EVL-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
50 ; IF-EVL-NEXT: br label [[FOR_BODY1:%.*]]
52 ; IF-EVL-NEXT: [[IV1:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
53 ; IF-EVL-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[IV1]]
54 ; IF-EVL-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
55 ; IF-EVL-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[IV1]]
56 ; IF-EVL-NEXT: store i32 [[TMP0]], ptr [[ARRAYIDX4]], align 4
57 ; IF-EVL-NEXT: [[IV_NEXT1]] = add nuw nsw i32 [[IV1]], 1
58 ; IF-EVL-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[IV_NEXT1]], [[N]]
59 ; IF-EVL-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
60 ; IF-EVL: for.cond.cleanup:
61 ; IF-EVL-NEXT: ret void
65 ; NO-VP-NEXT: [[TMP0:%.*]] = call i32 @llvm.vscale.i32()
66 ; NO-VP-NEXT: [[TMP10:%.*]] = mul i32 [[TMP0]], 4
67 ; NO-VP-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N:%.*]], [[TMP10]]
68 ; NO-VP-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
70 ; NO-VP-NEXT: [[TMP1:%.*]] = call i32 @llvm.vscale.i32()
71 ; NO-VP-NEXT: [[TMP11:%.*]] = mul i32 [[TMP1]], 4
72 ; NO-VP-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[N]], [[TMP11]]
73 ; NO-VP-NEXT: [[N_VEC:%.*]] = sub i32 [[N]], [[N_MOD_VF]]
74 ; NO-VP-NEXT: [[TMP2:%.*]] = call i32 @llvm.vscale.i32()
75 ; NO-VP-NEXT: [[TMP12:%.*]] = mul i32 [[TMP2]], 4
76 ; NO-VP-NEXT: br label [[VECTOR_BODY:%.*]]
78 ; NO-VP-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
79 ; NO-VP-NEXT: [[TMP3:%.*]] = add i32 [[INDEX]], 0
80 ; NO-VP-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[TMP3]]
81 ; NO-VP-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
82 ; NO-VP-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP5]], align 4
83 ; NO-VP-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[TMP3]]
84 ; NO-VP-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
85 ; NO-VP-NEXT: store <vscale x 4 x i32> [[WIDE_LOAD]], ptr [[TMP7]], align 4
86 ; NO-VP-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP12]]
87 ; NO-VP-NEXT: [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
88 ; NO-VP-NEXT: br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
89 ; NO-VP: middle.block:
90 ; NO-VP-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N]], [[N_VEC]]
91 ; NO-VP-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
93 ; NO-VP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
94 ; NO-VP-NEXT: br label [[FOR_BODY:%.*]]
96 ; NO-VP-NEXT: [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
97 ; NO-VP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[IV]]
98 ; NO-VP-NEXT: [[TMP9:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
99 ; NO-VP-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[IV]]
100 ; NO-VP-NEXT: store i32 [[TMP9]], ptr [[ARRAYIDX4]], align 4
101 ; NO-VP-NEXT: [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
102 ; NO-VP-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[IV_NEXT]], [[N]]
103 ; NO-VP-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
104 ; NO-VP: for.cond.cleanup:
105 ; NO-VP-NEXT: ret void
111 %iv = phi i32 [ 0, %entry ], [ %iv.next, %for.body ]
112 %arrayidx = getelementptr inbounds i32, ptr %b, i32 %iv
113 %0 = load i32, ptr %arrayidx, align 4
114 %arrayidx4 = getelementptr inbounds i32, ptr %a, i32 %iv
115 store i32 %0, ptr %arrayidx4, align 4
116 %iv.next = add nuw nsw i32 %iv, 1
117 %exitcond.not = icmp eq i32 %iv.next, %N
118 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body