1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 6
; RUN: opt -mattr=+simd128 -passes=loop-vectorize %s -S | FileCheck %s
; RUN: opt -mattr=+simd128 -passes=loop-vectorize -vectorizer-maximize-bandwidth %s -S | FileCheck %s --check-prefix=CHECK-MAX-BANDWIDTH
target triple = "wasm32"
define hidden i32 @accumulate_add_u8_u8(ptr noundef readonly %a, ptr noundef readonly %b, i32 noundef %N) {
; CHECK-LABEL: define hidden i32 @accumulate_add_u8_u8(
; CHECK-SAME: ptr noundef readonly [[A:%.*]], ptr noundef readonly [[B:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT: [[ENTRY:.*]]:
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK: [[VECTOR_PH]]:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[N]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i32 [[N]], [[N_MOD_VF]]
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK: [[VECTOR_BODY]]:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i32 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
; CHECK-NEXT: [[TMP3:%.*]] = zext <4 x i8> [[WIDE_LOAD]] to <4 x i32>
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i32 [[INDEX]]
; CHECK-NEXT: [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
; CHECK-NEXT: [[TMP6:%.*]] = zext <4 x i8> [[WIDE_LOAD1]] to <4 x i32>
; CHECK-NEXT: [[TMP7:%.*]] = add <4 x i32> [[VEC_PHI]], [[TMP3]]
; CHECK-NEXT: [[TMP8]] = add <4 x i32> [[TMP7]], [[TMP6]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: [[MIDDLE_BLOCK]]:
; CHECK-NEXT: [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP8]])
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
; CHECK: [[SCALAR_PH]]:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-NEXT: br label %[[FOR_BODY:.*]]
; CHECK: [[FOR_COND_CLEANUP]]:
; CHECK-NEXT: [[RESULT_0_LCSSA:%.*]] = phi i32 [ [[ADD3:%.*]], %[[FOR_BODY]] ], [ [[TMP10]], %[[MIDDLE_BLOCK]] ]
; CHECK-NEXT: ret i32 [[RESULT_0_LCSSA]]
; CHECK: [[FOR_BODY]]:
; CHECK-NEXT: [[IV:%.*]] = phi i32 [ [[INC:%.*]], %[[FOR_BODY]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
; CHECK-NEXT: [[RED:%.*]] = phi i32 [ [[ADD3]], %[[FOR_BODY]] ], [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ]
; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i32 [[IV]]
; CHECK-NEXT: [[TMP11:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
; CHECK-NEXT: [[CONV:%.*]] = zext i8 [[TMP11]] to i32
; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i32 [[IV]]
; CHECK-NEXT: [[TMP12:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
; CHECK-NEXT: [[CONV2:%.*]] = zext i8 [[TMP12]] to i32
; CHECK-NEXT: [[ADD:%.*]] = add i32 [[RED]], [[CONV]]
; CHECK-NEXT: [[ADD3]] = add i32 [[ADD]], [[CONV2]]
; CHECK-NEXT: [[INC]] = add nuw i32 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[INC]], [[N]]
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_COND_CLEANUP]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
;
; CHECK-MAX-BANDWIDTH-LABEL: define hidden i32 @accumulate_add_u8_u8(
; CHECK-MAX-BANDWIDTH-SAME: ptr noundef readonly [[A:%.*]], ptr noundef readonly [[B:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-MAX-BANDWIDTH-NEXT: [[ENTRY:.*]]:
; CHECK-MAX-BANDWIDTH-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[N]], 16
; CHECK-MAX-BANDWIDTH-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
; CHECK-MAX-BANDWIDTH: [[VECTOR_PH]]:
; CHECK-MAX-BANDWIDTH-NEXT: [[N_MOD_VF:%.*]] = urem i32 [[N]], 16
; CHECK-MAX-BANDWIDTH-NEXT: [[N_VEC:%.*]] = sub i32 [[N]], [[N_MOD_VF]]
; CHECK-MAX-BANDWIDTH-NEXT: br label %[[VECTOR_BODY:.*]]
; CHECK-MAX-BANDWIDTH: [[VECTOR_BODY]]:
; CHECK-MAX-BANDWIDTH-NEXT: [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-MAX-BANDWIDTH-NEXT: [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[PARTIAL_REDUCE2:%.*]], %[[VECTOR_BODY]] ]
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i32 [[INDEX]]
; CHECK-MAX-BANDWIDTH-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i32 [[INDEX]]
; CHECK-MAX-BANDWIDTH-NEXT: [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
; CHECK-MAX-BANDWIDTH-NEXT: [[PARTIAL_REDUCE:%.*]] = call <4 x i32> @llvm.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP3]])
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
; CHECK-MAX-BANDWIDTH-NEXT: [[PARTIAL_REDUCE2]] = call <4 x i32> @llvm.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[PARTIAL_REDUCE]], <16 x i32> [[TMP6]])
; CHECK-MAX-BANDWIDTH-NEXT: [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 16
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-MAX-BANDWIDTH-NEXT: br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK-MAX-BANDWIDTH: [[MIDDLE_BLOCK]]:
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP7:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[PARTIAL_REDUCE2]])
; CHECK-MAX-BANDWIDTH-NEXT: [[CMP_N:%.*]] = icmp eq i32 [[N]], [[N_VEC]]
; CHECK-MAX-BANDWIDTH-NEXT: br i1 [[CMP_N]], label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
; CHECK-MAX-BANDWIDTH: [[SCALAR_PH]]:
; CHECK-MAX-BANDWIDTH-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-MAX-BANDWIDTH-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP7]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
; CHECK-MAX-BANDWIDTH-NEXT: br label %[[FOR_BODY:.*]]
; CHECK-MAX-BANDWIDTH: [[FOR_COND_CLEANUP]]:
; CHECK-MAX-BANDWIDTH-NEXT: [[RESULT_0_LCSSA:%.*]] = phi i32 [ [[ADD3:%.*]], %[[FOR_BODY]] ], [ [[TMP7]], %[[MIDDLE_BLOCK]] ]
; CHECK-MAX-BANDWIDTH-NEXT: ret i32 [[RESULT_0_LCSSA]]
; CHECK-MAX-BANDWIDTH: [[FOR_BODY]]:
; CHECK-MAX-BANDWIDTH-NEXT: [[IV:%.*]] = phi i32 [ [[INC:%.*]], %[[FOR_BODY]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
; CHECK-MAX-BANDWIDTH-NEXT: [[RED:%.*]] = phi i32 [ [[ADD3]], %[[FOR_BODY]] ], [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ]
; CHECK-MAX-BANDWIDTH-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i32 [[IV]]
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP11:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
; CHECK-MAX-BANDWIDTH-NEXT: [[CONV:%.*]] = zext i8 [[TMP11]] to i32
; CHECK-MAX-BANDWIDTH-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i32 [[IV]]
; CHECK-MAX-BANDWIDTH-NEXT: [[TMP12:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
; CHECK-MAX-BANDWIDTH-NEXT: [[CONV2:%.*]] = zext i8 [[TMP12]] to i32
; CHECK-MAX-BANDWIDTH-NEXT: [[ADD:%.*]] = add i32 [[RED]], [[CONV]]
; CHECK-MAX-BANDWIDTH-NEXT: [[ADD3]] = add i32 [[ADD]], [[CONV2]]
; CHECK-MAX-BANDWIDTH-NEXT: [[INC]] = add nuw i32 [[IV]], 1
; CHECK-MAX-BANDWIDTH-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i32 [[INC]], [[N]]
; CHECK-MAX-BANDWIDTH-NEXT: br i1 [[EXITCOND_NOT]], label %[[FOR_COND_CLEANUP]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
;
entry:
br label %for.body
for.cond.cleanup: ; preds = %for.body
ret i32 %add3
for.body: ; preds = %entry, %for.body
%iv = phi i32 [ %inc, %for.body ], [ 0, %entry ]
%red = phi i32 [ %add3, %for.body ], [ 0, %entry ]
%arrayidx = getelementptr inbounds nuw i8, ptr %a, i32 %iv
%0 = load i8, ptr %arrayidx, align 1
%conv = zext i8 %0 to i32
%arrayidx1 = getelementptr inbounds nuw i8, ptr %b, i32 %iv
%1 = load i8, ptr %arrayidx1, align 1
%conv2 = zext i8 %1 to i32
%add = add i32 %red, %conv
%add3 = add i32 %add, %conv2
%inc = add nuw i32 %iv, 1
%exitcond.not = icmp eq i32 %inc, %N
br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
}
|
