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
|
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -passes=instcombine -S | FileCheck %s
target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
target triple = "armv8-arm-none-eabi"
; Turning a vld1 intrinsic into an llvm load is beneficial
; when the underlying object being addressed comes from a
; constant, since we get constant-folding for free.
; Bail the optimization if the alignment is not a constant.
define <2 x i64> @vld1_align(ptr %ptr, i32 %align) {
; CHECK-LABEL: @vld1_align(
; CHECK-NEXT: [[VLD1:%.*]] = call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0(ptr [[PTR:%.*]], i32 [[ALIGN:%.*]])
; CHECK-NEXT: ret <2 x i64> [[VLD1]]
;
%vld1 = call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0(ptr %ptr, i32 %align)
ret <2 x i64> %vld1
}
; Bail the optimization if the alignment is not power of 2.
define <2 x i64> @vld1_align_pow2(ptr %ptr) {
; CHECK-LABEL: @vld1_align_pow2(
; CHECK-NEXT: [[VLD1:%.*]] = call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0(ptr [[PTR:%.*]], i32 3)
; CHECK-NEXT: ret <2 x i64> [[VLD1]]
;
%vld1 = call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0(ptr %ptr, i32 3)
ret <2 x i64> %vld1
}
define <8 x i8> @vld1_8x8(ptr %ptr) {
; CHECK-LABEL: @vld1_8x8(
; CHECK-NEXT: [[TMP2:%.*]] = load <8 x i8>, ptr [[PTR:%.*]], align 1
; CHECK-NEXT: ret <8 x i8> [[TMP2]]
;
%vld1 = call <8 x i8> @llvm.arm.neon.vld1.v8i8.p0(ptr %ptr, i32 1)
ret <8 x i8> %vld1
}
define <4 x i16> @vld1_4x16(ptr %ptr) {
; CHECK-LABEL: @vld1_4x16(
; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i16>, ptr [[PTR:%.*]], align 2
; CHECK-NEXT: ret <4 x i16> [[TMP2]]
;
%vld1 = call <4 x i16> @llvm.arm.neon.vld1.v4i16.p0(ptr %ptr, i32 2)
ret <4 x i16> %vld1
}
define <2 x i32> @vld1_2x32(ptr %ptr) {
; CHECK-LABEL: @vld1_2x32(
; CHECK-NEXT: [[TMP2:%.*]] = load <2 x i32>, ptr [[PTR:%.*]], align 4
; CHECK-NEXT: ret <2 x i32> [[TMP2]]
;
%vld1 = call <2 x i32> @llvm.arm.neon.vld1.v2i32.p0(ptr %ptr, i32 4)
ret <2 x i32> %vld1
}
define <1 x i64> @vld1_1x64(ptr %ptr) {
; CHECK-LABEL: @vld1_1x64(
; CHECK-NEXT: [[TMP2:%.*]] = load <1 x i64>, ptr [[PTR:%.*]], align 8
; CHECK-NEXT: ret <1 x i64> [[TMP2]]
;
%vld1 = call <1 x i64> @llvm.arm.neon.vld1.v1i64.p0(ptr %ptr, i32 8)
ret <1 x i64> %vld1
}
define <8 x i16> @vld1_8x16(ptr %ptr) {
; CHECK-LABEL: @vld1_8x16(
; CHECK-NEXT: [[TMP2:%.*]] = load <8 x i16>, ptr [[PTR:%.*]], align 2
; CHECK-NEXT: ret <8 x i16> [[TMP2]]
;
%vld1 = call <8 x i16> @llvm.arm.neon.vld1.v8i16.p0(ptr %ptr, i32 2)
ret <8 x i16> %vld1
}
define <16 x i8> @vld1_16x8(ptr %ptr) {
; CHECK-LABEL: @vld1_16x8(
; CHECK-NEXT: [[TMP2:%.*]] = load <16 x i8>, ptr [[PTR:%.*]], align 1
; CHECK-NEXT: ret <16 x i8> [[TMP2]]
;
%vld1 = call <16 x i8> @llvm.arm.neon.vld1.v16i8.p0(ptr %ptr, i32 1)
ret <16 x i8> %vld1
}
define <4 x i32> @vld1_4x32(ptr %ptr) {
; CHECK-LABEL: @vld1_4x32(
; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[PTR:%.*]], align 4
; CHECK-NEXT: ret <4 x i32> [[TMP2]]
;
%vld1 = call <4 x i32> @llvm.arm.neon.vld1.v4i32.p0(ptr %ptr, i32 4)
ret <4 x i32> %vld1
}
define <2 x i64> @vld1_2x64(ptr %ptr) {
; CHECK-LABEL: @vld1_2x64(
; CHECK-NEXT: [[TMP2:%.*]] = load <2 x i64>, ptr [[PTR:%.*]], align 8
; CHECK-NEXT: ret <2 x i64> [[TMP2]]
;
%vld1 = call <2 x i64> @llvm.arm.neon.vld1.v2i64.p0(ptr %ptr, i32 8)
ret <2 x i64> %vld1
}
declare <8 x i8> @llvm.arm.neon.vld1.v8i8.p0(ptr, i32)
declare <4 x i16> @llvm.arm.neon.vld1.v4i16.p0(ptr, i32)
declare <2 x i32> @llvm.arm.neon.vld1.v2i32.p0(ptr, i32)
declare <1 x i64> @llvm.arm.neon.vld1.v1i64.p0(ptr, i32)
declare <8 x i16> @llvm.arm.neon.vld1.v8i16.p0(ptr, i32)
declare <16 x i8> @llvm.arm.neon.vld1.v16i8.p0(ptr, i32)
declare <4 x i32> @llvm.arm.neon.vld1.v4i32.p0(ptr, i32)
declare <2 x i64> @llvm.arm.neon.vld1.v2i64.p0(ptr, i32)
|
