1 # NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
2 # WARNING: update_mir_test_checks.py does not include the constant pools output,
3 # so this test requires manual fixing up after running the script.
5 # RUN: llc -mtriple=aarch64-- -O0 -run-pass=instruction-select -verify-machineinstrs %s -global-isel-abort=1 -o - | FileCheck %s
7 target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
8 target triple = "aarch64"
10 define <2 x float> @shuffle_v2f32(<2 x float> %a, <2 x float> %b) {
11 %shuf = shufflevector <2 x float> %a, <2 x float> %b, <2 x i32> <i32 1, i32 0>
15 define <4 x i32> @shuffle_v4i32(<4 x i32> %a, <4 x i32> %b) {
16 %shuf = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 0, i32 1, i32 3, i32 0>
20 define <4 x i32> @shuffle_tbl_v4i32(<4 x i32> %a, <4 x i32> %b) {
21 %shuf = shufflevector <4 x i32> %a, <4 x i32> %b, <4 x i32> <i32 5, i32 7, i32 1, i32 0>
25 define <2 x i64> @shuffle_v2i64(<2 x i64> %a, <2 x i64> %b) {
26 %shuf = shufflevector <2 x i64> %a, <2 x i64> %b, <2 x i32> zeroinitializer
36 tracksRegLiveness: true
41 ; CHECK-LABEL: name: shuffle_v2f32
44 ; CHECK: value: '<8 x i8> <i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3>'
46 ; CHECK: liveins: $d0, $d1
47 ; CHECK: [[COPY:%[0-9]+]]:fpr64 = COPY $d0
48 ; CHECK: [[COPY1:%[0-9]+]]:fpr64 = COPY $d1
49 ; CHECK: [[ADRP:%[0-9]+]]:gpr64common = ADRP target-flags(aarch64-page) %const.0
50 ; CHECK: [[LDRDui:%[0-9]+]]:fpr64 = LDRDui [[ADRP]], target-flags(aarch64-pageoff, aarch64-nc) %const.0
51 ; CHECK: [[DEF:%[0-9]+]]:fpr128 = IMPLICIT_DEF
52 ; CHECK: [[INSERT_SUBREG:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF]], [[COPY]], %subreg.dsub
53 ; CHECK: [[DEF1:%[0-9]+]]:fpr128 = IMPLICIT_DEF
54 ; CHECK: [[INSERT_SUBREG1:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF1]], [[COPY1]], %subreg.dsub
55 ; CHECK: [[INSvi64lane:%[0-9]+]]:fpr128 = INSvi64lane [[INSERT_SUBREG]], 1, [[INSERT_SUBREG1]], 0
56 ; CHECK: [[DEF2:%[0-9]+]]:fpr128 = IMPLICIT_DEF
57 ; CHECK: [[INSERT_SUBREG2:%[0-9]+]]:fpr128 = INSERT_SUBREG [[DEF2]], [[LDRDui]], %subreg.dsub
58 ; CHECK: [[TBLv16i8One:%[0-9]+]]:fpr128 = TBLv16i8One [[INSvi64lane]], [[INSERT_SUBREG2]]
59 ; CHECK: [[COPY2:%[0-9]+]]:fpr64 = COPY [[TBLv16i8One]].dsub
60 ; CHECK: $d0 = COPY [[COPY2]]
61 ; CHECK: RET_ReallyLR implicit $d0
62 %0:fpr(<2 x s32>) = COPY $d0
63 %1:fpr(<2 x s32>) = COPY $d1
64 %2:fpr(<2 x s32>) = G_SHUFFLE_VECTOR %0(<2 x s32>), %1, shufflemask(1, 0)
65 $d0 = COPY %2(<2 x s32>)
66 RET_ReallyLR implicit $d0
74 tracksRegLiveness: true
79 ; CHECK-LABEL: name: shuffle_v4i32
81 ; CHECK: value: '<16 x i8> <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3>'
82 ; CHECK: alignment: 16
83 ; CHECK: isTargetSpecific: false
84 ; CHECK: liveins: $q0, $q1
85 ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
86 ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
87 ; CHECK: [[ADRP:%[0-9]+]]:gpr64common = ADRP target-flags(aarch64-page) %const.0
88 ; CHECK: [[LDRQui:%[0-9]+]]:fpr128 = LDRQui [[ADRP]], target-flags(aarch64-pageoff, aarch64-nc) %const.0
89 ; CHECK: [[REG_SEQUENCE:%[0-9]+]]:qq = REG_SEQUENCE [[COPY]], %subreg.qsub0, [[COPY1]], %subreg.qsub1
90 ; CHECK: [[TBLv16i8Two:%[0-9]+]]:fpr128 = TBLv16i8Two [[REG_SEQUENCE]], [[LDRQui]]
91 ; CHECK: $q0 = COPY [[TBLv16i8Two]]
92 ; CHECK: RET_ReallyLR implicit $q0
93 %0:fpr(<4 x s32>) = COPY $q0
94 %1:fpr(<4 x s32>) = COPY $q1
95 %2:fpr(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(0, 1, 3, 0)
96 $q0 = COPY %2(<4 x s32>)
97 RET_ReallyLR implicit $q0
101 name: shuffle_tbl_v4i32
104 regBankSelected: true
105 tracksRegLiveness: true
110 ; CHECK-LABEL: name: shuffle_tbl_v4i32
112 ; CHECK: value: '<16 x i8> <i8 20, i8 21, i8 22, i8 23, i8 28, i8 29, i8 30, i8 31, i8 4, i8 5, i8 6, i8 7, i8 0, i8 1, i8 2, i8 3>'
113 ; CHECK: alignment: 16
114 ; CHECK: isTargetSpecific: false
115 ; CHECK: liveins: $q0, $q1
116 ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
117 ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
118 ; CHECK: [[ADRP:%[0-9]+]]:gpr64common = ADRP target-flags(aarch64-page) %const.0
119 ; CHECK: [[LDRQui:%[0-9]+]]:fpr128 = LDRQui [[ADRP]], target-flags(aarch64-pageoff, aarch64-nc) %const.0
120 ; CHECK: [[REG_SEQUENCE:%[0-9]+]]:qq = REG_SEQUENCE [[COPY]], %subreg.qsub0, [[COPY1]], %subreg.qsub1
121 ; CHECK: [[TBLv16i8Two:%[0-9]+]]:fpr128 = TBLv16i8Two [[REG_SEQUENCE]], [[LDRQui]]
122 ; CHECK: $q0 = COPY [[TBLv16i8Two]]
123 ; CHECK: RET_ReallyLR implicit $q0
124 %0:fpr(<4 x s32>) = COPY $q0
125 %1:fpr(<4 x s32>) = COPY $q1
126 %2:fpr(<4 x s32>) = G_SHUFFLE_VECTOR %0(<4 x s32>), %1, shufflemask(5, 7, 1, 0)
127 $q0 = COPY %2(<4 x s32>)
128 RET_ReallyLR implicit $q0
135 regBankSelected: true
136 tracksRegLiveness: true
141 ; CHECK-LABEL: name: shuffle_v2i64
143 ; CHECK: value: '<16 x i8> <i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>'
144 ; CHECK: alignment: 16
145 ; CHECK: isTargetSpecific: false
146 ; CHECK: liveins: $q0, $q1
147 ; CHECK: [[COPY:%[0-9]+]]:fpr128 = COPY $q0
148 ; CHECK: [[COPY1:%[0-9]+]]:fpr128 = COPY $q1
149 ; CHECK: [[ADRP:%[0-9]+]]:gpr64common = ADRP target-flags(aarch64-page) %const.0
150 ; CHECK: [[LDRQui:%[0-9]+]]:fpr128 = LDRQui [[ADRP]], target-flags(aarch64-pageoff, aarch64-nc) %const.0
151 ; CHECK: [[REG_SEQUENCE:%[0-9]+]]:qq = REG_SEQUENCE [[COPY]], %subreg.qsub0, [[COPY1]], %subreg.qsub1
152 ; CHECK: [[TBLv16i8Two:%[0-9]+]]:fpr128 = TBLv16i8Two [[REG_SEQUENCE]], [[LDRQui]]
153 ; CHECK: $q0 = COPY [[TBLv16i8Two]]
154 ; CHECK: RET_ReallyLR implicit $q0
155 %0:fpr(<2 x s64>) = COPY $q0
156 %1:fpr(<2 x s64>) = COPY $q1
157 %2:fpr(<2 x s64>) = G_SHUFFLE_VECTOR %0(<2 x s64>), %1, shufflemask(1, 0)
158 $q0 = COPY %2(<2 x s64>)
159 RET_ReallyLR implicit $q0