1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py$
2 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2 -force-streaming -verify-machineinstrs < %s | FileCheck %s
9 define <vscale x 8 x i16 > @multi_vector_qcvt_x2_s16_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2) {
10 ; CHECK-LABEL: multi_vector_qcvt_x2_s16_s32:
12 ; CHECK-NEXT: mov z3.d, z2.d
13 ; CHECK-NEXT: mov z2.d, z1.d
14 ; CHECK-NEXT: sqcvt z0.h, { z2.s, z3.s }
16 %res = call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x2.nxv4i32(<vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2)
17 ret <vscale x 8 x i16> %res
21 define <vscale x 16 x i8 > @multi_vector_qcvt_x4_s8_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2, <vscale x 4 x i32> %zn3, <vscale x 4 x i32> %zn4) {
22 ; CHECK-LABEL: multi_vector_qcvt_x4_s8_s32:
24 ; CHECK-NEXT: mov z7.d, z4.d
25 ; CHECK-NEXT: mov z6.d, z3.d
26 ; CHECK-NEXT: mov z5.d, z2.d
27 ; CHECK-NEXT: mov z4.d, z1.d
28 ; CHECK-NEXT: sqcvt z0.b, { z4.s - z7.s }
30 %res = call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvt.x4.nxv4i32(<vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2, <vscale x 4 x i32> %zn3, <vscale x 4 x i32> %zn4)
31 ret <vscale x 16 x i8> %res
34 define <vscale x 8 x i16> @multi_vector_qcvt_x4_s16_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zn1, <vscale x 2 x i64> %zn2, <vscale x 2 x i64> %zn3, <vscale x 2 x i64> %zn4) {
35 ; CHECK-LABEL: multi_vector_qcvt_x4_s16_s64:
37 ; CHECK-NEXT: mov z7.d, z4.d
38 ; CHECK-NEXT: mov z6.d, z3.d
39 ; CHECK-NEXT: mov z5.d, z2.d
40 ; CHECK-NEXT: mov z4.d, z1.d
41 ; CHECK-NEXT: sqcvt z0.h, { z4.d - z7.d }
43 %res = call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x4.nxv2i64(<vscale x 2 x i64> %zn1, <vscale x 2 x i64> %zn2, <vscale x 2 x i64> %zn3, <vscale x 2 x i64> %zn4)
44 ret <vscale x 8 x i16> %res
52 define <vscale x 8 x i16> @multi_vector_qcvt_x2_u16_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zn0, <vscale x 4 x i32> %zn1) {
53 ; CHECK-LABEL: multi_vector_qcvt_x2_u16_u32:
55 ; CHECK-NEXT: mov z3.d, z2.d
56 ; CHECK-NEXT: mov z2.d, z1.d
57 ; CHECK-NEXT: uqcvt z0.h, { z2.s, z3.s }
59 %res = call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x2.nxv4i32(<vscale x 4 x i32> %zn0, <vscale x 4 x i32> %zn1)
60 ret<vscale x 8 x i16> %res
64 define <vscale x 16 x i8> @multi_vector_qcvt_x4_u8_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2, <vscale x 4 x i32> %zn3, <vscale x 4 x i32> %zn4) {
65 ; CHECK-LABEL: multi_vector_qcvt_x4_u8_u32:
67 ; CHECK-NEXT: mov z7.d, z4.d
68 ; CHECK-NEXT: mov z6.d, z3.d
69 ; CHECK-NEXT: mov z5.d, z2.d
70 ; CHECK-NEXT: mov z4.d, z1.d
71 ; CHECK-NEXT: uqcvt z0.b, { z4.s - z7.s }
73 %res = call <vscale x 16 x i8> @llvm.aarch64.sve.uqcvt.x4.nxv4i32(<vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2, <vscale x 4 x i32> %zn3, <vscale x 4 x i32> %zn4)
74 ret <vscale x 16 x i8> %res
77 define <vscale x 8 x i16> @multi_vector_qcvt_x4_u16_u64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zn1, <vscale x 2 x i64> %zn2, <vscale x 2 x i64> %zn3, <vscale x 2 x i64> %zn4) {
78 ; CHECK-LABEL: multi_vector_qcvt_x4_u16_u64:
80 ; CHECK-NEXT: mov z7.d, z4.d
81 ; CHECK-NEXT: mov z6.d, z3.d
82 ; CHECK-NEXT: mov z5.d, z2.d
83 ; CHECK-NEXT: mov z4.d, z1.d
84 ; CHECK-NEXT: uqcvt z0.h, { z4.d - z7.d }
86 %res = call <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x4.nxv2i64(<vscale x 2 x i64> %zn1, <vscale x 2 x i64> %zn2, <vscale x 2 x i64> %zn3, <vscale x 2 x i64> %zn4)
87 ret <vscale x 8 x i16> %res
95 define <vscale x 8 x i16 > @multi_vector_qcvt_x2_s16_u32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2) {
96 ; CHECK-LABEL: multi_vector_qcvt_x2_s16_u32:
98 ; CHECK-NEXT: mov z3.d, z2.d
99 ; CHECK-NEXT: mov z2.d, z1.d
100 ; CHECK-NEXT: sqcvtu z0.h, { z2.s, z3.s }
102 %res = call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x2.nxv4i32(<vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2)
103 ret <vscale x 8 x i16> %res
107 define <vscale x 16 x i8> @multi_vector_qcvt_x4_u8_s32(<vscale x 4 x i32> %unused, <vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2, <vscale x 4 x i32> %zn3, <vscale x 4 x i32> %zn4) {
108 ; CHECK-LABEL: multi_vector_qcvt_x4_u8_s32:
110 ; CHECK-NEXT: mov z7.d, z4.d
111 ; CHECK-NEXT: mov z6.d, z3.d
112 ; CHECK-NEXT: mov z5.d, z2.d
113 ; CHECK-NEXT: mov z4.d, z1.d
114 ; CHECK-NEXT: sqcvtu z0.b, { z4.s - z7.s }
116 %res = call <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtu.x4.nxv4i32(<vscale x 4 x i32> %zn1, <vscale x 4 x i32> %zn2, <vscale x 4 x i32> %zn3, <vscale x 4 x i32> %zn4)
117 ret <vscale x 16 x i8> %res
120 define <vscale x 8 x i16> @multi_vector_qcvt_x4_u16_s64(<vscale x 2 x i64> %unused, <vscale x 2 x i64> %zn1, <vscale x 2 x i64> %zn2, <vscale x 2 x i64> %zn3, <vscale x 2 x i64> %zn4) {
121 ; CHECK-LABEL: multi_vector_qcvt_x4_u16_s64:
123 ; CHECK-NEXT: mov z7.d, z4.d
124 ; CHECK-NEXT: mov z6.d, z3.d
125 ; CHECK-NEXT: mov z5.d, z2.d
126 ; CHECK-NEXT: mov z4.d, z1.d
127 ; CHECK-NEXT: sqcvtu z0.h, { z4.d - z7.d }
129 %res = call <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x4.nxv2i64(<vscale x 2 x i64> %zn1, <vscale x 2 x i64> %zn2, <vscale x 2 x i64> %zn3, <vscale x 2 x i64> %zn4)
130 ret <vscale x 8 x i16> %res
133 declare <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x2.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
134 declare <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x2.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
135 declare <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x2.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
136 declare <vscale x 16 x i8> @llvm.aarch64.sve.sqcvt.x4.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>)
137 declare <vscale x 8 x i16> @llvm.aarch64.sve.sqcvt.x4.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>)
138 declare <vscale x 16 x i8> @llvm.aarch64.sve.uqcvt.x4.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>)
139 declare <vscale x 8 x i16> @llvm.aarch64.sve.uqcvt.x4.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>)
140 declare <vscale x 16 x i8> @llvm.aarch64.sve.sqcvtu.x4.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>, <vscale x 4 x i32>)
141 declare <vscale x 8 x i16> @llvm.aarch64.sve.sqcvtu.x4.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>, <vscale x 2 x i64>)