1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2 < %s | FileCheck %s
3 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme < %s | FileCheck %s
9 define <vscale x 8 x i16> @saddlbt_b(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
10 ; CHECK-LABEL: saddlbt_b:
12 ; CHECK-NEXT: saddlbt z0.h, z0.b, z1.b
14 %out = call <vscale x 8 x i16> @llvm.aarch64.sve.saddlbt.nxv8i16(<vscale x 16 x i8> %a,
15 <vscale x 16 x i8> %b)
16 ret <vscale x 8 x i16> %out
19 define <vscale x 4 x i32> @saddlbt_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
20 ; CHECK-LABEL: saddlbt_h:
22 ; CHECK-NEXT: saddlbt z0.s, z0.h, z1.h
24 %out = call <vscale x 4 x i32> @llvm.aarch64.sve.saddlbt.nxv4i32(<vscale x 8 x i16> %a,
25 <vscale x 8 x i16> %b)
26 ret <vscale x 4 x i32> %out
29 define <vscale x 2 x i64> @saddlbt_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
30 ; CHECK-LABEL: saddlbt_s:
32 ; CHECK-NEXT: saddlbt z0.d, z0.s, z1.s
34 %out = call <vscale x 2 x i64> @llvm.aarch64.sve.saddlbt.nxv2i64(<vscale x 4 x i32> %a,
35 <vscale x 4 x i32> %b)
36 ret <vscale x 2 x i64> %out
43 define <vscale x 8 x i16> @ssublbt_b(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
44 ; CHECK-LABEL: ssublbt_b:
46 ; CHECK-NEXT: ssublbt z0.h, z0.b, z1.b
48 %out = call <vscale x 8 x i16> @llvm.aarch64.sve.ssublbt.nxv8i16(<vscale x 16 x i8> %a,
49 <vscale x 16 x i8> %b)
50 ret <vscale x 8 x i16> %out
53 define <vscale x 4 x i32> @ssublbt_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
54 ; CHECK-LABEL: ssublbt_h:
56 ; CHECK-NEXT: ssublbt z0.s, z0.h, z1.h
58 %out = call <vscale x 4 x i32> @llvm.aarch64.sve.ssublbt.nxv4i32(<vscale x 8 x i16> %a,
59 <vscale x 8 x i16> %b)
60 ret <vscale x 4 x i32> %out
63 define <vscale x 2 x i64> @ssublbt_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
64 ; CHECK-LABEL: ssublbt_s:
66 ; CHECK-NEXT: ssublbt z0.d, z0.s, z1.s
68 %out = call <vscale x 2 x i64> @llvm.aarch64.sve.ssublbt.nxv2i64(<vscale x 4 x i32> %a,
69 <vscale x 4 x i32> %b)
70 ret <vscale x 2 x i64> %out
77 define <vscale x 8 x i16> @ssubltb_b(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b) {
78 ; CHECK-LABEL: ssubltb_b:
80 ; CHECK-NEXT: ssubltb z0.h, z0.b, z1.b
82 %out = call <vscale x 8 x i16> @llvm.aarch64.sve.ssubltb.nxv8i16(<vscale x 16 x i8> %a,
83 <vscale x 16 x i8> %b)
84 ret <vscale x 8 x i16> %out
87 define <vscale x 4 x i32> @ssubltb_h(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b) {
88 ; CHECK-LABEL: ssubltb_h:
90 ; CHECK-NEXT: ssubltb z0.s, z0.h, z1.h
92 %out = call <vscale x 4 x i32> @llvm.aarch64.sve.ssubltb.nxv4i32(<vscale x 8 x i16> %a,
93 <vscale x 8 x i16> %b)
94 ret <vscale x 4 x i32> %out
97 define <vscale x 2 x i64> @ssubltb_s(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b) {
98 ; CHECK-LABEL: ssubltb_s:
100 ; CHECK-NEXT: ssubltb z0.d, z0.s, z1.s
102 %out = call <vscale x 2 x i64> @llvm.aarch64.sve.ssubltb.nxv2i64(<vscale x 4 x i32> %a,
103 <vscale x 4 x i32> %b)
104 ret <vscale x 2 x i64> %out
107 declare <vscale x 8 x i16> @llvm.aarch64.sve.saddlbt.nxv8i16(<vscale x 16 x i8>, <vscale x 16 x i8>)
108 declare <vscale x 4 x i32> @llvm.aarch64.sve.saddlbt.nxv4i32(<vscale x 8 x i16>, <vscale x 8 x i16>)
109 declare <vscale x 2 x i64> @llvm.aarch64.sve.saddlbt.nxv2i64(<vscale x 4 x i32>, <vscale x 4 x i32>)
111 declare <vscale x 8 x i16> @llvm.aarch64.sve.ssublbt.nxv8i16(<vscale x 16 x i8>, <vscale x 16 x i8>)
112 declare <vscale x 4 x i32> @llvm.aarch64.sve.ssublbt.nxv4i32(<vscale x 8 x i16>, <vscale x 8 x i16>)
113 declare <vscale x 2 x i64> @llvm.aarch64.sve.ssublbt.nxv2i64(<vscale x 4 x i32>, <vscale x 4 x i32>)
115 declare <vscale x 8 x i16> @llvm.aarch64.sve.ssubltb.nxv8i16(<vscale x 16 x i8>, <vscale x 16 x i8>)
116 declare <vscale x 4 x i32> @llvm.aarch64.sve.ssubltb.nxv4i32(<vscale x 8 x i16>, <vscale x 8 x i16>)
117 declare <vscale x 2 x i64> @llvm.aarch64.sve.ssubltb.nxv2i64(<vscale x 4 x i32>, <vscale x 4 x i32>)