1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
2 ; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sve2p1 -mattr=+sve-b16b16 < %s | FileCheck %s
4 ; Replace pattern min(max(v1,v2),v3) by clamp
6 define <vscale x 16 x i8> @uclampi8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b, <vscale x 16 x i8> %c) {
7 ; CHECK-LABEL: uclampi8:
9 ; CHECK-NEXT: uclamp z0.b, z1.b, z2.b
11 %min = tail call <vscale x 16 x i8> @llvm.umax.nxv16i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
12 %res = tail call <vscale x 16 x i8> @llvm.umin.nxv16i8(<vscale x 16 x i8> %min, <vscale x 16 x i8> %c)
13 ret <vscale x 16 x i8> %res
16 define <vscale x 8 x i16> @uclampi16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
17 ; CHECK-LABEL: uclampi16:
19 ; CHECK-NEXT: uclamp z0.h, z1.h, z2.h
21 %min = tail call <vscale x 8 x i16> @llvm.umax.nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b)
22 %res = tail call <vscale x 8 x i16> @llvm.umin.nxv8i16(<vscale x 8 x i16> %min, <vscale x 8 x i16> %c)
23 ret <vscale x 8 x i16> %res
26 define <vscale x 4 x i32> @uclampi32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
27 ; CHECK-LABEL: uclampi32:
29 ; CHECK-NEXT: uclamp z0.s, z1.s, z2.s
31 %min = tail call <vscale x 4 x i32> @llvm.umax.nxv4i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b)
32 %res = tail call <vscale x 4 x i32> @llvm.umin.nxv4i32(<vscale x 4 x i32> %min, <vscale x 4 x i32> %c)
33 ret <vscale x 4 x i32> %res
36 define <vscale x 2 x i64> @uclampi64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
37 ; CHECK-LABEL: uclampi64:
39 ; CHECK-NEXT: uclamp z0.d, z1.d, z2.d
41 %min = tail call <vscale x 2 x i64> @llvm.umax.nxv2i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b)
42 %res = tail call <vscale x 2 x i64> @llvm.umin.nxv2i64(<vscale x 2 x i64> %min, <vscale x 2 x i64> %c)
43 ret <vscale x 2 x i64> %res
46 define <vscale x 16 x i8> @sclampi8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b, <vscale x 16 x i8> %c) {
47 ; CHECK-LABEL: sclampi8:
49 ; CHECK-NEXT: sclamp z0.b, z1.b, z2.b
51 %min = tail call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> %a, <vscale x 16 x i8> %b)
52 %res = tail call <vscale x 16 x i8> @llvm.smin.nxv16i8(<vscale x 16 x i8> %min, <vscale x 16 x i8> %c)
53 ret <vscale x 16 x i8> %res
56 define <vscale x 8 x i16> @sclampi16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b, <vscale x 8 x i16> %c) {
57 ; CHECK-LABEL: sclampi16:
59 ; CHECK-NEXT: sclamp z0.h, z1.h, z2.h
61 %min = tail call <vscale x 8 x i16> @llvm.smax.nxv8i16(<vscale x 8 x i16> %a, <vscale x 8 x i16> %b)
62 %res = tail call <vscale x 8 x i16> @llvm.smin.nxv8i16(<vscale x 8 x i16> %min, <vscale x 8 x i16> %c)
63 ret <vscale x 8 x i16> %res
66 define <vscale x 4 x i32> @sclampi32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b, <vscale x 4 x i32> %c) {
67 ; CHECK-LABEL: sclampi32:
69 ; CHECK-NEXT: sclamp z0.s, z1.s, z2.s
71 %min = tail call <vscale x 4 x i32> @llvm.smax.nxv4i32(<vscale x 4 x i32> %a, <vscale x 4 x i32> %b)
72 %res = tail call <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32> %min, <vscale x 4 x i32> %c)
73 ret <vscale x 4 x i32> %res
76 define <vscale x 2 x i64> @sclampi64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b, <vscale x 2 x i64> %c) {
77 ; CHECK-LABEL: sclampi64:
79 ; CHECK-NEXT: sclamp z0.d, z1.d, z2.d
81 %min = tail call <vscale x 2 x i64> @llvm.smax.nxv2i64(<vscale x 2 x i64> %a, <vscale x 2 x i64> %b)
82 %res = tail call <vscale x 2 x i64> @llvm.smin.nxv2i64(<vscale x 2 x i64> %min, <vscale x 2 x i64> %c)
83 ret <vscale x 2 x i64> %res
86 define <vscale x 8 x bfloat> @fclampbf16(<vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b, <vscale x 8 x bfloat> %c) {
87 ; CHECK-LABEL: fclampbf16:
89 ; CHECK-NEXT: bfclamp z0.h, z1.h, z2.h
91 %min = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fmaxnm.u.nxv8bf16(<vscale x 8 x i1> splat (i1 true), <vscale x 8 x bfloat> %a, <vscale x 8 x bfloat> %b)
92 %res = tail call <vscale x 8 x bfloat> @llvm.aarch64.sve.fminnm.u.nxv8bf16(<vscale x 8 x i1> splat (i1 true), <vscale x 8 x bfloat> %min, <vscale x 8 x bfloat> %c)
93 ret <vscale x 8 x bfloat> %res
96 define <vscale x 8 x half> @fclampf16(<vscale x 8 x half> %a, <vscale x 8 x half> %b, <vscale x 8 x half> %c) {
97 ; CHECK-LABEL: fclampf16:
99 ; CHECK-NEXT: fclamp z0.h, z1.h, z2.h
101 %min = call <vscale x 8 x half> @llvm.maxnum.nxv8f16(<vscale x 8 x half> %a, <vscale x 8 x half> %b)
102 %res = call <vscale x 8 x half> @llvm.minnum.nxv8f16(<vscale x 8 x half> %min, <vscale x 8 x half> %c)
103 ret <vscale x 8 x half> %res
106 define <vscale x 4 x float> @fclampf32(<vscale x 4 x float> %a, <vscale x 4 x float> %b, <vscale x 4 x float> %c) {
107 ; CHECK-LABEL: fclampf32:
109 ; CHECK-NEXT: fclamp z0.s, z1.s, z2.s
111 %min = tail call <vscale x 4 x float> @llvm.maxnum.nxv4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b)
112 %res = tail call <vscale x 4 x float> @llvm.minnum.nxv4f32(<vscale x 4 x float> %min, <vscale x 4 x float> %c)
113 ret <vscale x 4 x float> %res
116 define <vscale x 2 x double> @fclampf64(<vscale x 2 x double> %a, <vscale x 2 x double> %b, <vscale x 2 x double> %c) {
117 ; CHECK-LABEL: fclampf64:
119 ; CHECK-NEXT: fclamp z0.d, z1.d, z2.d
121 %min = tail call <vscale x 2 x double> @llvm.maxnum.nxv2f64(<vscale x 2 x double> %a, <vscale x 2 x double> %b)
122 %res = tail call <vscale x 2 x double> @llvm.minnum.nxv2f64(<vscale x 2 x double> %min, <vscale x 2 x double> %c)
123 ret <vscale x 2 x double> %res
126 declare <vscale x 16 x i8> @llvm.umax.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
127 declare <vscale x 16 x i8> @llvm.umin.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
128 declare <vscale x 8 x i16> @llvm.umax.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
129 declare <vscale x 8 x i16> @llvm.umin.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
130 declare <vscale x 4 x i32> @llvm.umax.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
131 declare <vscale x 4 x i32> @llvm.umin.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
132 declare <vscale x 2 x i64> @llvm.umax.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
133 declare <vscale x 2 x i64> @llvm.umin.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
135 declare <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
136 declare <vscale x 16 x i8> @llvm.smin.nxv16i8(<vscale x 16 x i8>, <vscale x 16 x i8>)
137 declare <vscale x 8 x i16> @llvm.smax.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
138 declare <vscale x 8 x i16> @llvm.smin.nxv8i16(<vscale x 8 x i16>, <vscale x 8 x i16>)
139 declare <vscale x 4 x i32> @llvm.smax.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
140 declare <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32>, <vscale x 4 x i32>)
141 declare <vscale x 2 x i64> @llvm.smax.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
142 declare <vscale x 2 x i64> @llvm.smin.nxv2i64(<vscale x 2 x i64>, <vscale x 2 x i64>)
144 declare <vscale x 8 x bfloat> @llvm.aarch64.sve.fmaxnm.u.nxv8bf16(<vscale x 8 x i1>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>)
145 declare <vscale x 8 x bfloat> @llvm.aarch64.sve.fminnm.u.nxv8bf16(<vscale x 8 x i1>, <vscale x 8 x bfloat>, <vscale x 8 x bfloat>)
146 declare <vscale x 8 x half> @llvm.maxnum.nxv8f16 (<vscale x 8 x half>, <vscale x 8 x half>)
147 declare <vscale x 8 x half> @llvm.minnum.nxv8f16 (<vscale x 8 x half>, <vscale x 8 x half>)
148 declare <vscale x 4 x float> @llvm.maxnum.nxv4f32 (<vscale x 4 x float>, <vscale x 4 x float>)
149 declare <vscale x 4 x float> @llvm.minnum.nxv4f32 (<vscale x 4 x float>, <vscale x 4 x float>)
150 declare <vscale x 2 x double> @llvm.maxnum.nxv2f64 (<vscale x 2 x double>, <vscale x 2 x double>)
151 declare <vscale x 2 x double> @llvm.minnum.nxv2f64 (<vscale x 2 x double>, <vscale x 2 x double>)