xref: /llvm-project/llvm/test/CodeGen/AArch64/complex-deinterleaving-f16-add.ll (revision bfc0317153dca75137fba00b5c28758d6f720963) 
1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc < %s --mattr=+complxnum,+neon,+fullfp16 -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
3; RUN: llc < %s --mattr=+complxnum,+neon,+fullfp16,+sve -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
4; RUN: llc < %s --mattr=+complxnum,+neon,+fullfp16,+sve2 -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
5; RUN: llc < %s --global-isel --global-isel-abort=2 --mattr=+complxnum,+neon,+fullfp16 -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
6; RUN: llc < %s --global-isel --global-isel-abort=2 --mattr=+complxnum,+neon,+fullfp16,+sve -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
7; RUN: llc < %s --global-isel --global-isel-abort=2 --mattr=+complxnum,+neon,+fullfp16,+sve2 -o - 2>&1 | FileCheck %s --check-prefixes=CHECK,CHECK-GI
8
9target triple = "aarch64"
10
11; CHECK-GI:      warning: Instruction selection used fallback path for complex_add_v16f16
12; CHECK-GI-NEXT: warning: Instruction selection used fallback path for complex_add_v32f16
13; CHECK-GI-NEXT: warning: Instruction selection used fallback path for complex_add_v16f16_with_intrinsic
14
15; Expected to not transform
16define <2 x half> @complex_add_v2f16(<2 x half> %a, <2 x half> %b) {
17; CHECK-SD-LABEL: complex_add_v2f16:
18; CHECK-SD:       // %bb.0: // %entry
19; CHECK-SD-NEXT:    // kill: def $d1 killed $d1 def $q1
20; CHECK-SD-NEXT:    // kill: def $d0 killed $d0 def $q0
21; CHECK-SD-NEXT:    mov h2, v0.h[1]
22; CHECK-SD-NEXT:    mov h3, v1.h[1]
23; CHECK-SD-NEXT:    fsub h1, h1, h2
24; CHECK-SD-NEXT:    fadd h0, h3, h0
25; CHECK-SD-NEXT:    mov v1.h[1], v0.h[0]
26; CHECK-SD-NEXT:    fmov d0, d1
27; CHECK-SD-NEXT:    ret
28;
29; CHECK-GI-LABEL: complex_add_v2f16:
30; CHECK-GI:       // %bb.0: // %entry
31; CHECK-GI-NEXT:    // kill: def $d0 killed $d0 def $q0
32; CHECK-GI-NEXT:    // kill: def $d1 killed $d1 def $q1
33; CHECK-GI-NEXT:    mov h2, v0.h[1]
34; CHECK-GI-NEXT:    mov h3, v1.h[1]
35; CHECK-GI-NEXT:    fsub h1, h1, h2
36; CHECK-GI-NEXT:    fadd h0, h3, h0
37; CHECK-GI-NEXT:    mov v1.h[1], v0.h[0]
38; CHECK-GI-NEXT:    fmov d0, d1
39; CHECK-GI-NEXT:    ret
40entry:
41  %a.real = shufflevector <2 x half> %a, <2 x half> zeroinitializer, <1 x i32> <i32 0>
42  %a.imag = shufflevector <2 x half> %a, <2 x half> zeroinitializer, <1 x i32> <i32 1>
43  %b.real = shufflevector <2 x half> %b, <2 x half> zeroinitializer, <1 x i32> <i32 0>
44  %b.imag = shufflevector <2 x half> %b, <2 x half> zeroinitializer, <1 x i32> <i32 1>
45  %0 = fsub fast <1 x half> %b.real, %a.imag
46  %1 = fadd fast <1 x half> %b.imag, %a.real
47  %interleaved.vec = shufflevector <1 x half> %0, <1 x half> %1, <2 x i32> <i32 0, i32 1>
48  ret <2 x half> %interleaved.vec
49}
50
51; Expected to transform
52define <4 x half> @complex_add_v4f16(<4 x half> %a, <4 x half> %b) {
53; CHECK-LABEL: complex_add_v4f16:
54; CHECK:       // %bb.0: // %entry
55; CHECK-NEXT:    fcadd v0.4h, v1.4h, v0.4h, #90
56; CHECK-NEXT:    ret
57entry:
58  %a.real = shufflevector <4 x half> %a, <4 x half> zeroinitializer, <2 x i32> <i32 0, i32 2>
59  %a.imag = shufflevector <4 x half> %a, <4 x half> zeroinitializer, <2 x i32> <i32 1, i32 3>
60  %b.real = shufflevector <4 x half> %b, <4 x half> zeroinitializer, <2 x i32> <i32 0, i32 2>
61  %b.imag = shufflevector <4 x half> %b, <4 x half> zeroinitializer, <2 x i32> <i32 1, i32 3>
62  %0 = fsub fast <2 x half> %b.real, %a.imag
63  %1 = fadd fast <2 x half> %b.imag, %a.real
64  %interleaved.vec = shufflevector <2 x half> %0, <2 x half> %1, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
65  ret <4 x half> %interleaved.vec
66}
67
68; Expected to transform
69define <8 x half> @complex_add_v8f16(<8 x half> %a, <8 x half> %b) {
70; CHECK-LABEL: complex_add_v8f16:
71; CHECK:       // %bb.0: // %entry
72; CHECK-NEXT:    fcadd v0.8h, v1.8h, v0.8h, #90
73; CHECK-NEXT:    ret
74entry:
75  %a.real = shufflevector <8 x half> %a, <8 x half> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
76  %a.imag = shufflevector <8 x half> %a, <8 x half> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
77  %b.real = shufflevector <8 x half> %b, <8 x half> zeroinitializer, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
78  %b.imag = shufflevector <8 x half> %b, <8 x half> zeroinitializer, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
79  %0 = fsub fast <4 x half> %b.real, %a.imag
80  %1 = fadd fast <4 x half> %b.imag, %a.real
81  %interleaved.vec = shufflevector <4 x half> %0, <4 x half> %1, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
82  ret <8 x half> %interleaved.vec
83}
84
85; Expected to transform
86define <16 x half> @complex_add_v16f16(<16 x half> %a, <16 x half> %b) {
87; CHECK-LABEL: complex_add_v16f16:
88; CHECK:       // %bb.0: // %entry
89; CHECK-NEXT:    fcadd v1.8h, v3.8h, v1.8h, #90
90; CHECK-NEXT:    fcadd v0.8h, v2.8h, v0.8h, #90
91; CHECK-NEXT:    ret
92entry:
93  %a.real = shufflevector <16 x half> %a, <16 x half> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
94  %a.imag = shufflevector <16 x half> %a, <16 x half> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
95  %b.real = shufflevector <16 x half> %b, <16 x half> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
96  %b.imag = shufflevector <16 x half> %b, <16 x half> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
97  %0 = fsub fast <8 x half> %b.real, %a.imag
98  %1 = fadd fast <8 x half> %b.imag, %a.real
99  %interleaved.vec = shufflevector <8 x half> %0, <8 x half> %1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
100  ret <16 x half> %interleaved.vec
101}
102
103; Expected to transform
104define <32 x half> @complex_add_v32f16(<32 x half> %a, <32 x half> %b) {
105; CHECK-LABEL: complex_add_v32f16:
106; CHECK:       // %bb.0: // %entry
107; CHECK-NEXT:    fcadd v2.8h, v6.8h, v2.8h, #90
108; CHECK-NEXT:    fcadd v0.8h, v4.8h, v0.8h, #90
109; CHECK-NEXT:    fcadd v1.8h, v5.8h, v1.8h, #90
110; CHECK-NEXT:    fcadd v3.8h, v7.8h, v3.8h, #90
111; CHECK-NEXT:    ret
112entry:
113  %a.real = shufflevector <32 x half> %a, <32 x half> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
114  %a.imag = shufflevector <32 x half> %a, <32 x half> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
115  %b.real = shufflevector <32 x half> %b, <32 x half> zeroinitializer, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
116  %b.imag = shufflevector <32 x half> %b, <32 x half> zeroinitializer, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
117  %0 = fsub fast <16 x half> %b.real, %a.imag
118  %1 = fadd fast <16 x half> %b.imag, %a.real
119  %interleaved.vec = shufflevector <16 x half> %0, <16 x half> %1, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
120  ret <32 x half> %interleaved.vec
121}
122
123; Expected to transform
124define <4 x half> @complex_add_v4f16_with_intrinsic(<4 x half> %a, <4 x half> %b) {
125; CHECK-LABEL: complex_add_v4f16_with_intrinsic:
126; CHECK:       // %bb.0: // %entry
127; CHECK-NEXT:    fcadd v0.4h, v1.4h, v0.4h, #90
128; CHECK-NEXT:    ret
129entry:
130  %a.deinterleaved = tail call { <2 x half>, <2 x half> } @llvm.vector.deinterleave2.v4f16(<4 x half> %a)
131  %a.real = extractvalue { <2 x half>, <2 x half> } %a.deinterleaved, 0
132  %a.imag = extractvalue { <2 x half>, <2 x half> } %a.deinterleaved, 1
133  %b.deinterleaved = tail call { <2 x half>, <2 x half> } @llvm.vector.deinterleave2.v4f16(<4 x half> %b)
134  %b.real = extractvalue { <2 x half>, <2 x half> } %b.deinterleaved, 0
135  %b.imag = extractvalue { <2 x half>, <2 x half> } %b.deinterleaved, 1
136  %0 = fsub fast <2 x half> %b.real, %a.imag
137  %1 = fadd fast <2 x half> %b.imag, %a.real
138  %interleaved.vec = tail call <4 x half> @llvm.vector.interleave2.v4f16(<2 x half> %0, <2 x half> %1)
139  ret <4 x half> %interleaved.vec
140}
141
142; Expected to transform
143define <8 x half> @complex_add_v8f16_with_intrinsic(<8 x half> %a, <8 x half> %b) {
144; CHECK-LABEL: complex_add_v8f16_with_intrinsic:
145; CHECK:       // %bb.0: // %entry
146; CHECK-NEXT:    fcadd v0.8h, v1.8h, v0.8h, #90
147; CHECK-NEXT:    ret
148entry:
149  %a.deinterleaved = tail call { <4 x half>, <4 x half> } @llvm.vector.deinterleave2.v8f16(<8 x half> %a)
150  %a.real = extractvalue { <4 x half>, <4 x half> } %a.deinterleaved, 0
151  %a.imag = extractvalue { <4 x half>, <4 x half> } %a.deinterleaved, 1
152  %b.deinterleaved = tail call { <4 x half>, <4 x half> } @llvm.vector.deinterleave2.v8f16(<8 x half> %b)
153  %b.real = extractvalue { <4 x half>, <4 x half> } %b.deinterleaved, 0
154  %b.imag = extractvalue { <4 x half>, <4 x half> } %b.deinterleaved, 1
155  %0 = fsub fast <4 x half> %b.real, %a.imag
156  %1 = fadd fast <4 x half> %b.imag, %a.real
157  %interleaved.vec = tail call <8 x half> @llvm.vector.interleave2.v8f16(<4 x half> %0, <4 x half> %1)
158  ret <8 x half> %interleaved.vec
159}
160
161; Expected to transform
162define <16 x half> @complex_add_v16f16_with_intrinsic(<16 x half> %a, <16 x half> %b) {
163; CHECK-LABEL: complex_add_v16f16_with_intrinsic:
164; CHECK:       // %bb.0: // %entry
165; CHECK-NEXT:    fcadd v1.8h, v3.8h, v1.8h, #90
166; CHECK-NEXT:    fcadd v0.8h, v2.8h, v0.8h, #90
167; CHECK-NEXT:    ret
168entry:
169  %a.deinterleaved = tail call { <8 x half>, <8 x half> } @llvm.vector.deinterleave2.v16f16(<16 x half> %a)
170  %a.real = extractvalue { <8 x half>, <8 x half> } %a.deinterleaved, 0
171  %a.imag = extractvalue { <8 x half>, <8 x half> } %a.deinterleaved, 1
172  %b.deinterleaved = tail call { <8 x half>, <8 x half> } @llvm.vector.deinterleave2.v16f16(<16 x half> %b)
173  %b.real = extractvalue { <8 x half>, <8 x half> } %b.deinterleaved, 0
174  %b.imag = extractvalue { <8 x half>, <8 x half> } %b.deinterleaved, 1
175  %0 = fsub fast <8 x half> %b.real, %a.imag
176  %1 = fadd fast <8 x half> %b.imag, %a.real
177  %interleaved.vec = tail call <16 x half> @llvm.vector.interleave2.v16f16(<8 x half> %0, <8 x half> %1)
178  ret <16 x half> %interleaved.vec
179}
180
181
182; Expected not to transform as it is integer
183define <16 x i16> @complex_add_v16i16(<16 x i16> %a, <16 x i16> %b) {
184; CHECK-SD-LABEL: complex_add_v16i16:
185; CHECK-SD:       // %bb.0: // %entry
186; CHECK-SD-NEXT:    uzp1 v4.8h, v2.8h, v3.8h
187; CHECK-SD-NEXT:    uzp1 v5.8h, v0.8h, v1.8h
188; CHECK-SD-NEXT:    uzp2 v0.8h, v0.8h, v1.8h
189; CHECK-SD-NEXT:    uzp2 v1.8h, v2.8h, v3.8h
190; CHECK-SD-NEXT:    sub v2.8h, v4.8h, v0.8h
191; CHECK-SD-NEXT:    add v1.8h, v1.8h, v5.8h
192; CHECK-SD-NEXT:    zip1 v0.8h, v2.8h, v1.8h
193; CHECK-SD-NEXT:    zip2 v1.8h, v2.8h, v1.8h
194; CHECK-SD-NEXT:    ret
195;
196; CHECK-GI-LABEL: complex_add_v16i16:
197; CHECK-GI:       // %bb.0: // %entry
198; CHECK-GI-NEXT:    uzp1 v4.8h, v0.8h, v1.8h
199; CHECK-GI-NEXT:    uzp2 v0.8h, v0.8h, v1.8h
200; CHECK-GI-NEXT:    uzp1 v1.8h, v2.8h, v3.8h
201; CHECK-GI-NEXT:    uzp2 v2.8h, v2.8h, v3.8h
202; CHECK-GI-NEXT:    sub v1.8h, v1.8h, v0.8h
203; CHECK-GI-NEXT:    add v2.8h, v2.8h, v4.8h
204; CHECK-GI-NEXT:    zip1 v0.8h, v1.8h, v2.8h
205; CHECK-GI-NEXT:    zip2 v1.8h, v1.8h, v2.8h
206; CHECK-GI-NEXT:    ret
207entry:
208  %a.real = shufflevector <16 x i16> %a, <16 x i16> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
209  %a.imag = shufflevector <16 x i16> %a, <16 x i16> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
210  %b.real = shufflevector <16 x i16> %b, <16 x i16> zeroinitializer, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
211  %b.imag = shufflevector <16 x i16> %b, <16 x i16> zeroinitializer, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
212  %0 = sub <8 x i16> %b.real, %a.imag
213  %1 = add <8 x i16> %b.imag, %a.real
214  %interleaved.vec = shufflevector <8 x i16> %0, <8 x i16> %1, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
215  ret <16 x i16> %interleaved.vec
216}
217
218
219declare { <2 x half>, <2 x half> } @llvm.vector.deinterleave2.v4f16(<4 x half>)
220declare <4 x half> @llvm.vector.interleave2.v4f16(<2 x half>, <2 x half>)
221
222declare { <4 x half>, <4 x half> } @llvm.vector.deinterleave2.v8f16(<8 x half>)
223declare <8 x half> @llvm.vector.interleave2.v8f16(<4 x half>, <4 x half>)
224
225declare { <8 x half>, <8 x half> } @llvm.vector.deinterleave2.v16f16(<16 x half>)
226declare <16 x half> @llvm.vector.interleave2.v16f16(<8 x half>, <8 x half>)
227