1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc < %s --mattr=+sve -o - | FileCheck %s
4 target triple = "aarch64"
6 ; Expected to transform
7 define <vscale x 4 x float> @complex_mul_v4f32(<vscale x 4 x float> %a, <vscale x 4 x float> %b) {
8 ; CHECK-LABEL: complex_mul_v4f32:
9 ; CHECK: // %bb.0: // %entry
10 ; CHECK-NEXT: mov z2.s, #0 // =0x0
11 ; CHECK-NEXT: ptrue p0.s
12 ; CHECK-NEXT: fcmla z2.s, p0/m, z1.s, z0.s, #0
13 ; CHECK-NEXT: fcmla z2.s, p0/m, z1.s, z0.s, #90
14 ; CHECK-NEXT: mov z0.d, z2.d
17 %a.deinterleaved = tail call { <vscale x 2 x float>, <vscale x 2 x float> } @llvm.vector.deinterleave2.nxv4f32(<vscale x 4 x float> %a)
18 %a.real = extractvalue { <vscale x 2 x float>, <vscale x 2 x float> } %a.deinterleaved, 0
19 %a.imag = extractvalue { <vscale x 2 x float>, <vscale x 2 x float> } %a.deinterleaved, 1
20 %b.deinterleaved = tail call { <vscale x 2 x float>, <vscale x 2 x float> } @llvm.vector.deinterleave2.nxv4f32(<vscale x 4 x float> %b)
21 %b.real = extractvalue { <vscale x 2 x float>, <vscale x 2 x float> } %b.deinterleaved, 0
22 %b.imag = extractvalue { <vscale x 2 x float>, <vscale x 2 x float> } %b.deinterleaved, 1
23 %0 = fmul fast <vscale x 2 x float> %b.imag, %a.real
24 %1 = fmul fast <vscale x 2 x float> %b.real, %a.imag
25 %2 = fadd fast <vscale x 2 x float> %1, %0
26 %3 = fmul fast <vscale x 2 x float> %b.real, %a.real
27 %4 = fmul fast <vscale x 2 x float> %a.imag, %b.imag
28 %5 = fsub fast <vscale x 2 x float> %3, %4
29 %interleaved.vec = tail call <vscale x 4 x float> @llvm.vector.interleave2.nxv4f32(<vscale x 2 x float> %5, <vscale x 2 x float> %2)
30 ret <vscale x 4 x float> %interleaved.vec
33 ; Expected to transform
34 define <vscale x 8 x float> @complex_mul_v8f32(<vscale x 8 x float> %a, <vscale x 8 x float> %b) {
35 ; CHECK-LABEL: complex_mul_v8f32:
36 ; CHECK: // %bb.0: // %entry
37 ; CHECK-NEXT: mov z4.s, #0 // =0x0
38 ; CHECK-NEXT: ptrue p0.s
39 ; CHECK-NEXT: mov z5.d, z4.d
40 ; CHECK-NEXT: fcmla z4.s, p0/m, z3.s, z1.s, #0
41 ; CHECK-NEXT: fcmla z5.s, p0/m, z2.s, z0.s, #0
42 ; CHECK-NEXT: fcmla z4.s, p0/m, z3.s, z1.s, #90
43 ; CHECK-NEXT: fcmla z5.s, p0/m, z2.s, z0.s, #90
44 ; CHECK-NEXT: mov z1.d, z4.d
45 ; CHECK-NEXT: mov z0.d, z5.d
48 %a.deinterleaved = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.vector.deinterleave2.nxv8f32(<vscale x 8 x float> %a)
49 %a.real = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } %a.deinterleaved, 0
50 %a.imag = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } %a.deinterleaved, 1
51 %b.deinterleaved = tail call { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.vector.deinterleave2.nxv8f32(<vscale x 8 x float> %b)
52 %b.real = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } %b.deinterleaved, 0
53 %b.imag = extractvalue { <vscale x 4 x float>, <vscale x 4 x float> } %b.deinterleaved, 1
54 %0 = fmul fast <vscale x 4 x float> %b.imag, %a.real
55 %1 = fmul fast <vscale x 4 x float> %b.real, %a.imag
56 %2 = fadd fast <vscale x 4 x float> %1, %0
57 %3 = fmul fast <vscale x 4 x float> %b.real, %a.real
58 %4 = fmul fast <vscale x 4 x float> %a.imag, %b.imag
59 %5 = fsub fast <vscale x 4 x float> %3, %4
60 %interleaved.vec = tail call <vscale x 8 x float> @llvm.vector.interleave2.nxv8f32(<vscale x 4 x float> %5, <vscale x 4 x float> %2)
61 ret <vscale x 8 x float> %interleaved.vec
64 ; Expected to transform
65 define <vscale x 16 x float> @complex_mul_v16f32(<vscale x 16 x float> %a, <vscale x 16 x float> %b) {
66 ; CHECK-LABEL: complex_mul_v16f32:
67 ; CHECK: // %bb.0: // %entry
68 ; CHECK-NEXT: mov z24.s, #0 // =0x0
69 ; CHECK-NEXT: ptrue p0.s
70 ; CHECK-NEXT: mov z25.d, z24.d
71 ; CHECK-NEXT: mov z26.d, z24.d
72 ; CHECK-NEXT: mov z27.d, z24.d
73 ; CHECK-NEXT: fcmla z24.s, p0/m, z7.s, z3.s, #0
74 ; CHECK-NEXT: fcmla z25.s, p0/m, z4.s, z0.s, #0
75 ; CHECK-NEXT: fcmla z26.s, p0/m, z5.s, z1.s, #0
76 ; CHECK-NEXT: fcmla z27.s, p0/m, z6.s, z2.s, #0
77 ; CHECK-NEXT: fcmla z24.s, p0/m, z7.s, z3.s, #90
78 ; CHECK-NEXT: fcmla z25.s, p0/m, z4.s, z0.s, #90
79 ; CHECK-NEXT: fcmla z26.s, p0/m, z5.s, z1.s, #90
80 ; CHECK-NEXT: fcmla z27.s, p0/m, z6.s, z2.s, #90
81 ; CHECK-NEXT: mov z3.d, z24.d
82 ; CHECK-NEXT: mov z0.d, z25.d
83 ; CHECK-NEXT: mov z1.d, z26.d
84 ; CHECK-NEXT: mov z2.d, z27.d
87 %a.deinterleaved = tail call { <vscale x 8 x float>, <vscale x 8 x float> } @llvm.vector.deinterleave2.nxv16f32(<vscale x 16 x float> %a)
88 %a.real = extractvalue { <vscale x 8 x float>, <vscale x 8 x float> } %a.deinterleaved, 0
89 %a.imag = extractvalue { <vscale x 8 x float>, <vscale x 8 x float> } %a.deinterleaved, 1
90 %b.deinterleaved = tail call { <vscale x 8 x float>, <vscale x 8 x float> } @llvm.vector.deinterleave2.nxv16f32(<vscale x 16 x float> %b)
91 %b.real = extractvalue { <vscale x 8 x float>, <vscale x 8 x float> } %b.deinterleaved, 0
92 %b.imag = extractvalue { <vscale x 8 x float>, <vscale x 8 x float> } %b.deinterleaved, 1
93 %0 = fmul fast <vscale x 8 x float> %b.imag, %a.real
94 %1 = fmul fast <vscale x 8 x float> %b.real, %a.imag
95 %2 = fadd fast <vscale x 8 x float> %1, %0
96 %3 = fmul fast <vscale x 8 x float> %b.real, %a.real
97 %4 = fmul fast <vscale x 8 x float> %a.imag, %b.imag
98 %5 = fsub fast <vscale x 8 x float> %3, %4
99 %interleaved.vec = tail call <vscale x 16 x float> @llvm.vector.interleave2.nxv16f32(<vscale x 8 x float> %5, <vscale x 8 x float> %2)
100 ret <vscale x 16 x float> %interleaved.vec
103 declare { <vscale x 2 x float>, <vscale x 2 x float> } @llvm.vector.deinterleave2.nxv4f32(<vscale x 4 x float>)
104 declare <vscale x 4 x float> @llvm.vector.interleave2.nxv4f32(<vscale x 2 x float>, <vscale x 2 x float>)
106 declare { <vscale x 4 x float>, <vscale x 4 x float> } @llvm.vector.deinterleave2.nxv8f32(<vscale x 8 x float>)
107 declare <vscale x 8 x float> @llvm.vector.interleave2.nxv8f32(<vscale x 4 x float>, <vscale x 4 x float>)
109 declare { <vscale x 8 x float>, <vscale x 8 x float> } @llvm.vector.deinterleave2.nxv16f32(<vscale x 16 x float>)
110 declare <vscale x 16 x float> @llvm.vector.interleave2.nxv16f32(<vscale x 8 x float>, <vscale x 8 x float>)