| blob | 3e5fa741c243abc4459a53e53e71edf1cb47ebe4 |
1 ; RUN: llc < %s -mcpu=cortex-a57 -mattr=+neon -fp-contract=fast -regalloc=pbqp -pbqp-coalescing | FileCheck %s --check-prefix CHECK --check-prefix CHECK-EVEN
2 ; RUN: llc < %s -mcpu=cortex-a57 -mattr=+neon -fp-contract=fast -regalloc=pbqp -pbqp-coalescing | FileCheck %s --check-prefix CHECK --check-prefix CHECK-ODD
3 ;
4 ; Test PBQP is able to fulfill the accumulator chaining constraint.
5 target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
6 target triple = "aarch64"
8 ; CHECK-LABEL: fir
9 ; CHECK-EVEN: fmadd {{d[0-9]*[02468]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[02468]}}
10 ; CHECK-EVEN: fmadd {{d[0-9]*[02468]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[02468]}}
11 ; CHECK-EVEN: fmadd {{d[0-9]*[02468]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[02468]}}
12 ; CHECK-EVEN: fmadd {{d[0-9]*[02468]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[02468]}}
13 ; CHECK-EVEN: fmadd {{d[0-9]*[02468]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[02468]}}
14 ; CHECK-EVEN: fmadd {{d[0-9]*[02468]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[02468]}}
15 ; CHECK-EVEN: fmadd {{d[0-9]*[02468]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[02468]}}
16 ; CHECK-ODD: fmadd {{d[0-9]*[13579]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[13579]}}
17 ; CHECK-ODD: fmadd {{d[0-9]*[13579]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[13579]}}
18 ; CHECK-ODD: fmadd {{d[0-9]*[13579]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[13579]}}
19 ; CHECK-ODD: fmadd {{d[0-9]*[13579]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[13579]}}
20 ; CHECK-ODD: fmadd {{d[0-9]*[13579]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[13579]}}
21 ; CHECK-ODD: fmadd {{d[0-9]*[13579]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[13579]}}
22 ; CHECK-ODD: fmadd {{d[0-9]*[13579]}}, {{d[0-9]*}}, {{d[0-9]*}}, {{d[0-9]*[13579]}}
23 define void @fir(double* nocapture %rx, double* nocapture %ry, double* nocapture %c, double* nocapture %x, double* nocapture %y) {
24 entry:
25 %0 = load double, double* %c, align 8
26 %1 = load double, double* %x, align 8
27 %mul = fmul fast double %1, %0
28 %2 = load double, double* %y, align 8
29 %mul7 = fmul fast double %2, %0
30 %arrayidx.1 = getelementptr inbounds double, double* %c, i64 1
31 %3 = load double, double* %arrayidx.1, align 8
32 %arrayidx2.1 = getelementptr inbounds double, double* %x, i64 1
33 %4 = load double, double* %arrayidx2.1, align 8
34 %mul.1 = fmul fast double %4, %3
35 %add.1 = fadd fast double %mul.1, %mul
36 %arrayidx6.1 = getelementptr inbounds double, double* %y, i64 1
37 %5 = load double, double* %arrayidx6.1, align 8
38 %mul7.1 = fmul fast double %5, %3
39 %add8.1 = fadd fast double %mul7.1, %mul7
40 %arrayidx.2 = getelementptr inbounds double, double* %c, i64 2
41 %6 = load double, double* %arrayidx.2, align 8
42 %arrayidx2.2 = getelementptr inbounds double, double* %x, i64 2
43 %7 = load double, double* %arrayidx2.2, align 8
44 %mul.2 = fmul fast double %7, %6
45 %add.2 = fadd fast double %mul.2, %add.1
46 %arrayidx6.2 = getelementptr inbounds double, double* %y, i64 2
47 %8 = load double, double* %arrayidx6.2, align 8
48 %mul7.2 = fmul fast double %8, %6
49 %add8.2 = fadd fast double %mul7.2, %add8.1
50 %arrayidx.3 = getelementptr inbounds double, double* %c, i64 3
51 %9 = load double, double* %arrayidx.3, align 8
52 %arrayidx2.3 = getelementptr inbounds double, double* %x, i64 3
53 %10 = load double, double* %arrayidx2.3, align 8
54 %mul.3 = fmul fast double %10, %9
55 %add.3 = fadd fast double %mul.3, %add.2
56 %arrayidx6.3 = getelementptr inbounds double, double* %y, i64 3
57 %11 = load double, double* %arrayidx6.3, align 8
58 %mul7.3 = fmul fast double %11, %9
59 %add8.3 = fadd fast double %mul7.3, %add8.2
60 %arrayidx.4 = getelementptr inbounds double, double* %c, i64 4
61 %12 = load double, double* %arrayidx.4, align 8
62 %arrayidx2.4 = getelementptr inbounds double, double* %x, i64 4
63 %13 = load double, double* %arrayidx2.4, align 8
64 %mul.4 = fmul fast double %13, %12
65 %add.4 = fadd fast double %mul.4, %add.3
66 %arrayidx6.4 = getelementptr inbounds double, double* %y, i64 4
67 %14 = load double, double* %arrayidx6.4, align 8
68 %mul7.4 = fmul fast double %14, %12
69 %add8.4 = fadd fast double %mul7.4, %add8.3
70 %arrayidx.5 = getelementptr inbounds double, double* %c, i64 5
71 %15 = load double, double* %arrayidx.5, align 8
72 %arrayidx2.5 = getelementptr inbounds double, double* %x, i64 5
73 %16 = load double, double* %arrayidx2.5, align 8
74 %mul.5 = fmul fast double %16, %15
75 %add.5 = fadd fast double %mul.5, %add.4
76 %arrayidx6.5 = getelementptr inbounds double, double* %y, i64 5
77 %17 = load double, double* %arrayidx6.5, align 8
78 %mul7.5 = fmul fast double %17, %15
79 %add8.5 = fadd fast double %mul7.5, %add8.4
80 %arrayidx.6 = getelementptr inbounds double, double* %c, i64 6
81 %18 = load double, double* %arrayidx.6, align 8
82 %arrayidx2.6 = getelementptr inbounds double, double* %x, i64 6
83 %19 = load double, double* %arrayidx2.6, align 8
84 %mul.6 = fmul fast double %19, %18
85 %add.6 = fadd fast double %mul.6, %add.5
86 %arrayidx6.6 = getelementptr inbounds double, double* %y, i64 6
87 %20 = load double, double* %arrayidx6.6, align 8
88 %mul7.6 = fmul fast double %20, %18
89 %add8.6 = fadd fast double %mul7.6, %add8.5
90 %arrayidx.7 = getelementptr inbounds double, double* %c, i64 7
91 %21 = load double, double* %arrayidx.7, align 8
92 %arrayidx2.7 = getelementptr inbounds double, double* %x, i64 7
93 %22 = load double, double* %arrayidx2.7, align 8
94 %mul.7 = fmul fast double %22, %21
95 %add.7 = fadd fast double %mul.7, %add.6
96 %arrayidx6.7 = getelementptr inbounds double, double* %y, i64 7
97 %23 = load double, double* %arrayidx6.7, align 8
98 %mul7.7 = fmul fast double %23, %21
99 %add8.7 = fadd fast double %mul7.7, %add8.6
100 store double %add.7, double* %rx, align 8
101 store double %add8.7, double* %ry, align 8
102 ret void
103 }