1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
|
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
; RUN: llc -mtriple=riscv32 -mattr=+zfbfmin -verify-machineinstrs \
; RUN: -target-abi ilp32f < %s | FileCheck -check-prefixes=CHECK,RV32IZFBFMIN %s
; RUN: llc -mtriple=riscv64 -mattr=+zfbfmin -verify-machineinstrs \
; RUN: -target-abi lp64f < %s | FileCheck -check-prefixes=CHECK,RV64IZFBFMIN %s
define bfloat @flh(ptr %a) nounwind {
; CHECK-LABEL: flh:
; CHECK: # %bb.0:
; CHECK-NEXT: flh fa5, 6(a0)
; CHECK-NEXT: flh fa4, 0(a0)
; CHECK-NEXT: fcvt.s.bf16 fa5, fa5
; CHECK-NEXT: fcvt.s.bf16 fa4, fa4
; CHECK-NEXT: fadd.s fa5, fa4, fa5
; CHECK-NEXT: fcvt.bf16.s fa0, fa5
; CHECK-NEXT: ret
%1 = load bfloat, ptr %a
%2 = getelementptr bfloat, ptr %a, i32 3
%3 = load bfloat, ptr %2
; Use both loaded values in an FP op to ensure an flh is used, even for the
; soft bfloat ABI
%4 = fadd bfloat %1, %3
ret bfloat %4
}
define dso_local void @fsh(ptr %a, bfloat %b, bfloat %c) nounwind {
; CHECK-LABEL: fsh:
; CHECK: # %bb.0:
; CHECK-NEXT: fcvt.s.bf16 fa5, fa1
; CHECK-NEXT: fcvt.s.bf16 fa4, fa0
; CHECK-NEXT: fadd.s fa5, fa4, fa5
; CHECK-NEXT: fcvt.bf16.s fa5, fa5
; CHECK-NEXT: fsh fa5, 0(a0)
; CHECK-NEXT: fsh fa5, 16(a0)
; CHECK-NEXT: ret
%1 = fadd bfloat %b, %c
store bfloat %1, ptr %a
%2 = getelementptr bfloat, ptr %a, i32 8
store bfloat %1, ptr %2
ret void
}
; Check load and store to a global
@G = dso_local global bfloat 0.0
define bfloat @flh_fsh_global(bfloat %a, bfloat %b) nounwind {
; Use %a and %b in an FP op to ensure bfloat precision floating point registers
; are used, even for the soft bfloat ABI
; CHECK-LABEL: flh_fsh_global:
; CHECK: # %bb.0:
; CHECK-NEXT: fcvt.s.bf16 fa5, fa1
; CHECK-NEXT: fcvt.s.bf16 fa4, fa0
; CHECK-NEXT: lui a0, %hi(G)
; CHECK-NEXT: fadd.s fa5, fa4, fa5
; CHECK-NEXT: flh fa4, %lo(G)(a0)
; CHECK-NEXT: fcvt.bf16.s fa0, fa5
; CHECK-NEXT: addi a1, a0, %lo(G)
; CHECK-NEXT: fsh fa0, %lo(G)(a0)
; CHECK-NEXT: flh fa5, 18(a1)
; CHECK-NEXT: fsh fa0, 18(a1)
; CHECK-NEXT: ret
%1 = fadd bfloat %a, %b
%2 = load volatile bfloat, ptr @G
store bfloat %1, ptr @G
%3 = getelementptr bfloat, ptr @G, i32 9
%4 = load volatile bfloat, ptr %3
store bfloat %1, ptr %3
ret bfloat %1
}
; Ensure that 1 is added to the high 20 bits if bit 11 of the low part is 1
define bfloat @flh_fsh_constant(bfloat %a) nounwind {
; RV32IZFBFMIN-LABEL: flh_fsh_constant:
; RV32IZFBFMIN: # %bb.0:
; RV32IZFBFMIN-NEXT: lui a0, 912092
; RV32IZFBFMIN-NEXT: flh fa5, -273(a0)
; RV32IZFBFMIN-NEXT: fcvt.s.bf16 fa4, fa0
; RV32IZFBFMIN-NEXT: fcvt.s.bf16 fa5, fa5
; RV32IZFBFMIN-NEXT: fadd.s fa5, fa4, fa5
; RV32IZFBFMIN-NEXT: fcvt.bf16.s fa0, fa5
; RV32IZFBFMIN-NEXT: fsh fa0, -273(a0)
; RV32IZFBFMIN-NEXT: ret
;
; RV64IZFBFMIN-LABEL: flh_fsh_constant:
; RV64IZFBFMIN: # %bb.0:
; RV64IZFBFMIN-NEXT: lui a0, 228023
; RV64IZFBFMIN-NEXT: slli a0, a0, 2
; RV64IZFBFMIN-NEXT: flh fa5, -273(a0)
; RV64IZFBFMIN-NEXT: fcvt.s.bf16 fa4, fa0
; RV64IZFBFMIN-NEXT: fcvt.s.bf16 fa5, fa5
; RV64IZFBFMIN-NEXT: fadd.s fa5, fa4, fa5
; RV64IZFBFMIN-NEXT: fcvt.bf16.s fa0, fa5
; RV64IZFBFMIN-NEXT: fsh fa0, -273(a0)
; RV64IZFBFMIN-NEXT: ret
%1 = inttoptr i32 3735928559 to ptr
%2 = load volatile bfloat, ptr %1
%3 = fadd bfloat %a, %2
store bfloat %3, ptr %1
ret bfloat %3
}
declare void @notdead(ptr)
define bfloat @flh_stack(bfloat %a) nounwind {
; RV32IZFBFMIN-LABEL: flh_stack:
; RV32IZFBFMIN: # %bb.0:
; RV32IZFBFMIN-NEXT: addi sp, sp, -16
; RV32IZFBFMIN-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IZFBFMIN-NEXT: fsw fs0, 8(sp) # 4-byte Folded Spill
; RV32IZFBFMIN-NEXT: fmv.s fs0, fa0
; RV32IZFBFMIN-NEXT: addi a0, sp, 4
; RV32IZFBFMIN-NEXT: call notdead
; RV32IZFBFMIN-NEXT: flh fa5, 4(sp)
; RV32IZFBFMIN-NEXT: fcvt.s.bf16 fa4, fs0
; RV32IZFBFMIN-NEXT: fcvt.s.bf16 fa5, fa5
; RV32IZFBFMIN-NEXT: fadd.s fa5, fa5, fa4
; RV32IZFBFMIN-NEXT: fcvt.bf16.s fa0, fa5
; RV32IZFBFMIN-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IZFBFMIN-NEXT: flw fs0, 8(sp) # 4-byte Folded Reload
; RV32IZFBFMIN-NEXT: addi sp, sp, 16
; RV32IZFBFMIN-NEXT: ret
;
; RV64IZFBFMIN-LABEL: flh_stack:
; RV64IZFBFMIN: # %bb.0:
; RV64IZFBFMIN-NEXT: addi sp, sp, -16
; RV64IZFBFMIN-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IZFBFMIN-NEXT: fsw fs0, 4(sp) # 4-byte Folded Spill
; RV64IZFBFMIN-NEXT: fmv.s fs0, fa0
; RV64IZFBFMIN-NEXT: mv a0, sp
; RV64IZFBFMIN-NEXT: call notdead
; RV64IZFBFMIN-NEXT: flh fa5, 0(sp)
; RV64IZFBFMIN-NEXT: fcvt.s.bf16 fa4, fs0
; RV64IZFBFMIN-NEXT: fcvt.s.bf16 fa5, fa5
; RV64IZFBFMIN-NEXT: fadd.s fa5, fa5, fa4
; RV64IZFBFMIN-NEXT: fcvt.bf16.s fa0, fa5
; RV64IZFBFMIN-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IZFBFMIN-NEXT: flw fs0, 4(sp) # 4-byte Folded Reload
; RV64IZFBFMIN-NEXT: addi sp, sp, 16
; RV64IZFBFMIN-NEXT: ret
%1 = alloca bfloat, align 4
call void @notdead(ptr %1)
%2 = load bfloat, ptr %1
%3 = fadd bfloat %2, %a ; force load in to FPR16
ret bfloat %3
}
define dso_local void @fsh_stack(bfloat %a, bfloat %b) nounwind {
; RV32IZFBFMIN-LABEL: fsh_stack:
; RV32IZFBFMIN: # %bb.0:
; RV32IZFBFMIN-NEXT: addi sp, sp, -16
; RV32IZFBFMIN-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IZFBFMIN-NEXT: fcvt.s.bf16 fa5, fa1
; RV32IZFBFMIN-NEXT: fcvt.s.bf16 fa4, fa0
; RV32IZFBFMIN-NEXT: fadd.s fa5, fa4, fa5
; RV32IZFBFMIN-NEXT: fcvt.bf16.s fa5, fa5
; RV32IZFBFMIN-NEXT: fsh fa5, 8(sp)
; RV32IZFBFMIN-NEXT: addi a0, sp, 8
; RV32IZFBFMIN-NEXT: call notdead
; RV32IZFBFMIN-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IZFBFMIN-NEXT: addi sp, sp, 16
; RV32IZFBFMIN-NEXT: ret
;
; RV64IZFBFMIN-LABEL: fsh_stack:
; RV64IZFBFMIN: # %bb.0:
; RV64IZFBFMIN-NEXT: addi sp, sp, -16
; RV64IZFBFMIN-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IZFBFMIN-NEXT: fcvt.s.bf16 fa5, fa1
; RV64IZFBFMIN-NEXT: fcvt.s.bf16 fa4, fa0
; RV64IZFBFMIN-NEXT: fadd.s fa5, fa4, fa5
; RV64IZFBFMIN-NEXT: fcvt.bf16.s fa5, fa5
; RV64IZFBFMIN-NEXT: fsh fa5, 4(sp)
; RV64IZFBFMIN-NEXT: addi a0, sp, 4
; RV64IZFBFMIN-NEXT: call notdead
; RV64IZFBFMIN-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IZFBFMIN-NEXT: addi sp, sp, 16
; RV64IZFBFMIN-NEXT: ret
%1 = fadd bfloat %a, %b ; force store from FPR16
%2 = alloca bfloat, align 4
store bfloat %1, ptr %2
call void @notdead(ptr %2)
ret void
}
|
