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
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
//! Zve64x vector load instructions
#[cfg(test)]
mod tests;
use crate::instructions::Instruction;
use crate::instructions::v::Eew;
use crate::registers::general_purpose::Register;
use crate::registers::vector::VReg;
use ab_riscv_macros::instruction;
use core::fmt;
/// RISC-V Zve64x vector load instruction.
///
/// Encoded under the LOAD-FP major opcode (0x07). All loads use rs1 (GPR) as a base address and vd
/// (vector register) as a destination.
#[instruction]
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
#[rustfmt::skip]
#[doc(hidden)]
pub enum Zve64xLoadInstruction<Reg> {
/// Unit-stride load: `vle{eew}.v vd, (rs1), vm`
///
/// mop=00, lumop=00000, nf=000
Vle { vd: VReg, rs1: Reg, vm: bool, eew: Eew },
/// Unit-stride fault-only-first load: `vle{eew}ff.v vd, (rs1), vm`
///
/// mop=00, lumop=10000, nf=000
Vleff { vd: VReg, rs1: Reg, vm: bool, eew: Eew },
/// Unit-stride mask load: `vlm.v vd, (rs1)`
///
/// mop=00, lumop=01011, nf=000, eew=e8, vm=1
Vlm { vd: VReg, rs1: Reg },
/// Strided load: `vlse{eew}.v vd, (rs1), rs2, vm`
///
/// mop=10, nf=000
Vlse { vd: VReg, rs1: Reg, rs2: Reg, vm: bool, eew: Eew },
/// Indexed-unordered load: `vluxei{eew}.v vd, (rs1), vs2, vm`
///
/// mop=01, nf=000. eew is the index element width.
Vluxei { vd: VReg, rs1: Reg, vs2: VReg, vm: bool, eew: Eew },
/// Indexed-ordered load: `vloxei{eew}.v vd, (rs1), vs2, vm`
///
/// mop=11, nf=000. eew is the index element width.
Vloxei { vd: VReg, rs1: Reg, vs2: VReg, vm: bool, eew: Eew },
/// Whole-register load: `vl{nreg}re{eew}.v vd, (rs1)`
///
/// mop=00, lumop=01000, vm=1. nreg must be 1, 2, 4, or 8.
Vlr { vd: VReg, rs1: Reg, nreg: u8, eew: Eew },
/// Unit-stride segment load: `vlseg{nf}e{eew}.v vd, (rs1), vm`
///
/// mop=00, lumop=00000, nf>0
Vlseg { vd: VReg, rs1: Reg, vm: bool, eew: Eew, nf: u8 },
/// Unit-stride fault-only-first segment load: `vlseg{nf}e{eew}ff.v vd, (rs1), vm`
///
/// mop=00, lumop=10000, nf>0
Vlsegff { vd: VReg, rs1: Reg, vm: bool, eew: Eew, nf: u8 },
/// Strided segment load: `vlsseg{nf}e{eew}.v vd, (rs1), rs2, vm`
///
/// mop=10, nf>0
Vlsseg { vd: VReg, rs1: Reg, rs2: Reg, vm: bool, eew: Eew, nf: u8 },
/// Indexed-unordered segment load: `vluxseg{nf}ei{eew}.v vd, (rs1), vs2, vm`
///
/// mop=01, nf>0
Vluxseg { vd: VReg, rs1: Reg, vs2: VReg, vm: bool, eew: Eew, nf: u8 },
/// Indexed-ordered segment load: `vloxseg{nf}ei{eew}.v vd, (rs1), vs2, vm`
///
/// mop=11, nf>0
Vloxseg { vd: VReg, rs1: Reg, vs2: VReg, vm: bool, eew: Eew, nf: u8 },
}
#[instruction]
impl<Reg> const Instruction for Zve64xLoadInstruction<Reg>
where
Reg: [const] Register,
{
type Reg = Reg;
#[inline(always)]
fn try_decode(instruction: u32) -> Option<Self> {
let opcode = (instruction & 0b111_1111) as u8;
// LOAD-FP major opcode
if opcode != 0b0000111 {
None?;
}
let vd_bits = ((instruction >> 7) & 0x1f) as u8;
let width = ((instruction >> 12) & 0b111) as u8;
let rs1_bits = ((instruction >> 15) & 0x1f) as u8;
let rs2_bits = ((instruction >> 20) & 0x1f) as u8;
let vm = ((instruction >> 25) & 1) != 0;
let mop = ((instruction >> 26) & 0b11) as u8;
let mew = ((instruction >> 28) & 1) as u8;
let nf = ((instruction >> 29) & 0b111) as u8;
// mew must be 0 (reserved for >=128-bit)
if mew != 0 {
None?;
}
let vd = VReg::from_bits(vd_bits)?;
let rs1 = Reg::from_bits(rs1_bits)?;
// nf encodes number of fields minus 1 (nf=0 means 1 field)
let nf_val = nf + 1;
match mop {
// Unit-stride
0b00 => {
let lumop = rs2_bits;
match lumop {
// Regular unit-stride load
0b00000 => {
let eew = Eew::from_width(width)?;
if nf == 0 {
Some(Self::Vle { vd, rs1, vm, eew })
} else {
Some(Self::Vlseg {
vd,
rs1,
vm,
eew,
nf: nf_val,
})
}
}
// Whole-register load
0b01000 => {
// vm must be 1 (unmasked)
if !vm {
None?;
}
let eew = Eew::from_width(width)?;
// nf encodes nreg: nf+1 must be 1, 2, 4, or 8
match nf_val {
1 | 2 | 4 | 8 => Some(Self::Vlr {
vd,
rs1,
nreg: nf_val,
eew,
}),
_ => None,
}
}
// Mask load
0b01011 => {
// Must be eew=e8, vm=1, nf=0
if width != 0b000 || !vm || nf != 0 {
None?;
}
Some(Self::Vlm { vd, rs1 })
}
// Fault-only-first
0b10000 => {
let eew = Eew::from_width(width)?;
if nf == 0 {
Some(Self::Vleff { vd, rs1, vm, eew })
} else {
Some(Self::Vlsegff {
vd,
rs1,
vm,
eew,
nf: nf_val,
})
}
}
_ => None,
}
}
// Indexed-unordered
0b01 => {
let eew = Eew::from_width(width)?;
let vs2 = VReg::from_bits(rs2_bits)?;
if nf == 0 {
Some(Self::Vluxei {
vd,
rs1,
vs2,
vm,
eew,
})
} else {
Some(Self::Vluxseg {
vd,
rs1,
vs2,
vm,
eew,
nf: nf_val,
})
}
}
// Strided
0b10 => {
let eew = Eew::from_width(width)?;
let rs2 = Reg::from_bits(rs2_bits)?;
if nf == 0 {
Some(Self::Vlse {
vd,
rs1,
rs2,
vm,
eew,
})
} else {
Some(Self::Vlsseg {
vd,
rs1,
rs2,
vm,
eew,
nf: nf_val,
})
}
}
// Indexed-ordered
0b11 => {
let eew = Eew::from_width(width)?;
let vs2 = VReg::from_bits(rs2_bits)?;
if nf == 0 {
Some(Self::Vloxei {
vd,
rs1,
vs2,
vm,
eew,
})
} else {
Some(Self::Vloxseg {
vd,
rs1,
vs2,
vm,
eew,
nf: nf_val,
})
}
}
_ => None,
}
}
#[inline(always)]
fn alignment() -> u8 {
size_of::<u32>() as u8
}
#[inline(always)]
fn size(&self) -> u8 {
size_of::<u32>() as u8
}
}
impl<Reg> fmt::Display for Zve64xLoadInstruction<Reg>
where
Reg: fmt::Display,
{
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
#[rustfmt::skip]
match self {
Self::Vle { vd, rs1, vm, eew } => write!(f, "vle{eew}.v {vd}, ({rs1}){}", mask_suffix(vm)),
Self::Vleff { vd, rs1, vm, eew } => write!(f, "vle{eew}ff.v {vd}, ({rs1}){}", mask_suffix(vm)),
Self::Vlm { vd, rs1 } => write!(f, "vlm.v {vd}, ({rs1})"),
Self::Vlse { vd, rs1, rs2, vm, eew } => write!(f, "vlse{eew}.v {vd}, ({rs1}), {rs2}{}", mask_suffix(vm)),
Self::Vluxei { vd, rs1, vs2, vm, eew } => write!(f, "vluxei{eew}.v {vd}, ({rs1}), {vs2}{}", mask_suffix(vm)),
Self::Vloxei { vd, rs1, vs2, vm, eew } => write!(f, "vloxei{eew}.v {vd}, ({rs1}), {vs2}{}", mask_suffix(vm)),
Self::Vlr { vd, rs1, nreg, eew } => write!(f, "vl{}re{}.v {}, ({})", nreg, eew, vd, rs1),
Self::Vlseg { vd, rs1, vm, eew, nf } => write!(f, "vlseg{nf}e{eew}.v {vd}, ({rs1}){}", mask_suffix(vm)),
Self::Vlsegff { vd, rs1, vm, eew, nf } => write!(f, "vlseg{nf}e{eew}ff.v {vd}, ({rs1}){}", mask_suffix(vm)),
Self::Vlsseg { vd, rs1, rs2, vm, eew, nf } => write!(f, "vlsseg{nf}e{eew}.v {vd}, ({rs1}), {rs2}{}", mask_suffix(vm)),
Self::Vluxseg { vd, rs1, vs2, vm, eew, nf } => write!(f, "vluxseg{nf}ei{eew}.v {vd}, ({rs1}), {vs2}{}", mask_suffix(vm)),
Self::Vloxseg { vd, rs1, vs2, vm, eew, nf } => write!(f, "vloxseg{nf}ei{eew}.v {vd}, ({rs1}), {vs2}{}", mask_suffix(vm)),
}
}
}
/// Format mask suffix for display
#[inline(always)]
fn mask_suffix(vm: &bool) -> &'static str {
if *vm { "" } else { ", v0.t" }
}