symjit 2.20.4

a lightweight just-in-time (JIT) optimizer compiler
Documentation 
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
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320

#[macro_use]
mod macros;

use crate::assembler::Assembler;
use crate::code::Func;
use crate::utils::Reg;

const SP: u8 = 31;

const MEM: u8 = 19; // first arg = mem if direct mode, otherwise null
const PARAMS: u8 = 20; // fourth arg = params
const STATES: u8 = 21; // second arg = states+obs if indirect mode, otherwise null
const IDX: u8 = 22; // third arg = index if indirect mode
const CALL: u8 = 23; // call pointer
const STACK: u8 = 24;

const SCRATCH1: u8 = 9;
const SCRATCH2: u8 = 10;
const SCRATCH3: u8 = 11;
const COUNTER: u8 = 12;
const TEMP: u8 = ϕ(Reg::Temp);

/*
 * registers v8 to v16 are ABI-preserved
 * registers v29-v31 can be temporary
 */
const FMAP: [u8; 30] = [
    2, 3, 4, 5, 6, 7, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 8, 9, 10, 11, 12, 13, 14,
    15, 29, 30, 31,
];

const T0: u8 = 29;
const T1: u8 = 30;
const T2: u8 = 31;

mod complex;
mod scalar;
mod vector;

pub use complex::ArmComplexGenerator;
pub use scalar::ArmGenerator;
pub use vector::ArmSimdGenerator;

const fn ϕ(r: Reg) -> u8 {
    match r {
        Reg::Ret => 0,  // d0
        Reg::Temp => 1, // d1
        Reg::Left => 0,
        Reg::Right => 1,
        Reg::Gen(dst) => FMAP[dst as usize],
        Reg::Static(..) => panic!("passing static registers to codegen"),
    }
}

fn emit(a: &mut Assembler, w: u32) {
    a.append_word(w);
}

fn load_long(a: &mut Assembler, reg: u8, label: &str) {
    let data = (a.ip() & 0xfffff000) as u32 | reg as u32;

    a.jump_abs(label, data, |offset, data| {
        let pg = (data & 0xfffff000) as i32;
        let reg = (data & 0xff) as u8;
        arm! {adrp x(reg), label((offset - pg) as u32)}
    });

    a.jump_abs(
        label,
        reg as u32,
        |offset, reg| arm! {ldr x(reg), [x(reg), #offset & 0x0fff]},
    );
}

fn load_d_from_mem(a: &mut Assembler, d: u8, base: u8, idx: u32) {
    if idx < 4096 {
        emit(a, arm! {ldr d(d), [x(base), #8*idx]});
    } else if idx < 65536 {
        emit(a, arm! {movz x(SCRATCH1), #idx});
        emit(a, arm! {ldr d(d), [x(base), x(SCRATCH1), lsl #3]});
    } else {
        emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
        emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
        emit(a, arm! {ldr d(d), [x(base), x(SCRATCH1), lsl #3]});
    }
}

fn save_d_to_mem(a: &mut Assembler, d: u8, base: u8, idx: u32) {
    if idx < 4096 {
        emit(a, arm! {str d(d), [x(base), #8*idx]});
    } else if idx < 65536 {
        emit(a, arm! {movz x(SCRATCH1), #idx});
        emit(a, arm! {str d(d), [x(base), x(SCRATCH1), lsl #3]});
    } else {
        emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
        emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
        emit(a, arm! {str d(d), [x(base), x(SCRATCH1), lsl #3]});
    }
}

fn load_paired_d_from_mem(a: &mut Assembler, d1: u8, d2: u8, base: u8, idx: u32) {
    assert!(idx & 1 == 0);

    match idx {
        0..64 => emit(a, arm! {ldp d(d1), d(d2), [x(base), #8*idx]}),
        64..256 => {
            emit(a, arm! {add x(SCRATCH1), x(base), #8*idx});
            emit(a, arm! {ldp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
        256..4096 => {
            emit(a, arm! {ldr d(d1), [x(base), #8*idx]});
            emit(a, arm! {ldr d(d2), [x(base), #8*(idx+1)]});
        }
        4096..65536 => {
            emit(a, arm! {movz x(SCRATCH1), #idx});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #3});
            emit(a, arm! {ldp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
        65536..131072 => {
            emit(a, arm! {movz x(SCRATCH1), #idx >> 1});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #4});
            emit(a, arm! {ldp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
        idx => {
            emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
            emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #3});
            emit(a, arm! {ldp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
    }
}

fn save_paired_d_to_mem(a: &mut Assembler, d1: u8, d2: u8, base: u8, idx: u32) {
    assert!(idx & 1 == 0);

    match idx {
        0..64 => emit(a, arm! {stp d(d1), d(d2), [x(base), #8*idx]}),
        64..256 => {
            emit(a, arm! {add x(SCRATCH1), x(base), #8*idx});
            emit(a, arm! {stp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
        256..4096 => {
            emit(a, arm! {str d(d1), [x(base), #8*idx]});
            emit(a, arm! {str d(d2), [x(base), #8*(idx+1)]});
        }
        4096..65536 => {
            emit(a, arm! {movz x(SCRATCH1), #idx});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #3});
            emit(a, arm! {stp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
        65536..131072 => {
            emit(a, arm! {movz x(SCRATCH1), #idx >> 1});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #4});
            emit(a, arm! {stp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
        idx => {
            emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
            emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #3});
            emit(a, arm! {stp d(d1), d(d2), [x(SCRATCH1), #0]});
        }
    }
}

fn load_q_from_mem(a: &mut Assembler, d: u8, base: u8, idx: u32) {
    if idx < 4096 {
        emit(a, arm! {ldr q(d), [x(base), #16*idx]});
    } else if idx < 65536 {
        emit(a, arm! {movz x(SCRATCH1), #idx});
        emit(a, arm! {ldr q(d), [x(base), x(SCRATCH1), lsl #4]});
    } else {
        emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
        emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
        emit(a, arm! {ldr q(d), [x(base), x(SCRATCH1), lsl #4]});
    }
}

fn save_q_to_mem(a: &mut Assembler, d: u8, base: u8, idx: u32) {
    if idx < 4096 {
        emit(a, arm! {str q(d), [x(base), #16*idx]});
    } else if idx < 65536 {
        emit(a, arm! {movz x(SCRATCH1), #idx});
        emit(a, arm! {str q(d), [x(base), x(SCRATCH1), lsl #4]});
    } else {
        emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
        emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
        emit(a, arm! {str q(d), [x(base), x(SCRATCH1), lsl #4]});
    }
}

fn load_paired_q_from_mem(a: &mut Assembler, d1: u8, d2: u8, base: u8, idx: u32) {
    assert!(idx & 1 == 0);

    match idx {
        0..64 => emit(a, arm! {ldp q(d1), q(d2), [x(base), #16*idx]}),
        64..256 => {
            emit(a, arm! {add x(SCRATCH1), x(base), #16*idx});
            emit(a, arm! {ldp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
        256..4096 => {
            emit(a, arm! {ldr q(d1), [x(base), #16*idx]});
            emit(a, arm! {ldr q(d2), [x(base), #16*(idx+1)]});
        }
        4096..65536 => {
            emit(a, arm! {movz x(SCRATCH1), #idx});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #4});
            emit(a, arm! {ldp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
        65536..131072 => {
            emit(a, arm! {movz x(SCRATCH1), #idx >> 1});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #5});
            emit(a, arm! {ldp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
        idx => {
            emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
            emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #4});
            emit(a, arm! {ldp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
    }
}

fn save_paired_q_to_mem(a: &mut Assembler, d1: u8, d2: u8, base: u8, idx: u32) {
    assert!(idx & 1 == 0);

    match idx {
        0..64 => emit(a, arm! {stp q(d1), q(d2), [x(base), #16*idx]}),
        64..256 => {
            emit(a, arm! {add x(SCRATCH1), x(base), #16*idx});
            emit(a, arm! {stp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
        256..4096 => {
            emit(a, arm! {str q(d1), [x(base), #16*idx]});
            emit(a, arm! {str q(d2), [x(base), #16*(idx+1)]});
        }
        4096..65536 => {
            emit(a, arm! {movz x(SCRATCH1), #idx});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #4});
            emit(a, arm! {stp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
        65536..131072 => {
            emit(a, arm! {movz x(SCRATCH1), #idx >> 1});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #5});
            emit(a, arm! {stp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
        idx => {
            emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
            emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
            emit(a, arm! {add x(SCRATCH1), x(base), x(SCRATCH1), lsl #4});
            emit(a, arm! {stp q(d1), q(d2), [x(SCRATCH1), #0]});
        }
    }
}

fn load_x_from_mem(a: &mut Assembler, r: u8, base: u8, idx: u32) {
    assert!(r != 9);

    if idx < 4096 {
        emit(a, arm! {ldr x(r), [x(base), #8*idx]});
    } else if idx < 65536 {
        emit(a, arm! {movz x(SCRATCH1), #idx});
        emit(a, arm! {ldr x(r), [x(base), x(SCRATCH1), lsl #3]});
    } else {
        emit(a, arm! {movz x(SCRATCH1), #idx & 0xffff});
        emit(a, arm! {movk_lsl16 x(SCRATCH1), #idx >> 16});
        emit(a, arm! {ldr x(r), [x(base), x(SCRATCH1), lsl #3]});
    }
}

fn load_x_from_label(a: &mut Assembler, dst: u8, label: &str) {
    load_long(a, dst, label);
}

fn add_consts(a: &mut Assembler, consts: &[f64]) {
    for (idx, val) in consts.iter().enumerate() {
        let label = format!("_const_{}_", idx);
        a.set_label(label.as_str());
        a.append_quad((*val).to_bits());
    }
}

fn add_func(a: &mut Assembler, op: &str, f: Func) {
    if let Func::Slice {
        f_scalar,
        f_simd,
        env,
        ..
    } = f
    {
        let label = format!("_func_{}_", op);
        a.set_label(label.as_str());
        a.append_quad(f_scalar as u64);

        let label = format!("_simd_{}_", op);
        a.set_label(label.as_str());
        a.append_quad(f_simd as u64);

        let label = format!("_env_{}_", op);
        a.set_label(label.as_str());
        a.append_quad(env as u64);
    } else {
        let label = format!("_func_{}_", op);
        a.set_label(label.as_str());
        a.append_quad(f.func_ptr());
    }
}

fn sub_stack(a: &mut Assembler, size: u32) {
    emit(a, arm! {sub sp, sp, #size & 0x0fff});
    if size >> 12 != 0 {
        emit(a, arm! {sub sp, sp, #size >> 12, lsl #12});
    }
}

fn add_stack(a: &mut Assembler, size: u32) {
    if size >> 12 != 0 {
        emit(a, arm! {add sp, sp, #size >> 12, lsl #12});
    }
    emit(a, arm! {add sp, sp, #size & 0x0fff});
}