llvm-native-core 0.1.10

LLVM-native core semantic engine — IR, CodeGen, X86 MC, Clang frontend pipeline
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

//! Hexagon MC Encoder — instruction encoding for Qualcomm Hexagon
//! VLIW DSP.
//!
//! Encodes Hexagon instructions into little-endian byte sequences.
//! All instructions are 32 bits wide and organized in 4-slot VLIW
//! packets (up to 4 instructions per packet).
//!
//! # Instruction Encoding (32-bit)
//!
//! ```text
//! Major opcode(4) | Sub-opcode(4) | Reg fields | Immediate | Parse bits
//! ```
//!
//! Slot types: 0=ALU/MPY, 1=LD/ST, 2=J/JR, 3=XTYPE (vector)

use super::hexagon_instr_info::HexagonOpcode;
use crate::codegen::{MachineFunction, MachineInstr, MachineOperand};

fn u32_to_le_bytes(word: u32) -> Vec<u8> {
    word.to_le_bytes().to_vec()
}

pub struct HexagonMCEncoder {
    pub is_64bit: bool,
    pub output: Vec<u8>,
}

impl HexagonMCEncoder {
    pub fn new(_is_64bit: bool) -> Self {
        Self {
            is_64bit: false,
            output: Vec::with_capacity(256),
        }
    }

    pub fn encode_instruction(&mut self, mi: &MachineInstr) -> Vec<u8> {
        let bytes = self.instruction_to_bytes(mi);
        self.output.extend_from_slice(&bytes);
        bytes
    }

    pub fn encode_function(&mut self, mf: &MachineFunction) -> Vec<u8> {
        for block in &mf.blocks {
            for instr in &block.instructions {
                self.encode_instruction(instr);
            }
        }
        std::mem::take(&mut self.output)
    }

    fn instruction_to_bytes(&self, mi: &MachineInstr) -> Vec<u8> {
        let word = self.encode_by_opcode(mi);
        u32_to_le_bytes(word)
    }

    fn encode_by_opcode(&self, mi: &MachineInstr) -> u32 {
        match get_hexagon_opcode(mi.opcode) {
            HexagonOpcode::ADD => self.encode_alu(0x00, mi),
            HexagonOpcode::SUB => self.encode_alu(0x01, mi),
            HexagonOpcode::AND => self.encode_alu(0x02, mi),
            HexagonOpcode::OR => self.encode_alu(0x03, mi),
            HexagonOpcode::XOR => self.encode_alu(0x04, mi),
            HexagonOpcode::NEG => self.encode_alu_2r(0x05, mi),
            HexagonOpcode::NOT => self.encode_alu_2r(0x06, mi),
            HexagonOpcode::ABS => self.encode_alu_2r(0x07, mi),
            HexagonOpcode::LSL => self.encode_alu(0x10, mi),
            HexagonOpcode::LSR => self.encode_alu(0x11, mi),
            HexagonOpcode::ASR => self.encode_alu(0x12, mi),
            HexagonOpcode::MPY => self.encode_alu(0x20, mi),
            HexagonOpcode::MPYU => self.encode_alu(0x21, mi),
            HexagonOpcode::MPYI => self.encode_alu(0x22, mi),
            HexagonOpcode::LDW => self.encode_mem(0x30, mi),
            HexagonOpcode::LDB => self.encode_mem(0x31, mi),
            HexagonOpcode::LDH => self.encode_mem(0x32, mi),
            HexagonOpcode::LDD => self.encode_mem(0x33, mi),
            HexagonOpcode::STW => self.encode_mem(0x34, mi),
            HexagonOpcode::STB => self.encode_mem(0x35, mi),
            HexagonOpcode::STH => self.encode_mem(0x36, mi),
            HexagonOpcode::STD => self.encode_mem(0x37, mi),
            HexagonOpcode::CMPEQ => self.encode_cmp(0x40, mi),
            HexagonOpcode::CMPGT => self.encode_cmp(0x41, mi),
            HexagonOpcode::CMPGTU => self.encode_cmp(0x42, mi),
            HexagonOpcode::JUMP => self.encode_jump(0x50, mi),
            HexagonOpcode::JUMPR => self.encode_jumpr(0x51, mi),
            HexagonOpcode::JUMPr => self.encode_jumpr(0x52, mi),
            HexagonOpcode::CALL => self.encode_call(0x53, mi),
            HexagonOpcode::CALLR => self.encode_jumpr(0x54, mi),
            HexagonOpcode::MOV => self.encode_alu_2r(0x0F, mi),
            HexagonOpcode::NOP => 0x7C00,
            HexagonOpcode::LI => self.encode_li(mi),
            HexagonOpcode::VADD => self.encode_vec(0x60, mi),
            HexagonOpcode::VSUB => self.encode_vec(0x61, mi),
            HexagonOpcode::VMPY => self.encode_vec(0x62, mi),
            _ => 0x7C00, // NOP
        }
    }

    fn encode_alu(&self, sub_op: u32, mi: &MachineInstr) -> u32 {
        let rd = Self::get_reg(mi, 0);
        let rs = Self::get_reg(mi, 1);
        let rt = Self::get_reg(mi, 2);
        (sub_op << 28) | (rd << 21) | (rs << 16) | (rt << 0)
    }

    fn encode_alu_2r(&self, sub_op: u32, mi: &MachineInstr) -> u32 {
        let rd = Self::get_reg(mi, 0);
        let rs = Self::get_reg(mi, 1);
        (sub_op << 28) | (rd << 21) | (rs << 16)
    }

    fn encode_mem(&self, sub_op: u32, mi: &MachineInstr) -> u32 {
        let rd = Self::get_reg(mi, 0);
        let rs = Self::get_reg(mi, 1);
        let offset: u32 = Self::get_imm(mi, 2) as u32 & 0xFFFF;
        (sub_op << 28) | (rd << 21) | (rs << 16) | offset
    }

    fn encode_cmp(&self, sub_op: u32, mi: &MachineInstr) -> u32 {
        let pd = Self::get_reg(mi, 0);
        let rs = Self::get_reg(mi, 1);
        let rt = Self::get_reg(mi, 2);
        (sub_op << 28) | (pd << 23) | (rs << 16) | (rt << 0)
    }

    fn encode_jump(&self, sub_op: u32, _mi: &MachineInstr) -> u32 {
        (sub_op << 28)
    }

    fn encode_jumpr(&self, sub_op: u32, mi: &MachineInstr) -> u32 {
        let rs = Self::get_reg(mi, 0);
        (sub_op << 28) | (rs << 16)
    }

    fn encode_call(&self, sub_op: u32, _mi: &MachineInstr) -> u32 {
        (sub_op << 28)
    }

    fn encode_li(&self, mi: &MachineInstr) -> u32 {
        let rd = Self::get_reg(mi, 0);
        let imm: u32 = Self::get_imm(mi, 1) as u32 & 0xFFFF;
        (0x0E << 28) | (rd << 21) | imm
    }

    fn encode_vec(&self, sub_op: u32, mi: &MachineInstr) -> u32 {
        let vd = Self::get_reg(mi, 0);
        let vs = Self::get_reg(mi, 1);
        let vt = Self::get_reg(mi, 2);
        (sub_op << 28) | (vd << 21) | (vs << 16) | (vt << 0)
    }

    fn get_reg(mi: &MachineInstr, index: usize) -> u32 {
        mi.operands
            .get(index)
            .map(|op| match op {
                MachineOperand::Reg(vr) => *vr & 0x1F,
                MachineOperand::PhysReg(pr) => *pr & 0x1F,
                _ => 0,
            })
            .unwrap_or(0) as u32
    }

    fn get_imm(mi: &MachineInstr, index: usize) -> i64 {
        mi.operands
            .get(index)
            .map(|op| match op {
                MachineOperand::Imm(imm) => *imm,
                _ => 0,
            })
            .unwrap_or(0)
    }
}

fn get_hexagon_opcode(opcode: u32) -> HexagonOpcode {
    match opcode {
        0 => HexagonOpcode::ADD,
        1 => HexagonOpcode::SUB,
        2 => HexagonOpcode::AND,
        3 => HexagonOpcode::OR,
        4 => HexagonOpcode::XOR,
        5 => HexagonOpcode::NEG,
        6 => HexagonOpcode::NOT,
        20 => HexagonOpcode::LSL,
        21 => HexagonOpcode::LSR,
        22 => HexagonOpcode::ASR,
        30 => HexagonOpcode::MPY,
        31 => HexagonOpcode::MPYU,
        32 => HexagonOpcode::MPYI,
        50 => HexagonOpcode::LDW,
        51 => HexagonOpcode::LDB,
        52 => HexagonOpcode::LDH,
        53 => HexagonOpcode::LDD,
        60 => HexagonOpcode::STW,
        61 => HexagonOpcode::STB,
        62 => HexagonOpcode::STH,
        63 => HexagonOpcode::STD,
        80 => HexagonOpcode::JUMP,
        81 => HexagonOpcode::JUMPR,
        82 => HexagonOpcode::JUMPr,
        83 => HexagonOpcode::CALL,
        84 => HexagonOpcode::CALLR,
        100 => HexagonOpcode::CMPEQ,
        101 => HexagonOpcode::CMPGT,
        102 => HexagonOpcode::CMPGTU,
        103 => HexagonOpcode::CMPLT,
        104 => HexagonOpcode::CMPLTU,
        120 => HexagonOpcode::VADD,
        121 => HexagonOpcode::VSUB,
        122 => HexagonOpcode::VMPY,
        140 => HexagonOpcode::NOP,
        141 => HexagonOpcode::LI,
        142 => HexagonOpcode::MOV,
        _ => HexagonOpcode::NOP,
    }
}