univdreams 0.4.0

Byte-identical binary <-> source round-tripping: binary formats, multi-arch decode/encode, a .ud decompiler/compiler, and an x86 emulator with Win16/Win32/Linux hosts.
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

//! Pattern: render common SIMD / vector ops as assignment-form.
//!
//! Most of the DLL's hot path is MMX/SSE vector math — `paddw`,
//! `pxor`, `movq`, `psrlw`, etc. — that's left as raw `@asm` lines
//! by every other pattern. The data dependencies are still clear
//! to anyone who reads x86 SIMD, but the visual style clashes
//! with the lifted GP code: everything else reads as
//! `dst = src op X`, the vector ops read as `op dst, src`.
//!
//! This pattern lifts the common shapes to the same assignment
//! form so a function body has a consistent rendering. It's a
//! cosmetic pass — the bytes still encode the original
//! instructions and the operations remain vector / SIMD; the
//! source just doesn't switch styles mid-function.
//!
//! Recognised:
//!
//! * Vector moves — `movq`, `movd`, `movaps`, `movups`, `movdqa`,
//!   `movdqu`, `movss`, `movsd`, `movhps`, `movlps` →
//!   `dst = src`.
//! * Vector arithmetic — `paddb`, `paddw`, `paddd`, `psubb`,
//!   `psubw`, `psubd`, `pmullw`, `pmulhw`, `pmullw`, `paddusb`,
//!   `paddusw`, `psubusb`, `psubusw`, `paddsb`, `paddsw`,
//!   `psubsb`, `psubsw` → `dst = dst + src` / `dst - src` / `dst
//!   * src`.
//! * Vector logical — `pand`, `por`, `pxor`, `pandn` → `dst =
//!   dst & src` / `| src` / `^ src` / `& ~src`.
//! * Vector shift — `psllw`, `pslld`, `psllq`, `psrlw`, `psrld`,
//!   `psrlq`, `psraw`, `psrad` → `dst = dst << src` / `>> src`.
//! * SSE float arith — `addss`, `addsd`, `addps`, `addpd`,
//!   `subss`, `subsd`, `subps`, `subpd`, `mulss`, `mulsd`, etc.
//!
//! Not lifted (would mislead): shuffles (`shufps`, `pshufw`),
//! pack/unpack (`packuswb`, `punpcklbw`), conversions (`cvtsi2sd`,
//! `cvtps2dq`), maskmov, prefetch. Those need their actual
//! semantics spelled out; leaving them as `@asm` is honest.

use crate::arch::x86::{DecodedInsn, Mnemonic, format_intel};
use crate::ast::Stmt;

use super::{Candidate, Pattern, PatternCtx};

pub struct SimdExpr;

impl Pattern for SimdExpr {
    fn name(&self) -> &'static str {
        "simd_expr"
    }

    fn tentative(
        &self,
        _ctx: &PatternCtx,
        insns: &[DecodedInsn],
        start: usize,
    ) -> Option<Candidate> {
        let ins = insns.get(start)?;
        let shape = classify(ins.iced.mnemonic())?;
        let full = format_intel(&ins.iced);
        let (prefix_len, kind) = match shape {
            Shape::Move(prefix) => (prefix.len(), Op::Move),
            Shape::Binary(prefix, op) => (prefix.len(), Op::Binary(op)),
        };
        let rest = full.get(prefix_len..)?;
        let (dst, src) = super::mov::split_two_operands(&full, &full[..prefix_len])?;
        let _ = rest;
        let new_src = match kind {
            Op::Move => src,
            Op::Binary(op) => format!("{dst} {op} {src}"),
        };
        Some(Candidate {
            pattern: self.name(),
            start,
            consumed: 1,
            // Below the structural lifts and the GP `mov` (so a
            // mem_via_reg/lms fold still wins) but above the bare
            // `@asm` fallback.
            priority: 30,
            stmts: vec![Stmt::Move {
                dst,
                src: new_src,
                bytes: ins.original_bytes.clone(),
            }],
        })
    }
}

enum Shape {
    Move(&'static str),
    Binary(&'static str, &'static str),
}

enum Op {
    Move,
    Binary(&'static str),
}

#[allow(clippy::too_many_lines)]
fn classify(m: Mnemonic) -> Option<Shape> {
    Some(match m {
        // Vector / FP moves: `<mnemonic> dst, src` → `dst = src`.
        Mnemonic::Movq => Shape::Move("movq "),
        Mnemonic::Movd => Shape::Move("movd "),
        Mnemonic::Movaps => Shape::Move("movaps "),
        Mnemonic::Movups => Shape::Move("movups "),
        Mnemonic::Movapd => Shape::Move("movapd "),
        Mnemonic::Movupd => Shape::Move("movupd "),
        Mnemonic::Movdqa => Shape::Move("movdqa "),
        Mnemonic::Movdqu => Shape::Move("movdqu "),
        Mnemonic::Movss => Shape::Move("movss "),
        Mnemonic::Movsd => Shape::Move("movsd "),
        Mnemonic::Movhps => Shape::Move("movhps "),
        Mnemonic::Movlps => Shape::Move("movlps "),
        Mnemonic::Movhpd => Shape::Move("movhpd "),
        Mnemonic::Movlpd => Shape::Move("movlpd "),

        // Packed integer add / sub / multiply.
        Mnemonic::Paddb => Shape::Binary("paddb ", "+"),
        Mnemonic::Paddw => Shape::Binary("paddw ", "+"),
        Mnemonic::Paddd => Shape::Binary("paddd ", "+"),
        Mnemonic::Paddq => Shape::Binary("paddq ", "+"),
        Mnemonic::Paddsb => Shape::Binary("paddsb ", "+"),
        Mnemonic::Paddsw => Shape::Binary("paddsw ", "+"),
        Mnemonic::Paddusb => Shape::Binary("paddusb ", "+"),
        Mnemonic::Paddusw => Shape::Binary("paddusw ", "+"),
        Mnemonic::Psubb => Shape::Binary("psubb ", "-"),
        Mnemonic::Psubw => Shape::Binary("psubw ", "-"),
        Mnemonic::Psubd => Shape::Binary("psubd ", "-"),
        Mnemonic::Psubq => Shape::Binary("psubq ", "-"),
        Mnemonic::Psubsb => Shape::Binary("psubsb ", "-"),
        Mnemonic::Psubsw => Shape::Binary("psubsw ", "-"),
        Mnemonic::Psubusb => Shape::Binary("psubusb ", "-"),
        Mnemonic::Psubusw => Shape::Binary("psubusw ", "-"),
        Mnemonic::Pmullw => Shape::Binary("pmullw ", "*"),
        Mnemonic::Pmulhw => Shape::Binary("pmulhw ", "*"),
        Mnemonic::Pmulhuw => Shape::Binary("pmulhuw ", "*"),
        Mnemonic::Pmuludq => Shape::Binary("pmuludq ", "*"),

        // Packed logical.
        Mnemonic::Pand => Shape::Binary("pand ", "&"),
        Mnemonic::Por => Shape::Binary("por ", "|"),
        Mnemonic::Pxor => Shape::Binary("pxor ", "^"),
        Mnemonic::Pandn => Shape::Binary("pandn ", "& ~"),

        // Packed shift (immediate or register count).
        Mnemonic::Psllw => Shape::Binary("psllw ", "<<"),
        Mnemonic::Pslld => Shape::Binary("pslld ", "<<"),
        Mnemonic::Psllq => Shape::Binary("psllq ", "<<"),
        Mnemonic::Psrlw => Shape::Binary("psrlw ", ">>"),
        Mnemonic::Psrld => Shape::Binary("psrld ", ">>"),
        Mnemonic::Psrlq => Shape::Binary("psrlq ", ">>"),
        Mnemonic::Psraw => Shape::Binary("psraw ", ">>"),
        Mnemonic::Psrad => Shape::Binary("psrad ", ">>"),

        // SSE scalar / packed float arithmetic.
        Mnemonic::Addss => Shape::Binary("addss ", "+"),
        Mnemonic::Addsd => Shape::Binary("addsd ", "+"),
        Mnemonic::Addps => Shape::Binary("addps ", "+"),
        Mnemonic::Addpd => Shape::Binary("addpd ", "+"),
        Mnemonic::Subss => Shape::Binary("subss ", "-"),
        Mnemonic::Subsd => Shape::Binary("subsd ", "-"),
        Mnemonic::Subps => Shape::Binary("subps ", "-"),
        Mnemonic::Subpd => Shape::Binary("subpd ", "-"),
        Mnemonic::Mulss => Shape::Binary("mulss ", "*"),
        Mnemonic::Mulsd => Shape::Binary("mulsd ", "*"),
        Mnemonic::Mulps => Shape::Binary("mulps ", "*"),
        Mnemonic::Mulpd => Shape::Binary("mulpd ", "*"),
        Mnemonic::Divss => Shape::Binary("divss ", "/"),
        Mnemonic::Divsd => Shape::Binary("divsd ", "/"),
        Mnemonic::Divps => Shape::Binary("divps ", "/"),
        Mnemonic::Divpd => Shape::Binary("divpd ", "/"),

        // SSE bitwise on float vectors.
        Mnemonic::Xorps => Shape::Binary("xorps ", "^"),
        Mnemonic::Xorpd => Shape::Binary("xorpd ", "^"),
        Mnemonic::Andps => Shape::Binary("andps ", "&"),
        Mnemonic::Andpd => Shape::Binary("andpd ", "&"),
        Mnemonic::Orps => Shape::Binary("orps ", "|"),
        Mnemonic::Orpd => Shape::Binary("orpd ", "|"),
        Mnemonic::Andnps => Shape::Binary("andnps ", "& ~"),
        Mnemonic::Andnpd => Shape::Binary("andnpd ", "& ~"),

        _ => return None,
    })
}