trueno 0.18.0

High-performance SIMD compute library with GPU support, LLM inference engine, and GGUF model loading
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

//! SSE2 elementwise operations (scale, abs, clamp, lerp, fma, relu, sqrt, recip, norms).

#[cfg(target_arch = "x86_64")]
use std::arch::x86_64::*;

/// SSE2 L1 norm.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn norm_l1(a: &[f32]) -> f32 {
    unsafe {
        if a.is_empty() {
            return 0.0;
        }
        let len = a.len();
        let mut i = 0;
        let mut acc = _mm_setzero_ps();
        let sign_mask = _mm_set1_ps(f32::from_bits(0x7FFF_FFFF));
        while i + 4 <= len {
            acc = _mm_add_ps(acc, _mm_and_ps(_mm_loadu_ps(a.as_ptr().add(i)), sign_mask));
            i += 4;
        }
        let mut result = {
            let temp = _mm_add_ps(acc, _mm_movehl_ps(acc, acc));
            let temp = _mm_add_ss(temp, _mm_shuffle_ps(temp, temp, 1));
            _mm_cvtss_f32(temp)
        };
        for &val in &a[i..] {
            result += val.abs();
        }
        result
    }
}

/// SSE2 L-infinity norm.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn norm_linf(a: &[f32]) -> f32 {
    unsafe {
        if a.is_empty() {
            return 0.0;
        }
        let len = a.len();
        let mut i = 0;
        let mut max_vec = _mm_setzero_ps();
        let sign_mask = _mm_set1_ps(f32::from_bits(0x7FFF_FFFF));
        while i + 4 <= len {
            let va = _mm_loadu_ps(a.as_ptr().add(i));
            max_vec = _mm_max_ps(max_vec, _mm_and_ps(va, sign_mask));
            i += 4;
        }
        let mut result = {
            let temp = _mm_max_ps(max_vec, _mm_movehl_ps(max_vec, max_vec));
            let temp = _mm_max_ss(temp, _mm_shuffle_ps(temp, temp, 1));
            _mm_cvtss_f32(temp)
        };
        for &val in &a[i..] {
            let abs_val = val.abs();
            if abs_val > result {
                result = abs_val;
            }
        }
        result
    }
}

/// SSE2 scalar multiply.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn scale(a: &[f32], scalar: f32, result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        let scalar_vec = _mm_set1_ps(scalar);
        while i + 4 <= len {
            _mm_storeu_ps(
                result.as_mut_ptr().add(i),
                _mm_mul_ps(_mm_loadu_ps(a.as_ptr().add(i)), scalar_vec),
            );
            i += 4;
        }
        for j in i..len {
            result[j] = a[j] * scalar;
        }
    }
}

/// SSE2 absolute value.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn abs(a: &[f32], result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        let sign_mask = _mm_set1_ps(f32::from_bits(0x7FFF_FFFF));
        while i + 4 <= len {
            _mm_storeu_ps(
                result.as_mut_ptr().add(i),
                _mm_and_ps(_mm_loadu_ps(a.as_ptr().add(i)), sign_mask),
            );
            i += 4;
        }
        for j in i..len {
            result[j] = a[j].abs();
        }
    }
}

/// SSE2 clamp.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn clamp(a: &[f32], min_val: f32, max_val: f32, result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        let min_vec = _mm_set1_ps(min_val);
        let max_vec = _mm_set1_ps(max_val);
        while i + 4 <= len {
            let va = _mm_loadu_ps(a.as_ptr().add(i));
            _mm_storeu_ps(result.as_mut_ptr().add(i), _mm_min_ps(_mm_max_ps(va, min_vec), max_vec));
            i += 4;
        }
        for j in i..len {
            result[j] = a[j].max(min_val).min(max_val);
        }
    }
}

/// SSE2 linear interpolation.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn lerp(a: &[f32], b: &[f32], t: f32, result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        let t_vec = _mm_set1_ps(t);
        while i + 4 <= len {
            let va = _mm_loadu_ps(a.as_ptr().add(i));
            let vb = _mm_loadu_ps(b.as_ptr().add(i));
            _mm_storeu_ps(
                result.as_mut_ptr().add(i),
                _mm_add_ps(va, _mm_mul_ps(t_vec, _mm_sub_ps(vb, va))),
            );
            i += 4;
        }
        for j in i..len {
            result[j] = a[j] + t * (b[j] - a[j]);
        }
    }
}

/// SSE2 fused multiply-add (emulated, no FMA instruction set).
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn fma(a: &[f32], b: &[f32], c: &[f32], result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        while i + 4 <= len {
            let va = _mm_loadu_ps(a.as_ptr().add(i));
            let vb = _mm_loadu_ps(b.as_ptr().add(i));
            let vc = _mm_loadu_ps(c.as_ptr().add(i));
            _mm_storeu_ps(result.as_mut_ptr().add(i), _mm_add_ps(_mm_mul_ps(va, vb), vc));
            i += 4;
        }
        for j in i..len {
            result[j] = a[j] * b[j] + c[j];
        }
    }
}

/// SSE2 ReLU activation.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn relu(a: &[f32], result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        let zero = _mm_setzero_ps();
        while i + 4 <= len {
            _mm_storeu_ps(
                result.as_mut_ptr().add(i),
                _mm_max_ps(_mm_loadu_ps(a.as_ptr().add(i)), zero),
            );
            i += 4;
        }
        for j in i..len {
            result[j] = a[j].max(0.0);
        }
    }
}

/// SSE2 square root.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn sqrt(a: &[f32], result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        while i + 4 <= len {
            _mm_storeu_ps(result.as_mut_ptr().add(i), _mm_sqrt_ps(_mm_loadu_ps(a.as_ptr().add(i))));
            i += 4;
        }
        for j in i..len {
            result[j] = a[j].sqrt();
        }
    }
}

/// SSE2 reciprocal.
#[inline]
#[target_feature(enable = "sse2")]
// SAFETY: caller ensures preconditions are met for this unsafe function
pub(crate) unsafe fn recip(a: &[f32], result: &mut [f32]) {
    unsafe {
        let len = a.len();
        let mut i = 0;
        let one = _mm_set1_ps(1.0);
        while i + 4 <= len {
            _mm_storeu_ps(
                result.as_mut_ptr().add(i),
                _mm_div_ps(one, _mm_loadu_ps(a.as_ptr().add(i))),
            );
            i += 4;
        }
        for j in i..len {
            result[j] = a[j].recip();
        }
    }
}