memx 0.2.0

memory functions like libc memcmp(), memchr(), memmem(), memcpy(), memset()
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

use crate::mem as basic;
use crate::utils::*;
use crate::RangeError;

#[cfg(target_arch = "x86")]
use core::arch::x86 as mmx;
#[cfg(target_arch = "x86_64")]
use core::arch::x86_64 as mmx;

use mmx::__m128i;
use mmx::_mm_load_si128;
use mmx::_mm_loadu_si128;
use mmx::_mm_store_si128;
use mmx::_mm_storeu_si128;
/*
use mmx::__m256i;
use mmx::_mm256_load_si256;
use mmx::_mm256_loadu_si256;
use mmx::_mm256_store_si256;
use mmx::_mm256_storeu_si256;
*/

#[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
use super::cpuid;

use core::sync::atomic::AtomicPtr;
use core::sync::atomic::Ordering;
type _FuncType = fn(&mut [u8], &[u8]) -> Result<(), RangeError>;

const _FUNC: _FuncType = _fnptr_setup_func;
static _FUNC_PTR_ATOM: AtomicPtr<_FuncType> = AtomicPtr::new(_FUNC as *mut _FuncType);

#[inline(never)]
fn _fnptr_setup_func(dst: &mut [u8], src: &[u8]) -> Result<(), RangeError> {
    #[cfg(target_arch = "x86_64")]
    let func = if cpuid::has_avx2() {
        _memcpy_avx2
    } else {
        _memcpy_sse2
    };
    #[cfg(target_arch = "x86")]
    let func = if cpuid::has_avx2() {
        _memcpy_avx2
    } else if cpuid::has_sse2() {
        _memcpy_sse2
    } else {
        _memcpy_basic
    };
    //
    _FUNC_PTR_ATOM.store(func as *mut _FuncType, Ordering::Relaxed);
    unsafe { func(dst, src) }
}

#[inline(always)]
pub(crate) fn _memcpy_impl(dst: &mut [u8], src: &[u8]) -> Result<(), RangeError> {
    let func_u = _FUNC_PTR_ATOM.load(Ordering::Relaxed);
    #[allow(clippy::crosspointer_transmute)]
    let func: _FuncType = unsafe { core::mem::transmute(func_u) };
    func(dst, src)
}

unsafe fn _memcpy_basic(dst: &mut [u8], src: &[u8]) -> Result<(), RangeError> {
    basic::_memcpy_impl(dst, src)
}

#[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
#[target_feature(enable = "sse2")]
#[allow(clippy::missing_safety_doc)]
pub unsafe fn _memcpy_sse2(dst: &mut [u8], src: &[u8]) -> Result<(), RangeError> {
    _memcpy_sse2_impl(dst, src)
}

#[cfg(any(target_arch = "x86_64", target_arch = "x86"))]
#[target_feature(enable = "avx2")]
#[allow(clippy::missing_safety_doc)]
pub unsafe fn _memcpy_avx2(dst: &mut [u8], src: &[u8]) -> Result<(), RangeError> {
    _memcpy_sse2_impl(dst, src)
}

#[inline(always)]
fn _memcpy_sse2_impl(dst: &mut [u8], src: &[u8]) -> Result<(), RangeError> {
    if src.is_empty() {
        return Ok(());
    }
    let dst_len = dst.len();
    let src_len = src.len();
    if dst_len < src_len {
        return Err(RangeError);
    }
    //
    let mut a_ptr = dst.as_mut_ptr();
    let mut b_ptr = src.as_ptr();
    let end_ptr = unsafe { a_ptr.add(src_len) };
    b_ptr.prefetch_read_data();
    //
    if src_len >= 16 {
        // to a aligned pointer
        {
            if !a_ptr.is_aligned_u128() {
                #[cfg(not(feature = "test_alignment_check"))]
                {
                    unsafe { _cpy_b16_uu_x1(a_ptr, b_ptr) };
                    let remaining_align = 0x10_usize - ((a_ptr as usize) & 0x0F_usize);
                    a_ptr = unsafe { a_ptr.add(remaining_align) };
                    b_ptr = unsafe { b_ptr.add(remaining_align) };
                }
                #[cfg(feature = "test_alignment_check")]
                {
                    let (ap, bp) = basic::_cpy_to_aligned_u128(a_ptr, b_ptr);
                    a_ptr = ap;
                    b_ptr = bp;
                }
            }
        }
        // the loop
        if b_ptr.is_aligned_u128() {
            {
                let unroll = 16;
                let loop_size = 16;
                while a_ptr.is_not_over(end_ptr, loop_size * unroll) {
                    b_ptr.prefetch_read_data();
                    _cpy_b16_aa_x16(a_ptr, b_ptr);
                    a_ptr = unsafe { a_ptr.add(loop_size * unroll) };
                    b_ptr = unsafe { b_ptr.add(loop_size * unroll) };
                }
            }
            {
                let unroll = 8;
                let loop_size = 16;
                while a_ptr.is_not_over(end_ptr, loop_size * unroll) {
                    b_ptr.prefetch_read_data();
                    _cpy_b16_aa_x8(a_ptr, b_ptr);
                    a_ptr = unsafe { a_ptr.add(loop_size * unroll) };
                    b_ptr = unsafe { b_ptr.add(loop_size * unroll) };
                }
            }
            {
                let unroll = 1;
                let loop_size = 16;
                while a_ptr.is_not_over(end_ptr, loop_size * unroll) {
                    unsafe { _cpy_b16_aa_x1(a_ptr, b_ptr) };
                    a_ptr = unsafe { a_ptr.add(loop_size * unroll) };
                    b_ptr = unsafe { b_ptr.add(loop_size * unroll) };
                }
            }
        } else {
            {
                let unroll = 16;
                let loop_size = 16;
                while a_ptr.is_not_over(end_ptr, loop_size * unroll) {
                    b_ptr.prefetch_read_data();
                    _cpy_b16_au_x16(a_ptr, b_ptr);
                    a_ptr = unsafe { a_ptr.add(loop_size * unroll) };
                    b_ptr = unsafe { b_ptr.add(loop_size * unroll) };
                }
            }
            {
                let unroll = 8;
                let loop_size = 16;
                while a_ptr.is_not_over(end_ptr, loop_size * unroll) {
                    b_ptr.prefetch_read_data();
                    _cpy_b16_au_x8(a_ptr, b_ptr);
                    a_ptr = unsafe { a_ptr.add(loop_size * unroll) };
                    b_ptr = unsafe { b_ptr.add(loop_size * unroll) };
                }
            }
            {
                let unroll = 1;
                let loop_size = 16;
                while a_ptr.is_not_over(end_ptr, loop_size * unroll) {
                    unsafe { _cpy_b16_au_x1(a_ptr, b_ptr) };
                    a_ptr = unsafe { a_ptr.add(loop_size * unroll) };
                    b_ptr = unsafe { b_ptr.add(loop_size * unroll) };
                }
            }
        }
    }
    // the remaining data is the max: 15 bytes.
    basic::_memcpy_remaining_15_bytes_impl(a_ptr, b_ptr, end_ptr);
    Ok(())
}

#[inline(always)]
unsafe fn _cpy_b16_uu_x1(a_ptr: *mut u8, b_ptr: *const u8) {
    unsafe {
        let mm_0_b = _mm_loadu_si128(b_ptr as *const __m128i);
        _mm_storeu_si128(a_ptr as *mut __m128i, mm_0_b);
    }
}

#[inline(always)]
unsafe fn _cpy_b16_au_x1(a_ptr: *mut u8, b_ptr: *const u8) {
    unsafe {
        let mm_0_b = _mm_loadu_si128(b_ptr as *const __m128i);
        _mm_store_si128(a_ptr as *mut __m128i, mm_0_b);
    }
}

#[inline(always)]
fn _cpy_b16_au_x2(a_ptr: *mut u8, b_ptr: *const u8) {
    unsafe { _cpy_b16_au_x1(a_ptr, b_ptr) };
    unsafe { _cpy_b16_au_x1(a_ptr.add(16), b_ptr.add(16)) }
}

#[inline(always)]
fn _cpy_b16_au_x4(a_ptr: *mut u8, b_ptr: *const u8) {
    _cpy_b16_au_x2(a_ptr, b_ptr);
    _cpy_b16_au_x2(unsafe { a_ptr.add(16 * 2) }, unsafe { b_ptr.add(16 * 2) });
}

#[inline(always)]
fn _cpy_b16_au_x8(a_ptr: *mut u8, b_ptr: *const u8) {
    _cpy_b16_au_x4(a_ptr, b_ptr);
    _cpy_b16_au_x4(unsafe { a_ptr.add(16 * 4) }, unsafe { b_ptr.add(16 * 4) });
}

#[inline(always)]
fn _cpy_b16_au_x16(a_ptr: *mut u8, b_ptr: *const u8) {
    _cpy_b16_au_x8(a_ptr, b_ptr);
    _cpy_b16_au_x8(unsafe { a_ptr.add(16 * 8) }, unsafe { b_ptr.add(16 * 8) });
}

#[inline(always)]
unsafe fn _cpy_b16_aa_x1(a_ptr: *mut u8, b_ptr: *const u8) {
    unsafe {
        let mm_0_b = _mm_load_si128(b_ptr as *const __m128i);
        _mm_store_si128(a_ptr as *mut __m128i, mm_0_b);
    }
}

#[inline(always)]
fn _cpy_b16_aa_x2(a_ptr: *mut u8, b_ptr: *const u8) {
    unsafe { _cpy_b16_aa_x1(a_ptr, b_ptr) };
    unsafe { _cpy_b16_aa_x1(a_ptr.add(16), b_ptr.add(16)) }
}

#[inline(always)]
fn _cpy_b16_aa_x4(a_ptr: *mut u8, b_ptr: *const u8) {
    _cpy_b16_aa_x2(a_ptr, b_ptr);
    _cpy_b16_aa_x2(unsafe { a_ptr.add(16 * 2) }, unsafe { b_ptr.add(16 * 2) });
}

#[inline(always)]
fn _cpy_b16_aa_x8(a_ptr: *mut u8, b_ptr: *const u8) {
    _cpy_b16_aa_x4(a_ptr, b_ptr);
    _cpy_b16_aa_x4(unsafe { a_ptr.add(16 * 4) }, unsafe { b_ptr.add(16 * 4) });
}

#[inline(always)]
fn _cpy_b16_aa_x16(a_ptr: *mut u8, b_ptr: *const u8) {
    _cpy_b16_aa_x8(a_ptr, b_ptr);
    _cpy_b16_aa_x8(unsafe { a_ptr.add(16 * 8) }, unsafe { b_ptr.add(16 * 8) });
}

#[cfg(test)]
mod disasm {
    use super::*;
    //
    #[test]
    fn do_procs() {
        let mut a = b"       ".to_vec();
        let b = b"abcdefg".to_vec();
        let a = a.as_mut_slice();
        let b = b.as_slice();
        assert_eq!(do_proc_basic(a, b), Ok(()));
        assert_eq!(do_proc_sse2(a, b), Ok(()));
        assert_eq!(do_proc_avx2(a, b), Ok(()));
    }

    #[inline(never)]
    fn do_proc_basic(a: &mut [u8], b: &[u8]) -> Result<(), RangeError> {
        unsafe { _memcpy_basic(a, b) }
    }
    #[inline(never)]
    fn do_proc_sse2(a: &mut [u8], b: &[u8]) -> Result<(), RangeError> {
        unsafe { _memcpy_sse2(a, b) }
    }
    #[inline(never)]
    fn do_proc_avx2(a: &mut [u8], b: &[u8]) -> Result<(), RangeError> {
        unsafe { _memcpy_avx2(a, b) }
    }
}