core/stdarch/crates/core_arch/src/x86_64/
avx512fp16.rs

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
use crate::core_arch::x86::*;
#[cfg(test)]
use stdarch_test::assert_instr;

/// Convert the signed 64-bit integer b to a half-precision (16-bit) floating-point element, store the
/// result in the lower element of dst, and copy the upper 3 packed elements from a to the upper elements
/// of dst.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvti64_sh)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtsi2sh))]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvti64_sh(a: __m128h, b: i64) -> __m128h {
    vcvtsi642sh(a, b, _MM_FROUND_CUR_DIRECTION)
}

/// Convert the signed 64-bit integer b to a half-precision (16-bit) floating-point element, store the
/// result in the lower element of dst, and copy the upper 3 packed elements from a to the upper elements
/// of dst.
///
/// Rounding is done according to the rounding parameter, which can be one of:
///
/// * [`_MM_FROUND_TO_NEAREST_INT`] | [`_MM_FROUND_NO_EXC`] : round to nearest and suppress exceptions
/// * [`_MM_FROUND_TO_NEG_INF`] | [`_MM_FROUND_NO_EXC`] : round down and suppress exceptions
/// * [`_MM_FROUND_TO_POS_INF`] | [`_MM_FROUND_NO_EXC`] : round up and suppress exceptions
/// * [`_MM_FROUND_TO_ZERO`] | [`_MM_FROUND_NO_EXC`] : truncate and suppress exceptions
/// * [`_MM_FROUND_CUR_DIRECTION`] : use `MXCSR.RC` - see [`_MM_SET_ROUNDING_MODE`]
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvt_roundi64_sh)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtsi2sh, ROUNDING = 8))]
#[rustc_legacy_const_generics(2)]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvt_roundi64_sh<const ROUNDING: i32>(a: __m128h, b: i64) -> __m128h {
    static_assert_rounding!(ROUNDING);
    vcvtsi642sh(a, b, ROUNDING)
}

/// Convert the unsigned 64-bit integer b to a half-precision (16-bit) floating-point element, store the
/// result in the lower element of dst, and copy the upper 1 packed elements from a to the upper elements
/// of dst.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvtu64_sh)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtusi2sh))]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvtu64_sh(a: __m128h, b: u64) -> __m128h {
    vcvtusi642sh(a, b, _MM_FROUND_CUR_DIRECTION)
}

/// Convert the unsigned 64-bit integer b to a half-precision (16-bit) floating-point element, store the
/// result in the lower element of dst, and copy the upper 1 packed elements from a to the upper elements
/// of dst.
///
/// Rounding is done according to the rounding parameter, which can be one of:
///
/// * [`_MM_FROUND_TO_NEAREST_INT`] | [`_MM_FROUND_NO_EXC`] : round to nearest and suppress exceptions
/// * [`_MM_FROUND_TO_NEG_INF`] | [`_MM_FROUND_NO_EXC`] : round down and suppress exceptions
/// * [`_MM_FROUND_TO_POS_INF`] | [`_MM_FROUND_NO_EXC`] : round up and suppress exceptions
/// * [`_MM_FROUND_TO_ZERO`] | [`_MM_FROUND_NO_EXC`] : truncate and suppress exceptions
/// * [`_MM_FROUND_CUR_DIRECTION`] : use `MXCSR.RC` - see [`_MM_SET_ROUNDING_MODE`]
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvt_roundu64_sh)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtusi2sh, ROUNDING = 8))]
#[rustc_legacy_const_generics(2)]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvt_roundu64_sh<const ROUNDING: i32>(a: __m128h, b: u64) -> __m128h {
    static_assert_rounding!(ROUNDING);
    vcvtusi642sh(a, b, ROUNDING)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit integer, and store
/// the result in dst.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvtsh_i64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtsh2si))]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvtsh_i64(a: __m128h) -> i64 {
    vcvtsh2si64(a, _MM_FROUND_CUR_DIRECTION)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit integer, and store
/// the result in dst.
///
/// Rounding is done according to the rounding parameter, which can be one of:
///
/// * [`_MM_FROUND_TO_NEAREST_INT`] | [`_MM_FROUND_NO_EXC`] : round to nearest and suppress exceptions
/// * [`_MM_FROUND_TO_NEG_INF`] | [`_MM_FROUND_NO_EXC`] : round down and suppress exceptions
/// * [`_MM_FROUND_TO_POS_INF`] | [`_MM_FROUND_NO_EXC`] : round up and suppress exceptions
/// * [`_MM_FROUND_TO_ZERO`] | [`_MM_FROUND_NO_EXC`] : truncate and suppress exceptions
/// * [`_MM_FROUND_CUR_DIRECTION`] : use `MXCSR.RC` - see [`_MM_SET_ROUNDING_MODE`]
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvt_roundsh_i64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtsh2si, ROUNDING = 8))]
#[rustc_legacy_const_generics(1)]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvt_roundsh_i64<const ROUNDING: i32>(a: __m128h) -> i64 {
    static_assert_rounding!(ROUNDING);
    vcvtsh2si64(a, ROUNDING)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit unsigned integer, and store
/// the result in dst.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvtsh_u64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtsh2usi))]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvtsh_u64(a: __m128h) -> u64 {
    vcvtsh2usi64(a, _MM_FROUND_CUR_DIRECTION)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit unsigned integer, and store
/// the result in dst.
///
/// Rounding is done according to the rounding parameter, which can be one of:
///
/// * [`_MM_FROUND_TO_NEAREST_INT`] | [`_MM_FROUND_NO_EXC`] : round to nearest and suppress exceptions
/// * [`_MM_FROUND_TO_NEG_INF`] | [`_MM_FROUND_NO_EXC`] : round down and suppress exceptions
/// * [`_MM_FROUND_TO_POS_INF`] | [`_MM_FROUND_NO_EXC`] : round up and suppress exceptions
/// * [`_MM_FROUND_TO_ZERO`] | [`_MM_FROUND_NO_EXC`] : truncate and suppress exceptions
/// * [`_MM_FROUND_CUR_DIRECTION`] : use `MXCSR.RC` - see [`_MM_SET_ROUNDING_MODE`]
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvt_roundsh_u64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvtsh2usi, ROUNDING = 8))]
#[rustc_legacy_const_generics(1)]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvt_roundsh_u64<const ROUNDING: i32>(a: __m128h) -> u64 {
    static_assert_rounding!(ROUNDING);
    vcvtsh2usi64(a, ROUNDING)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit integer with truncation,
/// and store the result in dst.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvttsh_i64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvttsh2si))]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvttsh_i64(a: __m128h) -> i64 {
    vcvttsh2si64(a, _MM_FROUND_CUR_DIRECTION)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit integer with truncation,
/// and store the result in dst.
///
/// Exceptions can be suppressed by passing _MM_FROUND_NO_EXC in the sae parameter.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvtt_roundsh_i64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvttsh2si, SAE = 8))]
#[rustc_legacy_const_generics(1)]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvtt_roundsh_i64<const SAE: i32>(a: __m128h) -> i64 {
    static_assert_sae!(SAE);
    vcvttsh2si64(a, SAE)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit unsigned integer with truncation,
/// and store the result in dst.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvttsh_u64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvttsh2usi))]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvttsh_u64(a: __m128h) -> u64 {
    vcvttsh2usi64(a, _MM_FROUND_CUR_DIRECTION)
}

/// Convert the lower half-precision (16-bit) floating-point element in a to a 64-bit unsigned integer with truncation,
/// and store the result in dst.
///
/// Exceptions can be suppressed by passing _MM_FROUND_NO_EXC in the sae parameter.
///
/// [Intel's documentation](https://software.intel.com/sites/landingpage/IntrinsicsGuide/#text=_mm_cvtt_roundsh_u64)
#[inline]
#[target_feature(enable = "avx512fp16")]
#[cfg_attr(test, assert_instr(vcvttsh2usi, SAE = 8))]
#[rustc_legacy_const_generics(1)]
#[unstable(feature = "stdarch_x86_avx512_f16", issue = "127213")]
pub unsafe fn _mm_cvtt_roundsh_u64<const SAE: i32>(a: __m128h) -> u64 {
    static_assert_sae!(SAE);
    vcvttsh2usi64(a, SAE)
}

#[allow(improper_ctypes)]
extern "C" {
    #[link_name = "llvm.x86.avx512fp16.vcvtsi642sh"]
    fn vcvtsi642sh(a: __m128h, b: i64, rounding: i32) -> __m128h;
    #[link_name = "llvm.x86.avx512fp16.vcvtusi642sh"]
    fn vcvtusi642sh(a: __m128h, b: u64, rounding: i32) -> __m128h;
    #[link_name = "llvm.x86.avx512fp16.vcvtsh2si64"]
    fn vcvtsh2si64(a: __m128h, rounding: i32) -> i64;
    #[link_name = "llvm.x86.avx512fp16.vcvtsh2usi64"]
    fn vcvtsh2usi64(a: __m128h, rounding: i32) -> u64;
    #[link_name = "llvm.x86.avx512fp16.vcvttsh2si64"]
    fn vcvttsh2si64(a: __m128h, sae: i32) -> i64;
    #[link_name = "llvm.x86.avx512fp16.vcvttsh2usi64"]
    fn vcvttsh2usi64(a: __m128h, sae: i32) -> u64;
}

#[cfg(test)]
mod tests {
    use crate::core_arch::{x86::*, x86_64::*};
    use stdarch_test::simd_test;

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvti64_sh() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvti64_sh(a, 10);
        let e = _mm_setr_ph(10.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        assert_eq_m128h(r, e);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvt_roundi64_sh() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvt_roundi64_sh::<{ _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC }>(a, 10);
        let e = _mm_setr_ph(10.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        assert_eq_m128h(r, e);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvtu64_sh() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvtu64_sh(a, 10);
        let e = _mm_setr_ph(10.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        assert_eq_m128h(r, e);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvt_roundu64_sh() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvt_roundu64_sh::<{ _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC }>(a, 10);
        let e = _mm_setr_ph(10.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        assert_eq_m128h(r, e);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvtsh_i64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvtsh_i64(a);
        assert_eq!(r, 1);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvt_roundsh_i64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvt_roundsh_i64::<{ _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC }>(a);
        assert_eq!(r, 1);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvtsh_u64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvtsh_u64(a);
        assert_eq!(r, 1);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvt_roundsh_u64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvt_roundsh_u64::<{ _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC }>(a);
        assert_eq!(r, 1);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvttsh_i64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvttsh_i64(a);
        assert_eq!(r, 1);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvtt_roundsh_i64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvtt_roundsh_i64::<{ _MM_FROUND_TO_NEAREST_INT | _MM_FROUND_NO_EXC }>(a);
        assert_eq!(r, 1);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvttsh_u64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvttsh_u64(a);
        assert_eq!(r, 1);
    }

    #[simd_test(enable = "avx512fp16")]
    unsafe fn test_mm_cvtt_roundsh_u64() {
        let a = _mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0);
        let r = _mm_cvtt_roundsh_u64::<_MM_FROUND_NO_EXC>(a);
        assert_eq!(r, 1);
    }
}