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
|
/*========================== begin_copyright_notice ============================
Copyright (C) 2021-2023 Intel Corporation
SPDX-License-Identifier: MIT
============================= end_copyright_notice ===========================*/
#include <cm-cl/math.h>
#include <cm-cl/vector.h>
#include "../helpers.h"
using namespace cm;
namespace {
template <int N>
CM_NODEBUG CM_INLINE vector<float, N> __impl_uitofp(vector<uint64_t, N> a) {
const vector<uint32_t, N> Zero(0);
const vector<uint32_t, N> Ones(0xffffffff);
const vector<uint32_t, N> One(1);
vector<uint32_t, 2 *N> LoHi = a.template format<uint32_t>();
vector<uint32_t, N> Lo = LoHi.template select<N, 2>(0);
vector<uint32_t, N> Hi = LoHi.template select<N, 2>(1);
vector<uint32_t, N> LZ = cm::math::count_leading_zeros(Hi);
// we need to get that nice first set bit into bit position 23.
// thus we shift our nice pair of values by 63 - 23 - clz,
// some bits will be dropped by shift thus we'll add 1 bits as R bit.
// uint8_t shift = 39 - lz;
vector<uint32_t, N> DroppedBits = vector<uint32_t, N>(39) - LZ;
// SI
vector<uint32_t, N> Sha = DroppedBits & vector<uint32_t, N>(0x3f);
vector<uint32_t, N> Vec32 = vector<int32_t, N>(32);
vector<uint32_t, N> Sh32 = Vec32 - Sha;
auto Flag_large_shift = (Sha >= Vec32);
auto Flag_zero_shift = (Sha == Zero);
vector<uint32_t, N> Mask1 = Ones;
Mask1.merge(Zero, Flag_large_shift);
vector<uint32_t, N> Mask0 = Ones;
Mask0.merge(Zero, Flag_zero_shift);
// partial shift
vector<uint32_t, N> TmpH1 = ((Hi & Mask0) << Sh32) & Mask1;
vector<uint32_t, N> TmpH2 = (Hi >> (Sha - Vec32)) & ~Mask1;
vector<uint32_t, N> TmpL = (Lo >> Sha) & Mask1;
vector<uint32_t, N> Mant = TmpL | TmpH1 | TmpH2;
vector<uint32_t, N> TmpSha = One << (-Sh32);
vector<uint32_t, N> TmpMask = TmpSha - One;
vector<uint32_t, N> StickyH = Hi & ~Mask1;
StickyH = StickyH & TmpMask;
// calculate RS
vector<uint32_t, N> L1 = Lo & ~Mask1;
vector<uint32_t, N> L2 = Lo & (Mask1 >> Sh32);
vector<uint32_t, N> StickyL = L1 | L2;
vector<uint32_t, N> S1 = StickyH | StickyL;
auto S = S1 == Zero;
vector<uint32_t, N> NotS = Zero;
NotS.merge(Ones, S);
// R is set but no S, round to even.
vector<uint32_t, N> R = Mant & One;
Mant = (Mant + One) >> One;
Mant &= ~(NotS & R);
vector<uint32_t, N> Exp = vector<uint32_t, N>(0xbd) - LZ;
vector<uint32_t, N> ResL = Exp << vector<uint32_t, N>(23);
ResL += Mant;
vector<float, N> ResultLarge = ResL.template format<float>();
vector<float, N> ResultSmall = Lo;
auto IsSmallPred = Hi == Zero;
vector<float, N> Result = ResultLarge;
Result.merge(ResultSmall, IsSmallPred);
return Result;
}
template <int N>
CM_NODEBUG CM_INLINE vector<float, N> __impl_sitofp(vector<int64_t, N> a) {
vector<uint64_t, N> Abs = math::absolute(a);
auto Res = __impl_uitofp(Abs);
Res.merge(-Res, a < 0);
return Res;
}
} // namespace
CM_NODEBUG CM_NOINLINE extern "C" float __vc_builtin_uitofp_f32(uint64_t a) {
vector<uint64_t, 1> va = a;
return __impl_uitofp(va)[0];
}
CM_NODEBUG CM_NOINLINE extern "C" float __vc_builtin_sitofp_f32(int64_t a) {
vector<int64_t, 1> va = a;
return __impl_sitofp(va)[0];
}
#define ITOFP(WIDTH) \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<float, WIDTH> \
__vc_builtin_uitofp_v##WIDTH##f32(cl_vector<uint64_t, WIDTH> a) { \
vector<uint64_t, WIDTH> va{a}; \
return __impl_uitofp(va).cl_vector(); \
} \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<float, WIDTH> \
__vc_builtin_sitofp_v##WIDTH##f32(cl_vector<int64_t, WIDTH> a) { \
vector<int64_t, WIDTH> va{a}; \
return __impl_sitofp(va).cl_vector(); \
}
ITOFP(1)
ITOFP(2)
ITOFP(4)
ITOFP(8)
ITOFP(16)
ITOFP(32)
|
