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
|
/*========================== begin_copyright_notice ============================
Copyright (C) 2021-2024 Intel Corporation
SPDX-License-Identifier: MIT
============================= end_copyright_notice ===========================*/
#include <cm-cl/math.h>
#include <cm-cl/vector.h>
using namespace cm;
namespace {
// RTZ rounding mode is needed for some of the steps
// It is fine to use RTZ mode throughout the FP computation
template <int N>
CM_NODEBUG CM_INLINE void
__impl_divrem(vector<int32_t, N> &q, vector<int32_t, N> &r,
vector<int32_t, N> sla, vector<int32_t, N> slb) {
vector<float, N> ah, al, bh, bl;
vector<float, N> y, Qh, Ql, Rh, Rl;
vector<uint32_t, N> la_h, lb_h, la_l, lb_l, lQh, lQl, lR;
vector<uint32_t, N> la, lb, corr_mask;
vector<int32_t, N> sgn_a, sgn_b, sgn_q;
// get signs and |sla|, |slb|
sgn_a = sla >> 31;
sgn_b = slb >> 31;
la = math::absolute(sla);
lb = math::absolute(slb);
// uint32 -> single precision convert, with truncation (RZ mode)
bh = lb;
// convert back to uint32, to get low part
lb_h = bh;
lb_l = lb - lb_h;
// low part of input, in single precision
bl = lb_l;
// uint32 -> single precision convert, with truncation (RZ mode)
ah = la;
// convert back to uint32, to get low part
la_h = ah;
la_l = la - la_h;
// low part of input, in single precision
al = la_l;
// y = RCP(bh)
y = math::reciprocal(bh);
// y = y*(1 - 3*2^(-23))
vector<float, N> sf = as_float(0xb4c00000u);
y = math::mad(sf, y, y);
// Qh = ah*y
Qh = ah * y;
// Qh = (unsigned)Qh, with truncation
lQh = Qh;
// convert lQh back to SP, any rounding mode is fine
Qh = lQh;
// ah - bh*Qh
Rh = math::mad(-bh, Qh, ah);
// al - bl*Qh
Rl = math::mad(-bl, Qh, al);
// Ql = y * (Rh + Rl)
Rl = Rh + Rl;
Ql = y * Rl;
// convert Ql to integer, with truncation
lQl = Ql;
// integer quotient
q = lQh + lQl;
sgn_q = sgn_a ^ sgn_b;
// integer remainder
lR = la - lb * q;
// apply correction if needed
// if (lR >= lb) { q++; lR -= lb; }
corr_mask = merge(0xffffffff, 0u, (lR >= lb));
q += (corr_mask & 1);
lR -= (lb & corr_mask);
// remainder
r = (sgn_a + lR) ^ sgn_a;
q = (sgn_q + q) ^ sgn_q;
}
template <int N>
CM_NODEBUG CM_INLINE void
__impl_divrem(vector<int16_t, N> &q, vector<int16_t, N> &r,
vector<int16_t, N> a, vector<int16_t, N> b) {
vector<float, N> fa = a;
vector<float, N> fb = b;
vector<float, N> fy = math::reciprocal(fb);
// s = 1 + 2^(-20)
vector<float, N> sf = as_float(0x3f800008U);
// a * (1 + 2^(-20))
fa = fa * sf;
// a * (1 + 2^(-20)) * fy
vector<float, N> fq = fa * fy;
// quotient: truncate to signed 16-bit integer
q = fq;
// remainder
r = a - q * b;
}
template <int N>
CM_NODEBUG CM_INLINE void
__impl_divrem(vector<int8_t, N> &q, vector<int8_t, N> &r, vector<int8_t, N> a,
vector<int8_t, N> b) {
vector<int16_t, N> xa = a, xb = b;
vector<int16_t, N> xq, xr;
__impl_divrem(xq, xr, xa, xb);
r = xr;
q = xq;
}
} // namespace
CM_NODEBUG CM_NOINLINE extern "C" int32_t
__vc_builtin_sdiv_i32__rtz_(int32_t a, int32_t b) {
vector<int32_t, 1> va = a, vb = b, vq, vr;
__impl_divrem(vq, vr, va, vb);
return vq[0];
}
CM_NODEBUG CM_NOINLINE extern "C" int16_t __vc_builtin_sdiv_i16(int16_t a,
int16_t b) {
vector<int16_t, 1> va = a, vb = b, vq, vr;
__impl_divrem(vq, vr, va, vb);
return vq[0];
}
CM_NODEBUG CM_NOINLINE extern "C" int8_t __vc_builtin_sdiv_i8(int8_t a,
int8_t b) {
vector<int8_t, 1> va = a, vb = b, vq, vr;
__impl_divrem(vq, vr, va, vb);
return vq[0];
}
#define SDIV(WIDTH) \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<int32_t, WIDTH> \
__vc_builtin_sdiv_v##WIDTH##i32__rtz_(cl_vector<int32_t, WIDTH> a, \
cl_vector<int32_t, WIDTH> b) { \
vector<int32_t, WIDTH> va = a, vb = b, vq, vr; \
__impl_divrem(vq, vr, va, vb); \
return vq.cl_vector(); \
} \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<int16_t, WIDTH> \
__vc_builtin_sdiv_v##WIDTH##i16(cl_vector<int16_t, WIDTH> a, \
cl_vector<int16_t, WIDTH> b) { \
vector<int16_t, WIDTH> va = a, vb = b, vq, vr; \
__impl_divrem(vq, vr, va, vb); \
return vq.cl_vector(); \
} \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<int8_t, WIDTH> \
__vc_builtin_sdiv_v##WIDTH##i8(cl_vector<int8_t, WIDTH> a, \
cl_vector<int8_t, WIDTH> b) { \
vector<int8_t, WIDTH> va = a, vb = b, vq, vr; \
__impl_divrem(vq, vr, va, vb); \
return vq.cl_vector(); \
}
SDIV(1)
SDIV(2)
SDIV(4)
SDIV(8)
SDIV(16)
SDIV(32)
CM_NODEBUG CM_NOINLINE extern "C" int32_t
__vc_builtin_srem_i32__rtz_(int32_t a, int32_t b) {
vector<int32_t, 1> va = a, vb = b, vq, vr;
__impl_divrem(vq, vr, va, vb);
return vr[0];
}
CM_NODEBUG CM_NOINLINE extern "C" int16_t __vc_builtin_srem_i16(int16_t a,
int16_t b) {
vector<int16_t, 1> va = a, vb = b, vq, vr;
__impl_divrem(vq, vr, va, vb);
return vr[0];
}
CM_NODEBUG CM_NOINLINE extern "C" int8_t __vc_builtin_srem_i8(int8_t a,
int8_t b) {
vector<int8_t, 1> va = a, vb = b, vq, vr;
__impl_divrem(vq, vr, va, vb);
return vr[0];
}
#define SREM(WIDTH) \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<int32_t, WIDTH> \
__vc_builtin_srem_v##WIDTH##i32__rtz_(cl_vector<int32_t, WIDTH> a, \
cl_vector<int32_t, WIDTH> b) { \
vector<int32_t, WIDTH> va = a, vb = b, vq, vr; \
__impl_divrem(vq, vr, va, vb); \
return vr.cl_vector(); \
} \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<int16_t, WIDTH> \
__vc_builtin_srem_v##WIDTH##i16(cl_vector<int16_t, WIDTH> a, \
cl_vector<int16_t, WIDTH> b) { \
vector<int16_t, WIDTH> va = a, vb = b, vq, vr; \
__impl_divrem(vq, vr, va, vb); \
return vr.cl_vector(); \
} \
CM_NODEBUG CM_NOINLINE extern "C" cl_vector<int8_t, WIDTH> \
__vc_builtin_srem_v##WIDTH##i8(cl_vector<int8_t, WIDTH> a, \
cl_vector<int8_t, WIDTH> b) { \
vector<int8_t, WIDTH> va = a, vb = b, vq, vr; \
__impl_divrem(vq, vr, va, vb); \
return vr.cl_vector(); \
}
SREM(1)
SREM(2)
SREM(4)
SREM(8)
SREM(16)
SREM(32)
|