1/*
2 * Copyright (c) 2014 Advanced Micro Devices, Inc.
3 * Copyright (c) 2016 Aaron Watry
4 *
5 * Permission is hereby granted, free of charge, to any person obtaining a copy
6 * of this software and associated documentation files (the "Software"), to deal
7 * in the Software without restriction, including without limitation the rights
8 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9 * copies of the Software, and to permit persons to whom the Software is
10 * furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice shall be included in
13 * all copies or substantial portions of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
21 * THE SOFTWARE.
22 */
23#if __CLC_FPSIZE == 32
24#ifdef __CLC_SCALAR
25_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE fdim(__CLC_GENTYPE x, __CLC_GENTYPE y) {
26    if (__builtin_isnan(x) || __builtin_isnan(y))
27        return as_float(QNANBITPATT_SP32);
28    return fmax(x - y, 0.0f);
29}
30#define __CLC_FDIM_VEC(width) \
31_CLC_OVERLOAD _CLC_DEF float##width fdim(float##width x, float##width y) { \
32    /* Determine if x or y is NaN. */ \
33    /* Vector true is -1, i.e. all-bits-set, and NaN==NaN is false. */ \
34    /* If either is NaN, then ~((x==x) & (y==y)) will be 0 (e.g. ~(-1)), as will n. */ \
35    int##width n = ~((x == x) & (y == y)) & QNANBITPATT_SP32; \
36    /* Calculate x-y if x>y, otherwise positive 0, again taking */ \
37    /* advantage of vector true being all-bits-set. */ \
38    int##width r = (x > y) & as_int##width(x - y); \
39    return as_float##width(n | r); \
40}
41__CLC_FDIM_VEC(2)
42__CLC_FDIM_VEC(3)
43__CLC_FDIM_VEC(4)
44__CLC_FDIM_VEC(8)
45__CLC_FDIM_VEC(16)
46#undef __CLC_FDIM_VEC
47#endif
48#endif
49
50#if __CLC_FPSIZE == 64
51#ifdef __CLC_SCALAR
52_CLC_OVERLOAD _CLC_DEF __CLC_GENTYPE fdim(__CLC_GENTYPE x, private __CLC_GENTYPE y) {
53    long n = -(isnan(x) | isnan(y)) & QNANBITPATT_DP64;
54    long r = -(x > y) & as_long(x - y);
55    return as_double(n | r);
56}
57#define __CLC_FDIM_VEC(width) \
58_CLC_OVERLOAD _CLC_DEF double##width fdim(double##width x, double##width y) { \
59    /* See comment in float implementation for explanation. */ \
60    long##width n = ~((x == x) & (y == y)) & QNANBITPATT_DP64; \
61    long##width r = (x > y) & as_long##width(x - y); \
62    return as_double##width(n | r); \
63}
64__CLC_FDIM_VEC(2)
65__CLC_FDIM_VEC(3)
66__CLC_FDIM_VEC(4)
67__CLC_FDIM_VEC(8)
68__CLC_FDIM_VEC(16)
69#undef __CLC_FDIM_VEC
70#endif
71#endif
72