1#include <clc/clc.h>
2
3#define VLOAD_VECTORIZE(PRIM_TYPE, ADDR_SPACE) \
4  _CLC_OVERLOAD _CLC_DEF PRIM_TYPE##2 vload2(size_t offset, const ADDR_SPACE PRIM_TYPE *x) { \
5    return (PRIM_TYPE##2)(x[2*offset] , x[2*offset+1]); \
6  } \
7\
8  _CLC_OVERLOAD _CLC_DEF PRIM_TYPE##3 vload3(size_t offset, const ADDR_SPACE PRIM_TYPE *x) { \
9    return (PRIM_TYPE##3)(x[3*offset] , x[3*offset+1], x[3*offset+2]); \
10  } \
11\
12  _CLC_OVERLOAD _CLC_DEF PRIM_TYPE##4 vload4(size_t offset, const ADDR_SPACE PRIM_TYPE *x) { \
13    return (PRIM_TYPE##4)(x[4*offset], x[4*offset+1], x[4*offset+2], x[4*offset+3]); \
14  } \
15\
16  _CLC_OVERLOAD _CLC_DEF PRIM_TYPE##8 vload8(size_t offset, const ADDR_SPACE PRIM_TYPE *x) { \
17    return (PRIM_TYPE##8)(vload4(0, &x[8*offset]), vload4(1, &x[8*offset])); \
18  } \
19\
20  _CLC_OVERLOAD _CLC_DEF PRIM_TYPE##16 vload16(size_t offset, const ADDR_SPACE PRIM_TYPE *x) { \
21    return (PRIM_TYPE##16)(vload8(0, &x[16*offset]), vload8(1, &x[16*offset])); \
22  } \
23
24#define VLOAD_ADDR_SPACES(__CLC_SCALAR_GENTYPE) \
25    VLOAD_VECTORIZE(__CLC_SCALAR_GENTYPE, __private) \
26    VLOAD_VECTORIZE(__CLC_SCALAR_GENTYPE, __local) \
27    VLOAD_VECTORIZE(__CLC_SCALAR_GENTYPE, __constant) \
28    VLOAD_VECTORIZE(__CLC_SCALAR_GENTYPE, __global) \
29
30#define VLOAD_TYPES() \
31    VLOAD_ADDR_SPACES(char) \
32    VLOAD_ADDR_SPACES(uchar) \
33    VLOAD_ADDR_SPACES(short) \
34    VLOAD_ADDR_SPACES(ushort) \
35    VLOAD_ADDR_SPACES(int) \
36    VLOAD_ADDR_SPACES(uint) \
37    VLOAD_ADDR_SPACES(long) \
38    VLOAD_ADDR_SPACES(ulong) \
39    VLOAD_ADDR_SPACES(float) \
40
41VLOAD_TYPES()
42
43#ifdef cl_khr_fp64
44#pragma OPENCL EXTENSION cl_khr_fp64 : enable
45    VLOAD_ADDR_SPACES(double)
46#endif
47