opencl.h File Reference

Maps OpenCL C 1.2 onto the kernel model of <cccc/kernel.h>. Included automatically ahead of every input read as OpenCL C (-x cl, .cl). See docs/opencl.md. More...

#include <cccc/kernel.h>

Include dependency graph for opencl.h:

Go to the source code of this file.

Macros
#define	__CL_VECTOR(base, name, bytes)    typedef base name __attribute__((vector_size(bytes)))
#define	__CL_VECTOR3(base, name, bytes)    typedef base name __attribute__((vector_size(bytes), cccc_visible_lanes(3)))
Version
#define	__OPENCL_C_VERSION__   120
#define	__OPENCL_VERSION__   120
#define	CL_VERSION_1_0   100
#define	CL_VERSION_1_1   110
#define	CL_VERSION_1_2   120
#define	__ENDIAN_LITTLE__   1
Qualifiers
#define	__kernel   [[cccc::__kernel__]]
#define	kernel   [[cccc::__kernel__]]
#define	__global   [[cccc::__global__]]
#define	global   [[cccc::__global__]]
#define	__local   [[cccc::__local__]]
#define	local   [[cccc::__local__]]
#define	__constant   [[cccc::__constant__]]
#define	constant   [[cccc::__constant__]]
#define	__private   [[cccc::__private__]]
#define	private   [[cccc::__private__]]
Work-item functions
Map onto the cccc_* builtins of kernel.h.
#define	get_work_dim()   cccc_work_dim()
#define	get_global_id(d)   cccc_global_id(d)
#define	get_local_id(d)   cccc_local_id(d)
#define	get_group_id(d)   cccc_group_id(d)
#define	get_global_size(d)   cccc_global_size(d)
#define	get_local_size(d)   cccc_local_size(d)
#define	get_num_groups(d)   cccc_num_groups(d)
#define	get_global_offset(d)   ((size_t)0)
Barriers and fences
#define	CLK_LOCAL_MEM_FENCE   CCCC_LOCAL_FENCE
#define	CLK_GLOBAL_MEM_FENCE   CCCC_GLOBAL_FENCE
#define	barrier(flags)   cccc_barrier(flags)
#define	mem_fence(flags)   __atomic_thread_fence(__ATOMIC_SEQ_CST)
#define	read_mem_fence(flags)   __atomic_thread_fence(__ATOMIC_SEQ_CST)
#define	write_mem_fence(flags)   __atomic_thread_fence(__ATOMIC_SEQ_CST)
Atomics
Return the old value. Also available under the OpenCL 1.0 atom_* spellings.
#define	atomic_add(p, v)   __atomic_fetch_add(p, v, __ATOMIC_SEQ_CST)
#define	atomic_sub(p, v)   __atomic_fetch_sub(p, v, __ATOMIC_SEQ_CST)
#define	atomic_and(p, v)   __atomic_fetch_and(p, v, __ATOMIC_SEQ_CST)
#define	atomic_or(p, v)   __atomic_fetch_or(p, v, __ATOMIC_SEQ_CST)
#define	atomic_xor(p, v)   __atomic_fetch_xor(p, v, __ATOMIC_SEQ_CST)
#define	atomic_inc(p)   __atomic_fetch_add(p, 1, __ATOMIC_SEQ_CST)
#define	atomic_dec(p)   __atomic_fetch_sub(p, 1, __ATOMIC_SEQ_CST)
#define	__CL_XCHG_F32(p, val)
#define	atomic_xchg(p, val)
#define	atomic_cmpxchg(p, cmp, val)
#define	__CL_ATOMIC_EXTREMUM(p, val, stops)
#define	atomic_min(p, val)   __CL_ATOMIC_EXTREMUM(p, val, __cl_v < __cl_o)
#define	atomic_max(p, val)   __CL_ATOMIC_EXTREMUM(p, val, __cl_v > __cl_o)
#define	atom_add   atomic_add
#define	atom_sub   atomic_sub
#define	atom_and   atomic_and
#define	atom_or   atomic_or
#define	atom_xor   atomic_xor
#define	atom_xchg   atomic_xchg
#define	atom_inc   atomic_inc
#define	atom_dec   atomic_dec
#define	atom_cmpxchg   atomic_cmpxchg
#define	atom_min   atomic_min
#define	atom_max   atomic_max

Typedefs
typedef unsigned	__cl_generic_uint
Scalar typedefs
typedef unsigned char	uchar
typedef unsigned short	ushort
typedef unsigned int	uint
typedef unsigned long	ulong

Functions
	__CL_VECTOR (signed char, char16, 16)
	__CL_VECTOR (unsigned char, uchar16, 16)
	__CL_VECTOR (short, short8, 16)
	__CL_VECTOR (short, short16, 32)
	__CL_VECTOR (unsigned short, ushort8, 16)
	__CL_VECTOR (unsigned short, ushort16, 32)
	__CL_VECTOR3 (int, int3, 16)
	__CL_VECTOR (int, int4, 16)
	__CL_VECTOR (int, int8, 32)
	__CL_VECTOR (int, int16, 64)
	__CL_VECTOR3 (unsigned int, uint3, 16)
	__CL_VECTOR (unsigned int, uint4, 16)
	__CL_VECTOR (unsigned int, uint8, 32)
	__CL_VECTOR (unsigned int, uint16, 64)
	__CL_VECTOR (long, long2, 16)
	__CL_VECTOR3 (long, long3, 32)
	__CL_VECTOR (long, long4, 32)
	__CL_VECTOR (long, long8, 64)
	__CL_VECTOR (unsigned long, ulong2, 16)
	__CL_VECTOR3 (unsigned long, ulong3, 32)
	__CL_VECTOR (unsigned long, ulong4, 32)
	__CL_VECTOR (unsigned long, ulong8, 64)
	__CL_VECTOR3 (float, float3, 16)
	__CL_VECTOR (float, float4, 16)
	__CL_VECTOR (float, float8, 32)
	__CL_VECTOR (float, float16, 64)

Detailed Description

Maps OpenCL C 1.2 onto the kernel model of <cccc/kernel.h>. Included automatically ahead of every input read as OpenCL C (-x cl, .cl). See docs/opencl.md.

Macro Definition Documentation

__CL_ATOMIC_EXTREMUM

#define __CL_ATOMIC_EXTREMUM	(		p,
			val,
			stops
	)

Value:

    ({                                                                         \
        auto __cl_p                   = (p);                                   \
        typeof_unqual(*__cl_p) __cl_v = (val);                                 \
        typeof_unqual(*__cl_p) __cl_o =                                        \
            __atomic_load_n(__cl_p, __ATOMIC_SEQ_CST);                         \
        while (stops && !__atomic_compare_exchange_n(__cl_p, &__cl_o, __cl_v,  \
                                                     0, __ATOMIC_SEQ_CST,      \
                                                     __ATOMIC_SEQ_CST)) {      \
        }                                                                      \
        __cl_o;                                                                \
    })

__CL_VECTOR

#define __CL_VECTOR	(		base,
			name,
			bytes
	)		typedef base name __attribute__((vector_size(bytes)))

__CL_VECTOR3

#define __CL_VECTOR3	(		base,
			name,
			bytes
	)		typedef base name __attribute__((vector_size(bytes), cccc_visible_lanes(3)))

__CL_XCHG_F32

#define __CL_XCHG_F32	(		p,
			val
	)

Value:

    ({                                                                         \
        union {                                                                \
            float    f;                                                        \
            unsigned u;                                                        \
        } __cl_in  = {.f = (val)}, __cl_out;                                   \
        __cl_out.u = __atomic_exchange_n((__cl_generic_uint *)(p), __cl_in.u,  \
                                         __ATOMIC_SEQ_CST);                    \
        __cl_out.f;                                                            \
    })

__constant

#define __constant   [[cccc::__constant__]]

__ENDIAN_LITTLE__

#define __ENDIAN_LITTLE__   1

__global

#define __global   [[cccc::__global__]]

__kernel

#define __kernel   [[cccc::__kernel__]]

__local

#define __local   [[cccc::__local__]]

__OPENCL_C_VERSION__

#define __OPENCL_C_VERSION__   120

__OPENCL_VERSION__

#define __OPENCL_VERSION__   120

__private

#define __private   [[cccc::__private__]]

atom_add

#define atom_add   atomic_add

atom_and

#define atom_and   atomic_and

atom_cmpxchg

#define atom_cmpxchg   atomic_cmpxchg

atom_dec

#define atom_dec   atomic_dec

atom_inc

#define atom_inc   atomic_inc

atom_max

#define atom_max   atomic_max

atom_min

#define atom_min   atomic_min

atom_or

#define atom_or   atomic_or

atom_sub

#define atom_sub   atomic_sub

atom_xchg

#define atom_xchg   atomic_xchg

atom_xor

#define atom_xor   atomic_xor

atomic_add

#define atomic_add	(		p,
			v
	)		__atomic_fetch_add(p, v, __ATOMIC_SEQ_CST)

atomic_and

#define atomic_and	(		p,
			v
	)		__atomic_fetch_and(p, v, __ATOMIC_SEQ_CST)

atomic_cmpxchg

#define atomic_cmpxchg	(		p,
			cmp,
			val
	)

Value:

    ({                                                                         \
        auto __cl_p                   = (p);                                   \
        typeof_unqual(*__cl_p) __cl_e = (cmp);                                 \
        __atomic_compare_exchange_n(__cl_p, &__cl_e, (val), 0,                 \
                                    __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST);       \
        __cl_e;                                                                \
    })

atomic_dec

#define atomic_dec

(

p

)

__atomic_fetch_sub(p, 1, __ATOMIC_SEQ_CST)

atomic_inc

#define atomic_inc

(

p

)

__atomic_fetch_add(p, 1, __ATOMIC_SEQ_CST)

atomic_max

#define atomic_max	(		p,
			val
	)		__CL_ATOMIC_EXTREMUM(p, val, __cl_v > __cl_o)

atomic_min

#define atomic_min	(		p,
			val
	)		__CL_ATOMIC_EXTREMUM(p, val, __cl_v < __cl_o)

atomic_or

#define atomic_or	(		p,
			v
	)		__atomic_fetch_or(p, v, __ATOMIC_SEQ_CST)

atomic_sub

#define atomic_sub	(		p,
			v
	)		__atomic_fetch_sub(p, v, __ATOMIC_SEQ_CST)

atomic_xchg

#define atomic_xchg	(		p,
			val
	)

Value:

    _Generic(*(p),                                                             \
        float: __CL_XCHG_F32(p, val),                                          \
        default: __atomic_exchange_n(p, val, __ATOMIC_SEQ_CST))

atomic_xor

#define atomic_xor	(		p,
			v
	)		__atomic_fetch_xor(p, v, __ATOMIC_SEQ_CST)

barrier

#define barrier

(

flags

)

cccc_barrier(flags)

CL_VERSION_1_0

#define CL_VERSION_1_0   100

CL_VERSION_1_1

#define CL_VERSION_1_1   110

CL_VERSION_1_2

#define CL_VERSION_1_2   120

CLK_GLOBAL_MEM_FENCE

#define CLK_GLOBAL_MEM_FENCE   CCCC_GLOBAL_FENCE

CLK_LOCAL_MEM_FENCE

#define CLK_LOCAL_MEM_FENCE   CCCC_LOCAL_FENCE

constant

#define constant   [[cccc::__constant__]]

get_global_id

#define get_global_id

(

d

)

cccc_global_id(d)

get_global_offset

#define get_global_offset

(

d

)

((size_t)0)

get_global_size

#define get_global_size

(

d

)

cccc_global_size(d)

get_group_id

#define get_group_id

(

d

)

cccc_group_id(d)

get_local_id

#define get_local_id

(

d

)

cccc_local_id(d)

get_local_size

#define get_local_size

(

d

)

cccc_local_size(d)

get_num_groups

#define get_num_groups

(

d

)

cccc_num_groups(d)

get_work_dim

#define get_work_dim

(

)

cccc_work_dim()

global

#define global   [[cccc::__global__]]

kernel

#define kernel   [[cccc::__kernel__]]

local

#define local   [[cccc::__local__]]

mem_fence

#define mem_fence

(

flags

)

__atomic_thread_fence(__ATOMIC_SEQ_CST)

private

#define private   [[cccc::__private__]]

read_mem_fence

#define read_mem_fence

(

flags

)

__atomic_thread_fence(__ATOMIC_SEQ_CST)

write_mem_fence

#define write_mem_fence

(

flags

)

__atomic_thread_fence(__ATOMIC_SEQ_CST)

Typedef Documentation

__cl_generic_uint

typedef unsigned __cl_generic_uint

uchar

typedef unsigned char uchar

uint

typedef unsigned int uint

ulong

typedef unsigned long ulong

ushort

typedef unsigned short ushort

Function Documentation

__CL_VECTOR() [1/21]

__CL_VECTOR	(	float	,
		float16	,
		64
	)

__CL_VECTOR() [2/21]

__CL_VECTOR	(	float	,
		float4	,
		16
	)

__CL_VECTOR() [3/21]

__CL_VECTOR	(	float	,
		float8	,
		32
	)

__CL_VECTOR() [4/21]

__CL_VECTOR	(	int	,
		int16	,
		64
	)

__CL_VECTOR() [5/21]

__CL_VECTOR	(	int	,
		int4	,
		16
	)

__CL_VECTOR() [6/21]

__CL_VECTOR	(	int	,
		int8	,
		32
	)

__CL_VECTOR() [7/21]

__CL_VECTOR	(	long	,
		long2	,
		16
	)

__CL_VECTOR() [8/21]

__CL_VECTOR	(	long	,
		long4	,
		32
	)

__CL_VECTOR() [9/21]

__CL_VECTOR	(	long	,
		long8	,
		64
	)

__CL_VECTOR() [10/21]

__CL_VECTOR	(	short	,
		short16	,
		32
	)

__CL_VECTOR() [11/21]

__CL_VECTOR	(	short	,
		short8	,
		16
	)

__CL_VECTOR() [12/21]

__CL_VECTOR	(	signed char	,
		char16	,
		16
	)

__CL_VECTOR() [13/21]

__CL_VECTOR	(	unsigned char	,
		uchar16	,
		16
	)

__CL_VECTOR() [14/21]

__CL_VECTOR	(	unsigned int	,
		uint16	,
		64
	)

__CL_VECTOR() [15/21]

__CL_VECTOR	(	unsigned int	,
		uint4	,
		16
	)

__CL_VECTOR() [16/21]

__CL_VECTOR	(	unsigned int	,
		uint8	,
		32
	)

__CL_VECTOR() [17/21]

__CL_VECTOR	(	unsigned long	,
		ulong2	,
		16
	)

__CL_VECTOR() [18/21]

__CL_VECTOR	(	unsigned long	,
		ulong4	,
		32
	)

__CL_VECTOR() [19/21]

__CL_VECTOR	(	unsigned long	,
		ulong8	,
		64
	)

__CL_VECTOR() [20/21]

__CL_VECTOR	(	unsigned short	,
		ushort16	,
		32
	)

__CL_VECTOR() [21/21]

__CL_VECTOR	(	unsigned short	,
		ushort8	,
		16
	)

__CL_VECTOR3() [1/5]

__CL_VECTOR3	(	float	,
		float3	,
		16
	)

__CL_VECTOR3() [2/5]

__CL_VECTOR3	(	int	,
		int3	,
		16
	)

__CL_VECTOR3() [3/5]

__CL_VECTOR3	(	long	,
		long3	,
		32
	)

__CL_VECTOR3() [4/5]

__CL_VECTOR3	(	unsigned int	,
		uint3	,
		16
	)

__CL_VECTOR3() [5/5]

__CL_VECTOR3	(	unsigned long	,
		ulong3	,
		32
	)