Implementation of vec::convert. More...

#include <sycl/detail/generic_type_traits.hpp>
#include <sycl/exception.hpp>
#include <cfenv>
#include <type_traits>

Include dependency graph for vector_convert.hpp:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Namespaces
	sycl

	sycl::_V1

	sycl::_V1::detail

Typedefs
template<typename T , typename R >
using	sycl::_V1::detail::is_sint_to_sint = std::bool_constant< is_sigeninteger_v< T > &&is_sigeninteger_v< R > >

template<typename T , typename R >
using	sycl::_V1::detail::is_uint_to_uint = std::bool_constant< is_sugeninteger_v< T > &&is_sugeninteger_v< R > >

template<typename T , typename R >
using	sycl::_V1::detail::is_sint_to_from_uint = std::bool_constant<(detail::is_sigeninteger_v< T > &&detail::is_sugeninteger_v< R >)\|\|(detail::is_sugeninteger_v< T > &&detail::is_sigeninteger_v< R >)>

template<typename T , typename R >
using	sycl::_V1::detail::is_sint_to_float = std::bool_constant< std::is_integral_v< T > &&!std::is_unsigned_v< T > &&detail::is_floating_point< R >::value >

template<typename T , typename R >
using	sycl::_V1::detail::is_uint_to_float = std::bool_constant< std::is_unsigned_v< T > &&detail::is_floating_point< R >::value >

template<typename T , typename R >
using	sycl::_V1::detail::is_int_to_float = std::bool_constant< std::is_integral_v< T > &&detail::is_floating_point< R >::value >

template<typename T , typename R >
using	sycl::_V1::detail::is_float_to_uint = std::bool_constant< detail::is_floating_point< T >::value &&std::is_unsigned_v< R > >

template<typename T , typename R >
using	sycl::_V1::detail::is_float_to_sint = std::bool_constant< detail::is_floating_point< T >::value &&std::is_integral_v< R > &&!std::is_unsigned_v< R > >

template<typename T , typename R >
using	sycl::_V1::detail::is_float_to_float = std::bool_constant< detail::is_floating_point< T >::value &&detail::is_floating_point< R >::value >

Enumerations
enum class	sycl::rounding_mode { sycl::automatic = 0 , sycl::rte = 1 , sycl::rtz = 2 , sycl::rtp = 3 , sycl::rtn = 4 }

Functions
float	sycl::_V1::ceil (float)

double	sycl::_V1::ceil (double)

float	sycl::_V1::floor (float)

double	sycl::_V1::floor (double)

float	sycl::_V1::rint (float)

double	sycl::_V1::rint (double)

float	sycl::_V1::trunc (float)

double	sycl::_V1::trunc (double)

template<typename From , typename To , int VecSize, typename Enable = std::enable_if_t<VecSize == 1>>
To	sycl::_V1::detail::SConvert (From Value)

template<typename From , typename To , int VecSize, typename Enable = std::enable_if_t<VecSize == 1>>
To	sycl::_V1::detail::UConvert (From Value)

template<typename From , typename To , int VecSize, typename Enable = std::enable_if_t<VecSize == 1>>
To	sycl::_V1::detail::ConvertSToF (From Value)

template<typename From , typename To , int VecSize, typename Enable = std::enable_if_t<VecSize == 1>>
To	sycl::_V1::detail::ConvertUToF (From Value)

template<typename From , typename To , int VecSize, typename Enable = std::enable_if_t<VecSize == 1>, sycl::rounding_mode RM>
To	sycl::_V1::detail::FConvert (From Value)

template<typename From , typename To , int VecSize, typename Enable = std::enable_if_t<VecSize == 1>, sycl::rounding_mode roundingMode>
To	sycl::_V1::detail::ConvertFToS (From Value)

template<typename From , typename To , int VecSize, typename Enable = std::enable_if_t<VecSize == 1>, sycl::rounding_mode roundingMode>
To	sycl::_V1::detail::ConvertFToU (From Value)

template<typename FromT , typename ToT , sycl::rounding_mode RoundingMode, int VecSize, typename NativeFromT , typename NativeToT >
NativeToT	sycl::_V1::detail::convertImpl (NativeFromT Value)
	Entry point helper for all kinds of converts between scalars and vectors, it dispatches to a right function depending on source and destination types. More...

Detailed Description

Implementation of vec::convert.

Implementation consists of a bunch of helper functions to do different conversions, as well as a single entry point function, which performs a dispatch into the right helper.

Helpers are made to match corresponding SPIR-V instructions defined in section 3.42.11. Conversion Instructions (SPIR-V specification version 1.6, revision 2, unified):

OpConvertFToU - float to unsigned integer
OpConvertFToS - float to signed integer
OpConvertSToF - signed integer to float
OpConvertUToF - unsigned integer to float
OpUConvert - unsigned integer to unsigned integer
OpSConvert - signed integer to signed integer
OpConvertF - float to float
OpSatConvertSToU - signed integer to unsigned integer
OpSatConvertUToS - unsigned integer to signed integer

To get the right SPIR-V instruction emitted from SYCL code, we need to make a call to a specific built-in in the following format: __spirv_[Op]_R[DestType][N?][_RoundingMode?] where:

[Op] is the name of instruction from the list above, without "Op" prefix.
[DestType] is the name of scalar return type
[N?] is vector size; omitted for scalars
[RoundingMode?] is rounding mode suffix, can be omitted

Implementation below is essentially split into two parts: for host and for device.

Host part is really simple, as we only have scalar conversions available in there. Most of them are implemented as regular static_cast with exception for float to integer conversions, which require rounding mode handling.

Device part is more complicated, because we need to generate calls to those __spirv* built-ins. To do so, macro code generation is used: we emit number of function templates, which are conditionally enabled using SFINAE for each destination type and vector size to each call their own __spriv* built-in.

Finally, there is single entry point which performs a dispatch to the right helper depending on source and destination types.

Definition in file vector_convert.hpp.

Namespaces

Typedefs

Enumerations

Functions

Detailed Description