cuda::rtc::compilation_output_t< cuda_cpp > Class Template Reference

Output of CUDA C++ code JIT-compilation. More...

#include <compilation_output.hpp>

Inheritance diagram for cuda::rtc::compilation_output_t< cuda_cpp >:

Collaboration diagram for cuda::rtc::compilation_output_t< cuda_cpp >:

Public Types
using	parent = compilation_output_base_t< cuda_cpp >
Public Types inherited from cuda::rtc::compilation_output_base_t< cuda_cpp >
using	handle_type = program::handle_t< source_kind >
using	status_type = status_t< source_kind >

Public Member Functions
span< char >	ptx (span< char > buffer) const
	Obtain a (nul-terminated) copy of the PTX result of the last compilation. More...
unique_span< char >	ptx () const
	Obtain a copy of the PTX resulting from the program compilation. More...
bool	has_ptx () const
const char *	get_mangling_of (const char *unmangled_name) const
	Obtain the mangled/lowered form of an expression registered earlier, after the compilation. More...
const char *	get_mangling_of (const ::std::string &unmangled_name) const
	Obtain the mangled/lowered form of an expression registered earlier, after the compilation. More...
Public Member Functions inherited from cuda::rtc::compilation_output_base_t< cuda_cpp >
bool	succeeded () const
bool	failed () const
	operator bool () const
const ::std::string &	program_name () const
handle_type	program_handle () const
span< char >	log (span< char > buffer) const
	Write a copy of the program compilation log into a user-provided buffer. More...
unique_span< char >	log () const
	Obtain a copy of the compilation log. More...
	compilation_output_base_t (compilation_output_base_t &&other) noexcept
compilation_output_base_t &	operator= (const compilation_output_base_t &other)=delete
compilation_output_base_t &	operator= (compilation_output_base_t &&other)=delete

Additional Inherited Members
Static Public Attributes inherited from cuda::rtc::compilation_output_base_t< cuda_cpp >
static constexpr const source_kind_t	source_kind

Detailed Description

template<>
class cuda::rtc::compilation_output_t< cuda_cpp >

Output of CUDA C++ code JIT-compilation.

Member Function Documentation

get_mangling_of() [1/2]

const char* cuda::rtc::compilation_output_t< cuda_cpp >::get_mangling_of ( const char *  unmangled_name ) const

inline

Obtain the mangled/lowered form of an expression registered earlier, after the compilation.

Parameters

unmangled_name

A name of a __global__ or __device__ function or variable.

Returns

The mangled name (which can actually be used for invoking kernels, moving data etc.). The memory is owned by the NVRTC program and will be released when it is destroyed.

get_mangling_of() [2/2]

const char* cuda::rtc::compilation_output_t< cuda_cpp >::get_mangling_of ( const ::std::string &  unmangled_name ) const

inline

Obtain the mangled/lowered form of an expression registered earlier, after the compilation.

Parameters

unmangled_name

A name of a __global__ or __device__ function or variable.

Returns

The mangled name (which can actually be used for invoking kernels, moving data etc.). The memory is owned by the NVRTC program and will be released when it is destroyed.

has_ptx()

bool cuda::rtc::compilation_output_t< cuda_cpp >::has_ptx ( ) const

inline

Returns

true if the program has been successfully compiled, with the result containing PTX code

ptx() [1/2]

span<char> cuda::rtc::compilation_output_t< cuda_cpp >::ptx ( span< char >  buffer ) const

inline

Obtain a (nul-terminated) copy of the PTX result of the last compilation.

Note

the PTX may be missing in cases such as compilation failure or link-time optimization compilation.

This will fail if the program has never been compiled. Write a copy of the PTX resulting from the compilation into a user-provided buffer

Parameters

[in,out]

buffer

A writable buffer large enough to contain the compiled program's PTX code.

Returns

The sub-buffer, starting at the beginning of buffer, containing exactly the compiled program's PTX (i.e. sized down to fit the contents)

Note

This will throw if the program has never been compiled, or if the buffer is not large enough to contain the compiled PTX code.

ptx() [2/2]

unique_span<char> cuda::rtc::compilation_output_t< cuda_cpp >::ptx ( ) const

inline

Obtain a copy of the PTX resulting from the program compilation.

Returns

an owning container with a nul-terminated copy of the PTX code

Note

This will fail if the program compilation has not produced any PTX

---

The documentation for this class was generated from the following file:

• src/cuda/rtc/compilation_output.hpp