 |
SyTen
|
|
|
SyTen
|
|
Support functions (memory allocation etc.) for CUDA-based GPUs. More...
Classes | |
| class | CudaAllocator |
| Cuda allocator class, uses the buddy system. More... | |
| class | CudaConstSpan |
| CudaConstSpan is a non-owning constant reference to a continuous array of objects in CUDA storage. More... | |
| class | CudaDynArray |
| The CUDA-equivalent of DynArray. More... | |
| class | CudaMutSpan |
| CudaMutSpan is a non-owning reference to a continuous array of objects in CUDA storage which can be mutated through the span. More... | |
| class | CudaPtr |
| Implementation for syten::CudaPtr. More... | |
| class | CudaStream |
| Implementation for syten::CudaStream. More... | |
| struct | Handle |
| Represents a cublasHandle_t object which may or may not be in use. More... | |
| struct | HandleBox |
| RAII wrapper around a handle, putting it out of use on destruction. More... | |
Functions | |
| void | addScaled (CudaDynArray< double > &a, CudaDynArray< double > const &b, double const factor) |
Adds b scaled by factor to a. More... | |
| void | addScaled (CudaDynArray< float > &a, CudaDynArray< float > const &b, float const factor) |
Adds b scaled by factor to a. More... | |
| void | addScaled (CudaDynArray< std::complex< double > > &a, CudaDynArray< std::complex< double > > const &b, std::complex< double > const factor) |
Adds b scaled by factor to a. More... | |
| void | addScaled (CudaDynArray< std::complex< float > > &a, CudaDynArray< std::complex< float > > const &b, std::complex< float > const factor) |
Adds b scaled by factor to a. More... | |
| std::uint16_t | allocator_get_max_size () |
| Returns the log2 of the maximal block size of the CUDA allocator. More... | |
| std::uint16_t | allocator_get_min_size () |
| Returns the log2 of the minimal block size of the CUDA allocator. More... | |
| void | allocator_print_status () |
| Prints the status of the CUDA allocator. More... | |
| std::uint16_t | allocator_set_max_size (std::uint16_t sz) |
| Sets the log2 of the maximal block size of the CUDA allocator. More... | |
| std::uint16_t | allocator_set_min_size (std::uint16_t sz) |
| Sets the log2 of the minimal block size of the CUDA allocator. More... | |
| void | conj_copy (CudaConstSpan< double > src, CudaMutSpan< double > dst) |
Copies all data from src to dst (no conjugation takes places as double is real). More... | |
| void | conj_copy (CudaConstSpan< float > src, CudaMutSpan< float > dst) |
Copies all data from src to dst (no conjugation takes places as float is real). More... | |
| void | conj_copy (CudaConstSpan< std::complex< double > > src, CudaMutSpan< std::complex< double > > dst) |
Copies all data from src to dst while complex-conjugating every value. More... | |
| void | conj_copy (CudaConstSpan< std::complex< float > > src, CudaMutSpan< std::complex< float > > dst) |
Copies all data from src to dst while complex-conjugating every value. More... | |
| template<typename T > | |
| void | copy (CudaConstSpan< T > const inp, CudaMutSpan< T > out) |
Copies all data from inp to out. More... | |
| bool | cuda_compiled () |
| Returns true if CUDA support is compiled in. More... | |
| void | cuda_dot_conj_kernel_impl (std::size_t sz, const std::complex< double > *to_be_conj_a, const std::complex< double > *b, std::complex< double > *result, void *cuda_stream) |
| Calculates the scalar product of two CUDA arrays. More... | |
| bool | cuda_enabled () |
| Returns true if the list of allowed devices is not empty. More... | |
| cudaError_t | cuda_handle_error_impl (cudaError_t err, std::string str, SourceLocation location=SourceLocation::current()) |
Implementation for helper function to handle Cuda return values by throwing an assertion failure if the return value is not cudaSuccess. More... | |
| cudaError_t | cuda_handle_error_impl (cudaError_t err, Vec< cudaError_t > acceptable_errors, std::string str, SourceLocation location=SourceLocation::current()) |
Implementation for helper function to handle Cuda return values by throwing an assertion failure if the return value is neither cudaSuccess nor in the list of acceptable errors. More... | |
| double | dot (CudaDynArray< double > const &a, CudaDynArray< double > const &b, Conj const conj=Conj::n()) |
| Returns the scalar product between two CudaDynArray objects. More... | |
| float | dot (CudaDynArray< float > const &a, CudaDynArray< float > const &b, Conj const conj=Conj::n()) |
| Returns the scalar product between two CudaDynArray objects. More... | |
| std::complex< double > | dot (CudaDynArray< std::complex< double > > const &a, CudaDynArray< std::complex< double > > const &b, Conj const conj=Conj::n()) |
| Returns the scalar product between two CudaDynArray objects. More... | |
| std::complex< float > | dot (CudaDynArray< std::complex< float > > const &a, CudaDynArray< std::complex< float > > const &b, Conj const conj=Conj::n()) |
| Returns the scalar product between two CudaDynArray objects. More... | |
| HandleBox | get_handle (CudaStream const &str) |
Returns a cublas handle associated to the device device and switches the current CUDA device to this device. More... | |
| cublasStatus_t | handle_error (cublasStatus_t err) |
| Checks the error code and asserts if it is not CUBLAS_STATUS_SUCCESS. More... | |
| template<typename T > | |
| int | host_device (CudaConstSpan< T > const inp) |
| Returns the device hosting the CudaConstSpan range. More... | |
| template<typename T > | |
| int | host_device (CudaMutSpan< T > const inp) |
| Returns the device hosting the CudaMutSpan range. More... | |
| void | initialise_handles () |
| Initialises the cublas handles for the currently allowed devices. More... | |
| template<typename T > | |
| T | max_element (CudaConstSpan< T > const inp) |
| Returns the absolute-value maximal element of the range. More... | |
| std::size_t | max_element_idx (CudaConstSpan< double > inp) |
| Returns the 0-based index of the absolute-value maximal element of the range. More... | |
| std::size_t | max_element_idx (CudaConstSpan< float > inp) |
| Returns the 0-based index of the absolute-value maximal element of the range. More... | |
| std::size_t | max_element_idx (CudaConstSpan< std::complex< double > > inp) |
| Returns the 0-based index of the absolute-value maximal element of the range. More... | |
| std::size_t | max_element_idx (CudaConstSpan< std::complex< float > > inp) |
| Returns the 0-based index of the absolute-value maximal element of the range. More... | |
| void | scale (CudaDynArray< double > &a, double const f) |
Scales the entries of a by f. More... | |
| void | scale (CudaDynArray< float > &a, float const f) |
Scales the entries of a by f. More... | |
| void | scale (CudaDynArray< std::complex< double > > &a, double const f) |
Scales the entries of a by f. More... | |
| void | scale (CudaDynArray< std::complex< double > > &a, std::complex< double > const &f) |
Scales the entries of a by f. More... | |
| void | scale (CudaDynArray< std::complex< float > > &a, float const f) |
Scales the entries of a by f. More... | |
| void | scale (CudaDynArray< std::complex< float > > &a, std::complex< float > const f) |
Scales the entries of a by f. More... | |
| void | set_allowed_devices ([[maybe_unused]] Vec< std::int16_t > const &devices) |
| Sets the allowed devices to the supplied list and enables inter-device memory access. More... | |
| std::string | status_description (cublasStatus_t err) |
| Returns a string describing the specified cuBLAS error. More... | |
| std::string | version () |
| Returns a version string describing the current CUDA version/compilation/enablement. More... | |
Allowed devices and allocation logic | |
| Vec< std::int16_t > const & | get_allowed_devices () |
| Returns a list of allowed devices. More... | |
| std::int16_t | get_alloc_device () |
| Returns the device ID of the next allocation device to use. More... | |
| void | setup () |
| Sets up CUDA to allow all existing devices and generates the associated cuBLAS handles for the calling thread. More... | |
| void | setup ([[maybe_unused]] Vec< std::int16_t > const &devices) |
| Sets up CUDA to allow the specified devices and generates the associated cuBLAS handles for the calling thread. More... | |
Memory management functions. | |
| CudaPtr< void > | alloc (std::size_t sz) |
Allocates sz bytes on the next CUDA allocation device. More... | |
| CudaPtr< void > | alloc_on_device (std::size_t sz,[[maybe_unused]] std::int16_t device) |
Allocates sz bytes on the specified CUDA allocation device. More... | |
| void | free (CudaPtr< void > ptr) |
Frees the allocation pointed to by ptr. More... | |
| template<typename T > | |
| void | free (CudaPtr< T > ptr) |
Frees the allocation pointed to by ptr. More... | |
| template<typename T > | |
| std::int16_t | host_device ([[maybe_unused]] CudaPtr< T > v) |
Returns the device hosting the allocation pointed to by v. More... | |
Copying objects and arrays. | |
| void | copy (CudaPtr< const char > src, CudaPtr< char > dst, std::size_t num) |
Copies num bytes from src to dst. More... | |
| void | copy (CudaPtr< const char > src, CudaPtr< char > dst, std::size_t num,[[maybe_unused]] CudaStream const &str) |
Copies num bytes from src to dst within stream str. More... | |
| template<typename T > | |
| void | copy (CudaPtr< const typename IdentityType< T >::type > src, CudaPtr< T > dst, std::size_t num) |
Copies num objects of type T from src to dst. More... | |
| template<typename T > | |
| void | copy (CudaPtr< const typename IdentityType< T >::type > src, T *dst, std::size_t num) |
Copies num objects of type T from src to dst. More... | |
| template<typename T > | |
| void | copy (const T *src, CudaPtr< T > dst, std::size_t num) |
Copies num objects of type T from src to dst. More... | |
| template<typename T > | |
| void | copy (const T *src, T *dst, std::size_t num) |
Copies num objects of type T from src to dst. More... | |
| template<typename T > | |
| void | copy (CudaPtr< const typename IdentityType< T >::type > src, CudaPtr< T > dst, std::size_t num, [[maybe_unused]] CudaStream const &str) |
Copies num objects of type T from src to dst within stream str. More... | |
| template<typename T > | |
| void | copy (CudaPtr< const typename IdentityType< T >::type > src, T *dst, std::size_t num, [[maybe_unused]] CudaStream const &str) |
Copies num objects of type T from src to dst within stream str. More... | |
| template<typename T > | |
| void | copy (const T *src, CudaPtr< T > dst, std::size_t num, [[maybe_unused]] CudaStream const &str) |
Copies num objects of type T from src to dst within stream str. More... | |
| template<typename T > | |
| void | copy (const T *src, T *dst, std::size_t num, [[maybe_unused]] CudaStream const &str) |
Copies num objects of type T from src to dst within stream str. More... | |
Device management and synchronisation. | |
| void | select_device ([[maybe_unused]] std::int16_t device) |
| Selects the specified device for the current thread. More... | |
| void | synchronise () |
| Synchronizes with the current device. More... | |
| Pair< Size, Size > | mem_status () |
| Returns a pair of free and total device memory. More... | |
Variables | |
| static CudaAllocator | allocator |
| Our CudaAllocator object, handling a buddy free list system. More... | |
| static Vec< std::int16_t > | allowed_devices = {} |
| Storage for the currently allowed devices. More... | |
| Vec< Pair< int, Vec< Handle > > > | handles |
| Preinitialised handles for each of the devices. More... | |
Support functions (memory allocation etc.) for CUDA-based GPUs.
1.9.1