BatchLAS Documentation BatchLASv0.1.0
batchlas::sycl_getrf Namespace Reference

Functions

◆ getrf_cta_max_n_for_slm()

template<typename T>
BATCHLAS_INTERNAL_API int batchlas::sycl_getrf::getrf_cta_max_n_for_slm ( std::size_t slm_budget_bytes,
int min_blocks_per_sm = resident::kMinBlocksPerSm )

◆ getrf_blocked_available()

template<typename T>
BATCHLAS_INTERNAL_API bool batchlas::sycl_getrf::getrf_blocked_available ( )

◆ getrf_tiny_max_n()

template<typename T>
BATCHLAS_INTERNAL_API int batchlas::sycl_getrf::getrf_tiny_max_n ( )

◆ getrf_tiny_buffer_size()

template<typename T>
BATCHLAS_INTERNAL_API std::size_t batchlas::sycl_getrf::getrf_tiny_buffer_size ( Queue & ctx,
const MatrixView< T, MatrixFormat::Dense > & A )

◆ getrf_tiny_dispatch()

template<typename T>
BATCHLAS_INTERNAL_API Event batchlas::sycl_getrf::getrf_tiny_dispatch ( Queue & ctx,
const MatrixView< T, MatrixFormat::Dense > & A,
Span< int64_t > pivots,
Span< std::byte > workspace,
Span< int32_t > info )

◆ getrf_cta_buffer_size()

template<typename T>
BATCHLAS_INTERNAL_API std::size_t batchlas::sycl_getrf::getrf_cta_buffer_size ( Queue & ctx,
const MatrixView< T, MatrixFormat::Dense > & A )

◆ getrf_blocked_buffer_size()

template<typename T>
BATCHLAS_INTERNAL_API std::size_t batchlas::sycl_getrf::getrf_blocked_buffer_size ( Queue & ctx,
const MatrixView< T, MatrixFormat::Dense > & A )

◆ getrf_blocked_debug_params()

template<typename T>
BATCHLAS_INTERNAL_API unsigned batchlas::sycl_getrf::getrf_blocked_debug_params ( Queue & ctx,
int n )

◆ getrf_blocked_debug_leaf()

template<typename T>
BATCHLAS_INTERNAL_API unsigned batchlas::sycl_getrf::getrf_blocked_debug_leaf ( Queue & ctx,
int n )

◆ getrf_cta_dispatch()

template<typename T>
BATCHLAS_INTERNAL_API Event batchlas::sycl_getrf::getrf_cta_dispatch ( Queue & ctx,
const MatrixView< T, MatrixFormat::Dense > & A,
Span< int64_t > pivots,
Span< std::byte > workspace,
Span< int32_t > info )

◆ getrf_blocked_dispatch()

template<typename T>
BATCHLAS_INTERNAL_API Event batchlas::sycl_getrf::getrf_blocked_dispatch ( Queue & ctx,
const MatrixView< T, MatrixFormat::Dense > & A,
Span< int64_t > pivots,
Span< std::byte > workspace,
Span< int32_t > info,
GetrfTrailingGemm< T > trailing_gemm,
GetrfPanelSolveTrsm< T > panel_trsm )

◆ getrf_cta_fits()

template<typename T>
BATCHLAS_INTERNAL_API bool batchlas::sycl_getrf::getrf_cta_fits ( int n,
std::size_t slm_budget_bytes,
int min_blocks_per_sm = resident::kMinBlocksPerSm )

◆ getrf_cta_debug_launch()

template<typename T>
BATCHLAS_INTERNAL_API unsigned batchlas::sycl_getrf::getrf_cta_debug_launch ( Queue & ctx,
int m,
int n )

◆ getrf_leaf_fits()

template<typename T>
BATCHLAS_INTERNAL_API bool batchlas::sycl_getrf::getrf_leaf_fits ( int m,
int n,
std::size_t slm_budget_bytes )

◆ getrf_panel_reg_nb()

template<typename T>
BATCHLAS_INTERNAL_API int batchlas::sycl_getrf::getrf_panel_reg_nb ( )

◆ getrf_panel_reg_max_m()

template<typename T>
BATCHLAS_INTERNAL_API int batchlas::sycl_getrf::getrf_panel_reg_max_m ( )

◆ getrf_panel_reg_fits()

template<typename T>
BATCHLAS_INTERNAL_API bool batchlas::sycl_getrf::getrf_panel_reg_fits ( int m,
int n,
int max_wg )

◆ getrf_panel_reg_debug_launch()

template<typename T>
BATCHLAS_INTERNAL_API unsigned batchlas::sycl_getrf::getrf_panel_reg_debug_launch ( Queue & ctx,
int m,
int n )

◆ getrf_panel_reg_factorize()

template<typename T>
BATCHLAS_INTERNAL_API Event batchlas::sycl_getrf::getrf_panel_reg_factorize ( Queue & ctx,
T * a_ptr,
int ld,
int stride,
int m,
int n,
int batch,
int * piv_ptr,
int piv_stride,
int piv_base,
int32_t * info_ptr )

◆ getrf_panel_factorize()

template<typename T>
BATCHLAS_INTERNAL_API Event batchlas::sycl_getrf::getrf_panel_factorize ( Queue & ctx,
T * a_ptr,
int ld,
int stride,
int m,
int n,
int batch,
int * piv_ptr,
int piv_stride,
int piv_base,
int32_t * info_ptr,
bool * used_resident_out )

Types

◆ GetrfTrailingGemm

template<typename T>
using batchlas::sycl_getrf::GetrfTrailingGemm
Initial value:
std::function<Event(
Queue&,
const MatrixView<T, MatrixFormat::Dense>&,
const MatrixView<T, MatrixFormat::Dense>&,
const MatrixView<T, MatrixFormat::Dense>&,
T, T, Transpose, Transpose, ComputePrecision)>
batchlas::MatrixView
Non-owning view of a batch of matrices: the argument type of every entry point.
Definition matrix.hh:485
batchlas::Transpose
Transpose
The operator op() applied to an operand: op(A) = A, A^T or A^H.
Definition enums.hh:171
batchlas::ComputePrecision
ComputePrecision
Internal compute precision of a GEMM (the cuBLAS cublasComputeType_t).
Definition enums.hh:505
batchlas::Event::Event
Event()
An empty Event, already complete.
batchlas::Queue
Where BatchLAS work runs: a device, a backend, an ordered SYCL queue and a workspace arena.
Definition sycl-device-queue.hh:321

Definition at line 70 of file getrf_native.hh.

◆ GetrfPanelSolveTrsm

template<typename T>
using batchlas::sycl_getrf::GetrfPanelSolveTrsm
Initial value:
std::function<Event(
Queue&,
const MatrixView<T, MatrixFormat::Dense>&,
const MatrixView<T, MatrixFormat::Dense>&,
T, Side, Uplo, Transpose, Diag)>
batchlas::Diag
Diag
Whether a triangular operand's diagonal is read or assumed to be one.
Definition enums.hh:303
batchlas::Uplo
Uplo
Which triangle of a symmetric, Hermitian or triangular operand is referenced.
Definition enums.hh:287
batchlas::Side
Side
Side on which a special operand multiplies: op(A) * B or B * op(A).
Definition enums.hh:319

Definition at line 79 of file getrf_native.hh.

Variables

◆ kGetrfTinyWgSize

int batchlas::sycl_getrf::kGetrfTinyWgSize = 64
inlineconstexpr

Definition at line 37 of file getrf_native.hh.

Generated by doxygen 1.18.0