Functions

◆ dev_zero()

template<typename D>

D batchlas::geqrf_native::dev_zero ( )

inline

Definition at line 27 of file geqrf_cta_device.hh.

◆ dev_add()

template<typename D>

D batchlas::geqrf_native::dev_add	(	D	a,
		D	b )

inline

Definition at line 36 of file geqrf_cta_device.hh.

◆ dev_imag()

template<typename D>

real_of< D > batchlas::geqrf_native::dev_imag ( D a )

inline

Definition at line 45 of file geqrf_cta_device.hh.

◆ dev_absmax()

template<typename D>

real_of< D > batchlas::geqrf_native::dev_absmax ( D a )

inline

Definition at line 56 of file geqrf_cta_device.hh.

◆ dev_abs2_scaled()

template<typename D>

real_of< D > batchlas::geqrf_native::dev_abs2_scaled	(	D	a,
		real_of< D >	s )

inline

Definition at line 66 of file geqrf_cta_device.hh.

◆ geqrf_larfg_scalars()

template<typename D>

LarfgScalars< D > batchlas::geqrf_native::geqrf_larfg_scalars	(	D	alpha,
		real_of< D >	s,
		real_of< D >	ssq )

inline

Definition at line 92 of file geqrf_cta_device.hh.

◆ geqrf_sg_sum()

template<typename D>

D batchlas::geqrf_native::geqrf_sg_sum	(	const sycl::sub_group &	sg,
		D	v )

inline

Definition at line 162 of file geqrf_cta_device.hh.

◆ geqrf_sg_max()

template<typename R>

R batchlas::geqrf_native::geqrf_sg_max	(	const sycl::sub_group &	sg,
		R	v )

inline

Definition at line 184 of file geqrf_cta_device.hh.

◆ geqr2_panel_device()

template<typename D, GeqrfScope SC, typename Tile>

void batchlas::geqrf_native::geqr2_panel_device	(	sycl::nd_item< 1 >	it,
		Tile	A,
		int	m,
		int	n,
		int	kmax,
		D *	tau_ptr )

inline

Definition at line 197 of file geqrf_cta_device.hh.

◆ geqrf_panel_reg_row_is_current()

template<typename D>

bool batchlas::geqrf_native::geqrf_panel_reg_row_is_current ( )

constexpr

Definition at line 63 of file geqrf_panel_reg_device.hh.

◆ geqrf_panel_reg_words()

int batchlas::geqrf_native::geqrf_panel_reg_words ( std::size_t scalar_bytes )

constexpr

Definition at line 68 of file geqrf_panel_reg_device.hh.

◆ geqrf_panel_reg_regs()

template<typename D>

int batchlas::geqrf_native::geqrf_panel_reg_regs ( )

constexpr

Definition at line 74 of file geqrf_panel_reg_device.hh.

◆ geqrf_panel_reg_max_rows()

template<typename D>

int batchlas::geqrf_native::geqrf_panel_reg_max_rows ( )

constexpr

Definition at line 91 of file geqrf_panel_reg_device.hh.

◆ geqrf_panel_reg_wg()

int batchlas::geqrf_native::geqrf_panel_reg_wg ( int m )

constexpr

Definition at line 109 of file geqrf_panel_reg_device.hh.

◆ geqrf_panel_reg_real_slots()

std::size_t batchlas::geqrf_native::geqrf_panel_reg_real_slots ( int wg )

constexpr

Definition at line 116 of file geqrf_panel_reg_device.hh.

◆ geqrf_panel_reg_scalar_slots()

std::size_t batchlas::geqrf_native::geqrf_panel_reg_scalar_slots	(	int	wg,
		int	cols )

constexpr

Definition at line 119 of file geqrf_panel_reg_device.hh.

◆ geqr2_panel_reg_device()

template<typename D, int N, typename Real>

void batchlas::geqrf_native::geqr2_panel_reg_device	(	sycl::nd_item< 1 >	it,
		D *	a,
		int	ld,
		int	m,
		int	n,
		int	kmax,
		D *	tau_ptr,
		Real *	sred,
		D *	sw )

inline

Definition at line 129 of file geqrf_panel_reg_device.hh.

◆ geqrf_tiny_reduce_fmax()

template<int N, typename R, typename Part>

R batchlas::geqrf_native::geqrf_tiny_reduce_fmax	(	const Part &	part,
		R	value )

inline

Definition at line 23 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_reduce_sum()

template<int N, typename R, typename Part>

R batchlas::geqrf_native::geqrf_tiny_reduce_sum	(	const Part &	part,
		R	value )

inline

Definition at line 32 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_reg_slot()

int batchlas::geqrf_native::geqrf_tiny_reg_slot ( int n_pad )

constexpr

Definition at line 45 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_words()

template<typename D>

int batchlas::geqrf_native::geqrf_tiny_words ( )

constexpr

Definition at line 82 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_regs()

int batchlas::geqrf_native::geqrf_tiny_regs	(	int	n_pad,
		int	words )

constexpr

Definition at line 89 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_regs_for()

template<typename D, int N>

int batchlas::geqrf_native::geqrf_tiny_regs_for ( )

constexpr

Definition at line 94 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_ceil8()

int batchlas::geqrf_native::geqrf_tiny_ceil8 ( int regs )

constexpr

Definition at line 109 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_blocks_by_regs()

int batchlas::geqrf_native::geqrf_tiny_blocks_by_regs	(	int	regs,
		int	wg )

constexpr

Definition at line 111 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_slm_elems()

std::size_t batchlas::geqrf_native::geqrf_tiny_slm_elems	(	int	n_pad,
		int	c )

constexpr

Definition at line 122 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_blocks_by_slm()

int batchlas::geqrf_native::geqrf_tiny_blocks_by_slm	(	int	n_pad,
		int	c,
		std::size_t	elem_bytes,
		int	matrices_per_wg )

constexpr

Definition at line 130 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_matrices_per_wg()

template<int N>

int batchlas::geqrf_native::geqrf_tiny_matrices_per_wg ( )

constexpr

Definition at line 145 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_chunk()

template<typename D, int N>

int batchlas::geqrf_native::geqrf_tiny_chunk ( )

constexpr

Definition at line 161 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_chunk_is_fixed_point()

template<typename D, int N>

bool batchlas::geqrf_native::geqrf_tiny_chunk_is_fixed_point ( )

constexpr

Definition at line 175 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_cell_is_resident()

template<typename D, int N>

bool batchlas::geqrf_native::geqrf_tiny_cell_is_resident ( )

constexpr

Definition at line 184 of file geqrf_tiny_device.hh.

◆ geqrf_tiny_slda()

int batchlas::geqrf_native::geqrf_tiny_slda ( int c )

constexpr

Definition at line 192 of file geqrf_tiny_device.hh.

Types

◆ real_of

template<typename D>

using batchlas::geqrf_native::real_of = typename RealOf<D>::type

Definition at line 24 of file geqrf_cta_device.hh.

◆ GeqrfRawTile

template<typename D>

using batchlas::geqrf_native::GeqrfRawTile = GeqrfGlobalTile<D>

Definition at line 149 of file geqrf_cta_device.hh.

Enumerations

◆ GeqrfScope

enum class batchlas::geqrf_native::GeqrfScope

strong

Enumerator
SubGroup
WorkGroup

Definition at line 157 of file geqrf_cta_device.hh.

Variables

◆ kGeqrfPanelRegOverhead

int batchlas::geqrf_native::kGeqrfPanelRegOverhead = 96

inlineconstexpr

Definition at line 58 of file geqrf_panel_reg_device.hh.

◆ kGeqrfPanelRegMargin

int batchlas::geqrf_native::kGeqrfPanelRegMargin = 16

inlineconstexpr

Definition at line 59 of file geqrf_panel_reg_device.hh.

◆ kGeqrfPanelRegMaxWg

int batchlas::geqrf_native::kGeqrfPanelRegMaxWg = 1024

inlineconstexpr

Definition at line 84 of file geqrf_panel_reg_device.hh.

◆ kGeqrfPanelRegPolicyRows

int batchlas::geqrf_native::kGeqrfPanelRegPolicyRows = 128

inlineconstexpr

Definition at line 107 of file geqrf_panel_reg_device.hh.

◆ kGeqrfTinyRegMargin

int batchlas::geqrf_native::kGeqrfTinyRegMargin = 8

inlineconstexpr

Definition at line 43 of file geqrf_tiny_device.hh.

◆ kGeqrfTinyRegOverhead

int batchlas::geqrf_native::kGeqrfTinyRegOverhead = 88

inlineconstexpr

Definition at line 88 of file geqrf_tiny_device.hh.

◆ kGeqrfTinyRegsPerSm

int batchlas::geqrf_native::kGeqrfTinyRegsPerSm = 65536

inlineconstexpr

Definition at line 105 of file geqrf_tiny_device.hh.

◆ kGeqrfTinyThreadsPerSm

int batchlas::geqrf_native::kGeqrfTinyThreadsPerSm = 1536

inlineconstexpr

Definition at line 106 of file geqrf_tiny_device.hh.

◆ kGeqrfTinyMaxBlocksPerSm

int batchlas::geqrf_native::kGeqrfTinyMaxBlocksPerSm = 24

inlineconstexpr

Definition at line 107 of file geqrf_tiny_device.hh.

◆ kGeqrfTinyReferenceSlm

std::size_t batchlas::geqrf_native::kGeqrfTinyReferenceSlm

inlineconstexpr

Initial value:

=

97280

Definition at line 127 of file geqrf_tiny_device.hh.

◆ kGeqrfTinyWg

int batchlas::geqrf_native::kGeqrfTinyWg = tiny_native::kTinyWgSize

inlineconstexpr

Definition at line 140 of file geqrf_tiny_device.hh.

Classes
struct	GeqrfGlobalTile
struct	GeqrfPanelRegPlan
struct	GeqrfPanelRegPlan< double >
struct	GeqrfPanelRegPlan< float >
struct	GeqrfPanelRegPlan< sycl_device::Cx< double > >
struct	GeqrfPanelRegPlan< sycl_device::Cx< float > >
struct	GeqrfTinyRegs
struct	GeqrfTinyRegs< double >
struct	GeqrfTinyRegs< float >
struct	GeqrfTinyRegs< sycl_device::Cx< double > >
struct	GeqrfTinyRegs< sycl_device::Cx< float > >
struct	LarfgScalars
struct	RealOf
struct	RealOf< Cx< R > >

Functions

◆ dev_zero()

◆ dev_add()

◆ dev_imag()

◆ dev_absmax()

◆ dev_abs2_scaled()

◆ geqrf_larfg_scalars()

◆ geqrf_sg_sum()

◆ geqrf_sg_max()

◆ geqr2_panel_device()

◆ geqrf_panel_reg_row_is_current()

◆ geqrf_panel_reg_words()

◆ geqrf_panel_reg_regs()

◆ geqrf_panel_reg_max_rows()

◆ geqrf_panel_reg_wg()

◆ geqrf_panel_reg_real_slots()

◆ geqrf_panel_reg_scalar_slots()

◆ geqr2_panel_reg_device()

◆ geqrf_tiny_reduce_fmax()

◆ geqrf_tiny_reduce_sum()

◆ geqrf_tiny_reg_slot()

◆ geqrf_tiny_words()

◆ geqrf_tiny_regs()

◆ geqrf_tiny_regs_for()

◆ geqrf_tiny_ceil8()

◆ geqrf_tiny_blocks_by_regs()

◆ geqrf_tiny_slm_elems()

◆ geqrf_tiny_blocks_by_slm()

◆ geqrf_tiny_matrices_per_wg()

◆ geqrf_tiny_chunk()

◆ geqrf_tiny_chunk_is_fixed_point()

◆ geqrf_tiny_cell_is_resident()

◆ geqrf_tiny_slda()

Types

◆ real_of

◆ GeqrfRawTile

Enumerations

◆ GeqrfScope

Variables

◆ kGeqrfPanelRegOverhead

◆ kGeqrfPanelRegMargin

◆ kGeqrfPanelRegMaxWg

◆ kGeqrfPanelRegPolicyRows

◆ kGeqrfTinyRegMargin

◆ kGeqrfTinyRegOverhead

◆ kGeqrfTinyRegsPerSm

◆ kGeqrfTinyThreadsPerSm

◆ kGeqrfTinyMaxBlocksPerSm

◆ kGeqrfTinyReferenceSlm

◆ kGeqrfTinyWg

Classes