batchlas::ops::getrf Namespace Reference

Functions

candidates()

template<class T>

auto batchlas::ops::getrf::candidates ( )

constexpr

Every compiled choice, once, in tie-break order (§6.3).

Definition at line 22 of file choice.hh.

Types

GetrfChoice

using batchlas::ops::getrf::GetrfChoice = std::variant<Tiny, Cta, Blocked, Vendor>

natives: GPU, sub-group 32, square, not NETLIB

Definition at line 19 of file choice.hh.

Variables

spec

select::OpSpec batchlas::ops::getrf::spec {Op::getrf, select::Lib::factorization}

inlineconstexpr

§5.5: blocked, then vendor (CPU)

Definition at line 24 of file choice.hh.

key_names

std::array<std::string_view, 2> batchlas::ops::getrf::key_names {"n:log:3", "batch:log"}

inlineconstexpr

work ~ n^3 batch

Definition at line 26 of file choice.hh.

grid_n

std::array<int, 28> batchlas::ops::getrf::grid_n

inlineconstexpr

Initial value:

{1,  2,  3,  4,   5,   6,   7,   8,   9,   12,  16,  17,  24,  25,
                                            32, 33, 48, 64,  96,  128, 192, 255, 256, 384, 511, 512, 768, 1024}

Both sides of every transcribed threshold (4|5, 7|8|9, 16|17, 24|25, 32|33, 255|256, 511|512; batch 255|256), a coarse log grid elsewhere.

The transcriber (tuned/README.md) spells the same grid.

Definition at line 30 of file choice.hh.

grid_batch

std::array<int, 10> batchlas::ops::getrf::grid_batch {1, 8, 64, 128, 255, 256, 512, 2048, 8192, 32768}

inlineconstexpr

Definition at line 32 of file choice.hh.

Classes
struct	Tiny
	getrf_tiny_dispatch: register-resident, n <= 32 (cdouble 16) More...
struct	Cta
	getrf_cta_dispatch: SLM-resident panel, n <= the SLM ceiling More...
struct	Blocked
	getrf_blocked_dispatch + the public gemm and trsm More...
struct	Vendor
	backend::getrf_vendor (cuBLAS / rocSOLVER / LAPACKE) More...