Imports

Types

CPUFeatureX86 = enum x86_Generic, x86_SSE, x86_SSE2, x86_SSE4_1, x86_AVX, x86_AVX_FMA, x86_AVX2, x86_AVX512: Source Edit
MicroKernel = object mr*, nr*: int cpu_simd*: CPUFeatureX86 nb_scalars*: int nb_vecs_nr*: int c_unit_stride*: bool pt*: int: Source Edit
Tiles[T] = ref object a*: ptr UncheckedArray[T] b*: ptr UncheckedArray[T] mc*, nc*, kc*: int ic_num_tasks*: int upanelA_size*: int: Source Edit

func get_num_tiles(dim_size, tile_size: int): int {.inline, ...raises: [], tags: [], forbids: [].}: Get the number of tiles along a dimension depending on the tile size Source Edit
proc newTiles(ukernel: static MicroKernel; T: typedesc; M, N, K: Natural): Tiles[ T]: Source Edit
func partitionMNK(ukernel: static MicroKernel; T: typedesc; M, N, K: Natural): tuple[ mc, nc, kc: int]: Source Edit
func x86_ukernel(cpu: CPUFeatureX86; T: typedesc; c_unit_stride: bool): MicroKernel: Source Edit

macro extract_c_unit_stride(ukernel: static MicroKernel): untyped: Source Edit
macro extract_cpu_simd(ukernel: static MicroKernel): untyped: Source Edit
macro extract_mr(ukernel: static MicroKernel): untyped: Source Edit
macro extract_nb_scalars(ukernel: static MicroKernel): untyped: Source Edit
macro extract_nb_vecs_nr(ukernel: static MicroKernel): untyped: Source Edit
macro extract_nr(ukernel: static MicroKernel): untyped: Source Edit
macro extract_pt(ukernel: static MicroKernel): untyped: Source Edit