Struct WmmaLoad 

pub struct WmmaLoad<D: Dialect> { /* private fields */ }

Implementations

impl<D: Dialect> WmmaLoad<D>

pub fn new(frag: Fragment<D>, layout: Option<FragmentLayout<D>>) -> Self

Constructs a new WmmaLoad.

impl<D: Dialect> WmmaLoad<D>

pub fn fn_name(&self) -> String

pub fn format_extension(&self, f: &mut Formatter<'_>) -> Result

Matrix A must be in column major layout (so fragments correspond to a row) Matrices B, C and D must be in row major layout (so fragments correspond to a column)

Each lane is a thread so each column get 8 VGPRs used to store fragments Here is the layout for C and D matrices and how they map to registers

Lane index 0 1 2 3 … 13 14 15 … 17 18 … 30 31

VGPR0 | 1,1 | 1,2 | 1,3 | 1,4 | … | 1,13 | 1,14 | 1,15 | … | 2,1 | 2,2 | … | 2,15 | 2,16 |

VGPR1 | 3,1 | 3,2 | 3,3 | 3,4 | … | 3,13 | 3,14 | 3,15 | … | 4,1 | 4,2 | … | 4,15 | 4,16 |

VGPR2 | 5,1 | 5,2 | 5,3 | 5,4 | … | 5,13 | 5,14 | 5,15 | … | 6,1 | 6,2 | … | 6,15 | 6,16 |

VGPR3 | 7,1 | 7,2 | 7,3 | 7,4 | … | 7,13 | 7,14 | 7,15 | … | 8,1 | 8,2 | … | 8,15 | 8,16 |

VGPR4 | 9,1 | 9,2 | 9,3 | 9,4 | … | 9,13 | 9,14 | 9,15 | … | 10,1 | 10,2 | … | 10,15| 10,16|

VGPR5 | 11,1 | 11,2 | 11,3 | 11,4 | … | 11,13| 11,14| 11,15| … | 12,1 | 12,2 | … | 12,15| 12,16|

VGPR6 | 13,1 | 13,2 | 13,3 | 13,4 | … | 13,13| 13,14| 13,15| … | 14,1 | 14,2 | … | 14,15| 14,16|

VGPR7 | 15,1 | 15,2 | 15,3 | 15,4 | … | 15,13| 15,14| 15,15| … | 16,1 | 16,2 | … | 16,15| 16,16|

Trait Implementations

impl<D: Clone + Dialect> Clone for WmmaLoad<D>

fn clone(&self) -> WmmaLoad<D>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<D: Debug + Dialect> Debug for WmmaLoad<D>

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl<D: PartialEq + Dialect> PartialEq for WmmaLoad<D>

fn eq(&self, other: &WmmaLoad<D>) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl<D: Dialect> StructuralPartialEq for WmmaLoad<D>