Skip to main content

/

DocsCode
Log in

Mojo struct

VBuffer

struct VBuffer[dtype: DType, layout: Layout, address_space: AddressSpace, alignment: Int, origin: Origin[], masked: Bool, //, mma_shape: IndexList[3], k_group_size: Int, BN: Int, BK: Int, depth: Int, num_threads: Int]

Fields​

  • ​load_tile (LayoutTensor[dtype, row_major((0 if ((4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]() * num_threads) == 0) else (div_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) + -1) if (((rem_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) == 0) ^ True) & (((min(depth, 128) * BK) < 0) ^ ((4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]() * num_threads) < 0))) else div_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) * 0 if (min(depth, 128) == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, 0), {1}, apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128})), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, 0), {1}, apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128})), "value">) == 0) ^ True) & ((min(depth, 128) < 0) ^ (depth < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, 0), {1}, apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128})), "value">)), 4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]()), MutableAnyOrigin, address_space=AddressSpace(5)]):
  • ​mma_tile (LayoutTensor[dtype, row_major((ceildiv[::CeilDivable](BK, (mma_shape.__getitem__[::Indexer](2) * k_group_size)) * 0 if (mma_shape.__getitem__[::Indexer](0) == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">) == 0) ^ True) & ((mma_shape.__getitem__[::Indexer](0) < 0) ^ (depth < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">)), simdwidthof[::DType,__mlir_type.!kgen.target]()), MutableAnyOrigin, address_space=AddressSpace(5)]):
  • ​smem_iter (LayoutTensorIter[dtype, blocked_product(row_major(pad[::Int](), simdwidthof[::DType,__mlir_type.!kgen.target]()), row_major(1, 0 if (simdwidthof[::DType,__mlir_type.!kgen.target]() == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) == 0) ^ True) & ((simdwidthof[::DType,__mlir_type.!kgen.target]() < 0) ^ (BK < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">)), True), MutableAnyOrigin, address_space=AddressSpace(3), circular=True]):
  • ​global_iterator (LayoutTensorIter[dtype, _compute_tile_layout[*::Int]()[0], origin, address_space=address_space, axis=OptionalReg[Int]({:@stdlib::@builtin::@int::@Int {0}, 0}), layout_int_type=_get_layout_type(layout, address_space), linear_idx_type=_get_index_type(layout, address_space), masked=masked if masked else _tile_is_masked[::Layout,*::Int]()]):
  • ​global_base_tile (LayoutTensor[dtype, layout, origin, address_space=address_space, masked=masked, alignment=alignment]):

Implemented traits​

AnyType, UnknownDestructibility

Aliases​

__del__is_trivial​

alias __del__is_trivial = True

base_layout​

alias base_layout = row_major(pad[::Int](), simdwidthof[::DType,__mlir_type.!kgen.target]())

depth_tile_size​

alias depth_tile_size = min(depth, 128)

GlobalTensorType​

alias GlobalTensorType = LayoutTensor[dtype, layout, origin, address_space=address_space, masked=masked, alignment=alignment]

GlobalTiledIteratorType​

alias GlobalTiledIteratorType = LayoutTensorIter[dtype, _compute_tile_layout[*::Int]()[0], origin, address_space=address_space, axis=OptionalReg[Int]({:@stdlib::@builtin::@int::@Int {0}, 0}), layout_int_type=_get_layout_type(layout, address_space), linear_idx_type=_get_index_type(layout, address_space), masked=masked if masked else _tile_is_masked[::Layout,*::Int]()]

load_width​

alias load_width = 4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]()

loads_per_thread_per_depth_tile​

alias loads_per_thread_per_depth_tile = 0 if ((4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]() * num_threads) == 0) else (div_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) + -1) if (((rem_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) == 0) ^ True) & (((min(depth, 128) * BK) < 0) ^ ((4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]() * num_threads) < 0))) else div_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">)

LoadTileType​

alias LoadTileType = LayoutTensor[dtype, row_major((0 if ((4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]() * num_threads) == 0) else (div_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) + -1) if (((rem_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) == 0) ^ True) & (((min(depth, 128) * BK) < 0) ^ ((4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]() * num_threads) < 0))) else div_s(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">), #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">), 0), {1}, {value = mul(#lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, 64), {4}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int num_threads, "value">)}), "value">) * 0 if (min(depth, 128) == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, 0), {1}, apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128})), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, 0), {1}, apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128})), "value">) == 0) ^ True) & ((min(depth, 128) < 0) ^ (depth < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128}), "value">, 0), {1}, apply(:!lit.generator<("x": !lit.struct<@stdlib::@builtin::@int::@Int>, "y": !lit.struct<@stdlib::@builtin::@int::@Int>, |) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@builtin::@math::@"min(::Int,::Int)", depth, {128})), "value">)), 4 if (depth == 64) else simdwidthof[::DType,__mlir_type.!kgen.target]()), MutableAnyOrigin, address_space=AddressSpace(5)]

MMA_K​

alias MMA_K = mma_shape.__getitem__[::Indexer](2)

MMA_M​

alias MMA_M = mma_shape.__getitem__[::Indexer](0)

MMATileType​

alias MMATileType = LayoutTensor[dtype, row_major((ceildiv[::CeilDivable](BK, (mma_shape.__getitem__[::Indexer](2) * k_group_size)) * 0 if (mma_shape.__getitem__[::Indexer](0) == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">) == 0) ^ True) & ((mma_shape.__getitem__[::Indexer](0) < 0) ^ (depth < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">)), simdwidthof[::DType,__mlir_type.!kgen.target]()), MutableAnyOrigin, address_space=AddressSpace(5)]

num_depth_tiles​

alias num_depth_tiles = 0 if (mma_shape.__getitem__[::Indexer](0) == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">) == 0) ^ True) & ((mma_shape.__getitem__[::Indexer](0) < 0) ^ (depth < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int depth, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> rebind(:!lit.generator<[1]("self": !lit.struct<@stdlib::@utils::@index::@IndexList<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}>>, "idx": !lit.ref<@stdlib::@builtin::@int::@Int, imm *[0,0]> read_mem) -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@utils::@index::@IndexList::@"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:@stdlib::@builtin::@int::@Int {3}, :@stdlib::@builtin::@dtype::@DType {:dtype si64}, :trait<@stdlib::@builtin::@int::@Indexer> @stdlib::@builtin::@int::@Int>), mma_shape, store_to_mem({0}))), "value">)

num_k_tiles​

alias num_k_tiles = ceildiv[::CeilDivable](BK, (mma_shape.__getitem__[::Indexer](2) * k_group_size))

num_repeats​

alias num_repeats = 0 if (simdwidthof[::DType,__mlir_type.!kgen.target]() == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) == 0) ^ True) & ((simdwidthof[::DType,__mlir_type.!kgen.target]() < 0) ^ (BK < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">)

SharedIterType​

alias SharedIterType = LayoutTensorIter[dtype, blocked_product(row_major(pad[::Int](), simdwidthof[::DType,__mlir_type.!kgen.target]()), row_major(1, 0 if (simdwidthof[::DType,__mlir_type.!kgen.target]() == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) == 0) ^ True) & ((simdwidthof[::DType,__mlir_type.!kgen.target]() < 0) ^ (BK < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">)), True), MutableAnyOrigin, address_space=AddressSpace(3), circular=True]

SharedTileType​

alias SharedTileType = LayoutTensor[dtype, blocked_product(row_major(pad[::Int](), simdwidthof[::DType,__mlir_type.!kgen.target]()), row_major(1, 0 if (simdwidthof[::DType,__mlir_type.!kgen.target]() == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) == 0) ^ True) & ((simdwidthof[::DType,__mlir_type.!kgen.target]() < 0) ^ (BK < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">)), True), MutableAnyOrigin, address_space=AddressSpace(3), layout_int_type=_get_index_type(AddressSpace(3)), linear_idx_type=_get_index_type(AddressSpace(3))]

simd_width​

alias simd_width = simdwidthof[::DType,__mlir_type.!kgen.target]()

smem_layout​

alias smem_layout = blocked_product(row_major(pad[::Int](), simdwidthof[::DType,__mlir_type.!kgen.target]()), row_major(1, 0 if (simdwidthof[::DType,__mlir_type.!kgen.target]() == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) == 0) ^ True) & ((simdwidthof[::DType,__mlir_type.!kgen.target]() < 0) ^ (BK < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">)), True)

tiler_layout​

alias tiler_layout = row_major(1, 0 if (simdwidthof[::DType,__mlir_type.!kgen.target]() == 0) else (div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) + -1) if (((rem_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">) == 0) ^ True) & ((simdwidthof[::DType,__mlir_type.!kgen.target]() < 0) ^ (BK < 0))) else div_s(#lit.struct.extract<:@stdlib::@builtin::@int::@Int BK, "value">, #lit.struct.extract<:@stdlib::@builtin::@int::@Int cond(eq(#lit.struct.extract<:@stdlib::@builtin::@int::@Int apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>), "value">, 0), {1}, apply(:!lit.generator<() -> !lit.struct<@stdlib::@builtin::@int::@Int>> @stdlib::@sys::@info::@"simdwidthof[::DType,__mlir_type.!kgen.target]()"<:@stdlib::@builtin::@dtype::@DType dtype, :target apply(:!lit.generator<() -> !kgen.target> @stdlib::@sys::@info::@"_current_target()")>)), "value">))

Methods​

__init__​

__init__(out self, global_tile: LayoutTensor[dtype, layout, origin, address_space=address_space, masked=masked, alignment=alignment], shared_ptr: UnsafePointer[SIMD[dtype, 1], address_space=AddressSpace(3), alignment=alignment, mut=mut, origin=origin])

pad​

static pad[dim: Int]() -> Int

Returns:

Int

load_from_dram​

load_from_dram(mut self)

get_mma_tile​

get_mma_tile(self) -> LayoutTensor[dtype, row_major((ceildiv[::CeilDivable](BK, (mma_shape.__getitem__[::Indexer](2) * k_group_size)) * 0 if (mma_shape.__getitem__[::Indexer](0) == 0) else (div_s(#lit.struct.extract<:_stdlib::_builtin::_int::_Int depth, "value">, #lit.struct.extract<:_stdlib::_builtin::_int::_Int cond(eq(#lit.struct.extract<:_stdlib::_builtin::_int::_Int apply(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> rebind(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm *[0,0]> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> _stdlib::_utils::_index::_IndexList::_"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}, :trait<_stdlib::_builtin::_int::_Indexer> _stdlib::_builtin::_int::_Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> rebind(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm *[0,0]> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> _stdlib::_utils::_index::_IndexList::_"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}, :trait<_stdlib::_builtin::_int::_Indexer> _stdlib::_builtin::_int::_Int>), mma_shape, store_to_mem({0}))), "value">) + -1) if (((rem_s(#lit.struct.extract<:_stdlib::_builtin::_int::_Int depth, "value">, #lit.struct.extract<:_stdlib::_builtin::_int::_Int cond(eq(#lit.struct.extract<:_stdlib::_builtin::_int::_Int apply(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> rebind(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm *[0,0]> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> _stdlib::_utils::_index::_IndexList::_"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}, :trait<_stdlib::_builtin::_int::_Indexer> _stdlib::_builtin::_int::_Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> rebind(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm *[0,0]> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> _stdlib::_utils::_index::_IndexList::_"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}, :trait<_stdlib::_builtin::_int::_Indexer> _stdlib::_builtin::_int::_Int>), mma_shape, store_to_mem({0}))), "value">) == 0) ^ True) & ((mma_shape.__getitem__[::Indexer](0) < 0) ^ (depth < 0))) else div_s(#lit.struct.extract<:_stdlib::_builtin::_int::_Int depth, "value">, #lit.struct.extract<:_stdlib::_builtin::_int::_Int cond(eq(#lit.struct.extract<:_stdlib::_builtin::_int::_Int apply(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> rebind(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm *[0,0]> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> _stdlib::_utils::_index::_IndexList::_"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}, :trait<_stdlib::_builtin::_int::_Indexer> _stdlib::_builtin::_int::_Int>), mma_shape, store_to_mem({0})), "value">, 0), {1}, apply(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm #lit.comptime.origin> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> rebind(:!lit.generator<[1]("self": !lit.struct<_stdlib::_utils::_index::_IndexList<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}>>, "idx": !lit.ref<_stdlib::_builtin::_int::_Int, imm *[0,0]> read_mem) -> !lit.struct<_stdlib::_builtin::_int::_Int>> _stdlib::_utils::_index::_IndexList::_"__getitem__[::Indexer](::IndexList[$0, $1],$2)"<:_stdlib::_builtin::_int::_Int {3}, :_stdlib::_builtin::_dtype::_DType {:dtype si64}, :trait<_stdlib::_builtin::_int::_Indexer> _stdlib::_builtin::_int::_Int>), mma_shape, store_to_mem({0}))), "value">)), simdwidthof[::DType,__mlir_type.!kgen.target]()), MutableAnyOrigin, address_space=AddressSpace(5)]

Returns:

LayoutTensor

copy_to_shared​

copy_to_shared(self)

load_from_shared​

load_from_shared(self)

View source

View source

Was this page helpful?

PRODUCT

Mammoth

MAX

Mojo

Editions

Install

QUICK START

Documentation

Tutorials

Model Repo

Recipes

GPU Puzzles

Quickstart

SOLUTIONS

AI Agents

AI Inference

Batch processing

Chatbots

Code Generation

RAG & CAG

Research

DEVELOPERS

Modular Changelog

Tech Talks

Mojo🔥 Playground

Hackathons

AI Coding Assistant

FAQ

CONNECT

Blog

Community

Community Forum

Community Highlights

Report Issue

COMPANY

About Us

Culture

Careers

Talk to Us

@ Copyright - Modular Inc - 2025

Terms,

Privacy

&

Acceptable Use