MCPcopy Create free account

hub / github.com/tile-ai/tilelang / functions

Functions7,254 in github.com/tile-ai/tilelang

Functionieee_add
IEEE-compliant addition with specified rounding mode Parameters ---------- x : PrimExpr First operand. y : PrimExpr S
tilelang/language/math_intrinsics.py:152
Functionieee_fadd
IEEE-754 add with explicit rounding mode.
tilelang/contrib/cutedsl/ieee_math.py:274
Functionieee_fdiv
IEEE-754 divide with explicit rounding mode.
tilelang/contrib/cutedsl/ieee_math.py:304
Functionieee_fdiv
IEEE-compliant division with specified rounding mode Parameters ---------- x : PrimExpr Dividend. y : PrimExpr Diviso
tilelang/language/math_intrinsics.py:312
Functionieee_fmaf
IEEE-754 fused multiply-add with explicit rounding mode.
tilelang/contrib/cutedsl/ieee_math.py:289
Functionieee_fmaf
IEEE-compliant fused multiply-add with specified rounding mode Parameters ---------- x : PrimExpr First operand. y : PrimExpr
tilelang/language/math_intrinsics.py:226
Functionieee_fmul
IEEE-754 multiply with explicit rounding mode.
tilelang/contrib/cutedsl/ieee_math.py:284
Functionieee_frcp
IEEE-754 reciprocal with explicit rounding mode.
tilelang/contrib/cutedsl/ieee_math.py:294
Functionieee_frcp
IEEE-compliant reciprocal with specified rounding mode Parameters ---------- x : PrimExpr Input operand. rounding_mode : str,
tilelang/language/math_intrinsics.py:253
Functionieee_frsqrt
IEEE-compliant reciprocal square root (round to nearest only) Parameters ---------- x : PrimExpr Input operand. Returns
tilelang/language/math_intrinsics.py:295
Functionieee_fsqrt
IEEE-754 square root with explicit rounding mode.
tilelang/contrib/cutedsl/ieee_math.py:299
Functionieee_fsqrt
IEEE-compliant square root with specified rounding mode Parameters ---------- x : PrimExpr Input operand. rounding_mode : str
tilelang/language/math_intrinsics.py:274
Functionieee_fsub
IEEE-754 subtract with explicit rounding mode.
tilelang/contrib/cutedsl/ieee_math.py:279
Functionieee_mul
IEEE-compliant multiplication with specified rounding mode Parameters ---------- x : PrimExpr First operand. y : PrimExpr
tilelang/language/math_intrinsics.py:202
Functionieee_sub
IEEE-compliant subtraction with specified rounding mode Parameters ---------- x : PrimExpr First operand. y : PrimExpr
tilelang/language/math_intrinsics.py:178
Functionif_then_else
Conditional selection expression. Parameters ---------- cond : PrimExpr The condition t : PrimExpr The result expres
tilelang/language/tir/op.py:3228
Methodifexp
(self, cond: Any, then: Callable[[], Any], otherwise: Callable[[], Any])
tilelang/language/eager/ast.py:226
Methodifexp
(self, cond, then, otherwise)
tilelang/language/eager/builder.py:633
Functionimage_assets
()
examples/bitnet-1.58b/vllm_workspace/conftest.py:126
Functionimport_source
(source: str | None = None)
tilelang/language/__init__.py:178
Functioninc_max_nreg
Increment the maximum number of registers to use.
tilelang/language/builtin.py:398
Functionincrease_descriptor_offset
(desc: GmmaDescriptor, offset)
tilelang/contrib/cutedsl/gemm_v2.py:51
Functionincrease_descriptor_offset
Increase the offset of a memory descriptor. Parameters: descriptor (PrimExpr): The memory descriptor to modify. offset (Prim
tilelang/language/builtin.py:1231
Functionindex_map
Create a TIR Index mapping
tilelang/language/ast/ir.py:1667
Functionindexdiv
Compute floor(a / b) where a and b are non-negative. Parameters ---------- a : PrimExpr The left hand operand, known to be non-ne
tilelang/language/tir/op.py:3286
Functionindexmod
Compute the remainder of indexdiv. a and b are non-negative. Parameters ---------- a : PrimExpr The left hand operand, known to b
tilelang/language/tir/op.py:3314
Methodinfer
(self, shape, rstep: dict[str, int] = None)
tilelang/carver/roller/shape_inference/common.py:43
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/rocm/op/gemm/gemm_wmma.py:43
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/rocm/op/gemm/gemm_mfma.py:21
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/cpu/op/gemm/gemm_scalar.py:16
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/tileop/gemm/gemm_base.py:40
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/tileop/gemm_sp/gemm_sp_wgmma.py:35
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/tileop/gemm_sp/gemm_sp_base.py:15
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/metal/op/gemm/gemm_metal.py:20
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/cuda/op/gemm/gemm_mma_sm70.py:22
Methodinfer_layout
Infer swizzled layouts for operands and accumulator. For SS: both A and B get swizzled shared-memory layouts. For TS: A and C get TME
tilelang/cuda/op/gemm/gemm_tcgen05.py:70
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/cuda/op/gemm/gemm_wgmma.py:51
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/cuda/op/gemm/gemm_mma.py:42
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/cuda/op/gemm_sp/gemm_sp_wgmma.py:35
Methodinfer_layout
(self, target: Target, thread_nums: int)
tilelang/cuda/op/gemm_sp/gemm_sp_mma.py:17
Methodinfer_node_smem_usage
(self, td: TileDict, node: PrimFuncNode)
tilelang/carver/roller/policy/tensorcore.py:81
Functioninfinity
infinity value of dtype Parameters ---------- dtype : str The data type. span : Optional[Span] The location of this
tilelang/language/tir/op.py:2177
Functioninit
The block initialization statement. Returns ------- res : frame.BlockInitFrame The BlockInitFrame.
tilelang/language/ast/ir.py:363
Methodinitialize_function
Defines and initializes the GEMV computation function. This method sets up placeholders for input matrices, computes the mat
tilelang/carver/template/gemv.py:41
Methodinitialize_function
Defines and initializes the matrix multiplication computation. This method sets up placeholders for input matrices, computes
tilelang/carver/template/flashattention.py:41
Methodinitialize_function
Defines and initializes the matrix multiplication computation. This method sets up placeholders for input matrices, computes
tilelang/carver/template/matmul.py:54
Methodinitialize_function
Parse the structure (e.g., 'SSR'), build the TVM compute definition with the appropriate spatial and reduce axes, and store it in sel
tilelang/carver/template/general_reduce.py:21
Methodinitialize_function
Defines and initializes the convolution computation. This method sets up placeholders for input matrices, computes the convo
tilelang/carver/template/conv.py:61
Methodinitialize_function
Initializes the element-wise computation function. Defines a simple element-wise computation: B = A + 1, where A is an input tensor.
tilelang/carver/template/elementwise.py:38
Functioninitialize_tcgen05_descriptor
Pack the tcgen05 SMEM descriptor bitfields. Matches the C++ ``initialize_tcgen05_descriptor`` in common.h: Low 32 bits (reg32_[0]):
tilelang/contrib/cutedsl/gemm_tcgen05.py:67
Functioninitialize_tcgen05_descriptor
Initialize a TCGEN05 shared-memory descriptor.
tilelang/language/builtin.py:1197
Functioninitialize_wgmma_descriptor
(layout_type, leading_byte_offset, stride_byte_offset, desc: GmmaDescriptor, start_address: cute.Pointer)
tilelang/contrib/cutedsl/gemm_v2.py:37
Functioninitialize_wgmma_descriptor
Initialize a WGMMA/UTCMMA shared-memory descriptor.
tilelang/language/builtin.py:1167
Functioninline_transpose_block
(sch: Schedule, blocks: list[SBlockRV])
tilelang/carver/matmul_analysis.py:472
Functioninner
(func)
tilelang/testing/__init__.py:167
Methodinner
(**config_arg)
tilelang/autotuner/tuner.py:552
Functioninput_guard
A decorator to make sure all input tensors are contiguous and set the device based on input tensors.
examples/kda/FLA_KDA/fla_utils.py:155
Methodinputs
(self)
tilelang/carver/roller/node.py:129
Methodinter_transform_a
(self)
tilelang/carver/roller/hint.py:146
Methodinter_transform_b
(self)
tilelang/carver/roller/hint.py:150
Methodis_after_reduce_stage
(block)
tilelang/carver/roller/node.py:460
Methodis_ampere_arch
Checks if the current architecture is an Ampere architecture. Returns: bool: True if the architecture is Ampere, False o
tilelang/carver/template/base.py:81
Methodis_blockscaled
(self)
tilelang/tileop/gemm/gemm_base.py:200
Functionis_broadcast_epilogue
Check if the epilogue block is a broadcast pattern
tilelang/carver/analysis.py:246
Methodis_broadcast_pattern
(buffer, output_buffer)
tilelang/carver/roller/node.py:453
Methodis_cdna_arch
Checks if the current architecture is a CDNA architecture. Returns: bool: True if the architecture is CDNA, False otherw
tilelang/carver/template/base.py:90
Methodis_cont
(shape, vec)
tilelang/carver/roller/policy/default.py:776
Functionis_cpu_arch
(arch: TileDevice)
tilelang/carver/arch/cpu.py:6
Functionis_darwin
()
tilelang/contrib/cc.py:100
Functionis_device_call
(func: tirx.PrimFunc)
tilelang/engine/lower.py:43
Functionis_device_call_c_device
(func: tirx.PrimFunc)
tilelang/engine/lower.py:31
Methodis_dynamic
Indicates whether the kernel handles dynamic shapes.
tilelang/jit/adapter/cython/adapter.py:383
Methodis_elementwise
Whether the block is elementwise, i.e. trivial mapping between read/write region
tilelang/carver/analysis.py:80
Methodis_float4
Checks if the parameter represents a float4 type. Returns: bool: True if parameter is a float4 type, False otherwise
tilelang/engine/param.py:103
Methodis_gemm
Whether the block is a GEMM workload.
tilelang/carver/analysis.py:109
Methodis_gemm_rr
Return True if both A and B are in registers (RR variant).
tilelang/tileop/gemm/gemm_base.py:69
Methodis_gemm_rr
(self)
tilelang/tileop/gemm_sp/gemm_sp_wgmma.py:160
Methodis_gemm_rr
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:30
Methodis_gemm_rr
(self)
tilelang/cuda/op/gemm/gemm_mma_sm70.py:182
Methodis_gemm_rr
(self)
tilelang/cuda/op/gemm/gemm_wgmma.py:177
Methodis_gemm_rr
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_wgmma.py:160
Methodis_gemm_rs
Return True if A is in registers and B is in shared memory (RS variant).
tilelang/tileop/gemm/gemm_base.py:61
Methodis_gemm_rs
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:27
Methodis_gemm_sr
Return True if A is in shared memory and B is in registers (SR variant).
tilelang/tileop/gemm/gemm_base.py:57
Methodis_gemm_sr
(self)
tilelang/tileop/gemm_sp/gemm_sp_wgmma.py:154
Methodis_gemm_sr
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:24
Methodis_gemm_sr
(self)
tilelang/cuda/op/gemm/gemm_mma_sm70.py:176
Methodis_gemm_sr
(self)
tilelang/cuda/op/gemm/gemm_wgmma.py:171
Methodis_gemm_sr
(self)
tilelang/cuda/op/gemm_sp/gemm_sp_wgmma.py:154
Methodis_gemm_ss
Return True if both A and B are in shared memory (SS variant).
tilelang/tileop/gemm/gemm_base.py:53
Methodis_gemm_ss
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:21
Methodis_gemv
Whether the block is a GEMV workload.
tilelang/carver/analysis.py:105
Functionis_hip
()
benchmark/blocksparse_attention/benchmark_tilelang_block_sparse_fmha.py:10
Functionis_hopper
(target)
tilelang/contrib/nvcc.py:582
Functionis_identity_block
(block_stmt: tirx.SBlock)
tilelang/carver/matmul_analysis.py:464
Methodis_input_8bit
(self)
tilelang/carver/roller/hint.py:134
Methodis_inter_warp_transform
(self)
tilelang/common/transform_kind.py:14
Methodis_intra_warp_transform
(self)
tilelang/common/transform_kind.py:17
Methodis_ld_matrix_transform
(self)
tilelang/common/transform_kind.py:20
Functionis_local_buffer
Check if a buffer is local (register-level), including local.var and metal.simdgroup.
tilelang/transform/decouple_type_cast.py:74
← previousnext →4,801–4,900 of 7,254, ranked by callers