Code
Hub
Workspaces
Following
Trending
Connect
MCP
copy
Create free account
hub
/
github.com/tile-ai/tilelang
/ functions
Functions
7,254 in github.com/tile-ai/tilelang
⨍
Functions
7,254
◇
Types & classes
525
↳
Endpoints
224
Function
LegalizeVectorizedLoop
LegalizeLoopVectorize Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:173
Function
Let
Create a Let expression binding
tilelang/language/ast/ir.py:904
Function
LetStmt
Create a LetStmt binding Parameters ---------- value : PrimExpr The value to be bound. type_annotation : Optional[Type] = Non
tilelang/language/ast/ir.py:873
Function
LoopUnswitching
LoopUnswitching: Hoist loop-invariant if statements out of loops. Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:162
Function
LowerBlackwell2SM
Lower 2SM TCGEN5MMA and related on Blackwell target Returns: fpass : tvm.transform.Pass The result pass
tilelang/cuda/transform/__init__.py:35
Function
LowerDeviceKernelLaunch
Create and return a transform pass that lowers device kernel launch constructs to target-specific IR. This pass transforms high-level device
tilelang/transform/__init__.py:375
Function
LowerHopperIntrin
LowerHopperIntrin
tilelang/cuda/transform/__init__.py:45
Function
LowerIntrin
LowerIntrin
tilelang/transform/__init__.py:370
Function
LowerL2Persistent
LowerL2Persistent
tilelang/cuda/transform/__init__.py:52
Function
LowerLDGSTG
Lower Ramp-based global memory load/store to ldg/stg intrinsics. This pass transforms vectorized global memory loads and stores (using Ramp indic
tilelang/cuda/transform/__init__.py:72
Function
LowerOpaqueBlock
LowerOpaqueBlock
tilelang/transform/__init__.py:360
Function
LowerSharedBarrier
LowerSharedBarrier
tilelang/cuda/transform/__init__.py:62
Function
LowerSharedTmem
LowerSharedTmem
tilelang/cuda/transform/__init__.py:57
Function
LowerThreadAllreduce
LowerThreadAllreduce
tilelang/transform/__init__.py:365
Function
LowerTileOp
LowerTileOp Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:68
Method
M
(self)
tilelang/tileop/gemm/gemm_base.py:74
Method
M
(self)
tilelang/tileop/gemm/__init__.py:66
Method
M
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:34
Function
MakePackedAPI
MakePackedAPI Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:206
Function
MarkCudaSyncCalls
MarkCudaSyncCalls
tilelang/cuda/transform/__init__.py:94
Function
MergeIfStmt
MergeIfStmt Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:151
Function
MergeSharedMemoryAllocations
MergeSharedMemoryAllocations Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:312
Function
MetalPassPipelineBody
(mod: IRModule, target: Target)
tilelang/metal/pipeline.py:19
Method
N
(self)
tilelang/tileop/gemm/gemm_base.py:78
Method
N
(self)
tilelang/tileop/gemm/__init__.py:70
Method
N
(self)
tilelang/tileop/gemm_sp/gemm_sp_base.py:38
Function
NestedLoopChecker
User-friendly pass which identifies any invalid any nested-loop pattern. Nested loops is an annoying problem in tilelang or other polyhedral
tilelang/analysis/nested_loop_checker.py:61
Function
Parallel
Tools to construct nested parallel for loop. This can be used to create element-wise tensor expression. Parameters ---------- exte
tilelang/language/loop.py:13
Function
PersistThreadblock
PersistThreadblock
tilelang/cuda/transform/__init__.py:143
Function
Persistent
Tools to construct persistent for loop. Parameters ---------- domain : List[tirx.PrimExpr] The list of dominators. wave_size
tilelang/language/loop.py:90
Function
PipelinePlanning
infer the fragment/shared memory layout Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:30
Function
Pipelined
Tools to construct pipelined for loop. Parameters ---------- start : PrimExpr The minimum value of iteration. stop : PrimExpr
tilelang/language/loop.py:112
Function
PlanAndUpdateBufferAllocationLocation
Plan and update buffer allocation locations within PrimFuncs. Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:323
Function
ProducerConsumerWarpSpecializedTiled
Compatibility alias for ``ProducerConsumerWarpSpecialized``. The tiled tile-op implementation is now the canonical ``ProducerConsumerWarpSpec
tilelang/cuda/transform/__init__.py:21
Function
ROCMPassPipelineBody
(mod: IRModule, target: Target)
tilelang/rocm/pipeline.py:18
Method
SFARegion
(self)
tilelang/tileop/gemm/gemm_base.py:188
Method
SFBRegion
(self)
tilelang/tileop/gemm/gemm_base.py:192
Function
Serial
Alias of T.serial.
tilelang/language/loop.py:303
Function
SplitHostDevice
Split host/device functions even for empty kernels. Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:228
Function
StorageRewrite
StorageRewrite Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:349
Function
TVMBackendAllocWorkspace
Backend function to allocate temporal workspace Parameters ---------- device_type : int The device type which the space will be a
tilelang/language/tir/op.py:3494
Function
TVMBackendFreeWorkspace
Backend function to free temporal workspace. Parameters ---------- device_type : int The device type which the space will be allo
tilelang/language/tir/op.py:3522
Function
Then
Create a then. Returns ------- res : frame.ThenFrame The result ThenFrame.
tilelang/language/ast/ir.py:1111
Function
ThreadIdx
Return the current CUDA thread index.
tilelang/contrib/cutedsl/threadblock_swizzle.py:28
Function
ThreadSync
Insert sync between parallel read/write of shared buffers. Parameters ---------- storage_scope: str The target storage scope.
tilelang/transform/__init__.py:124
Function
Unroll
Alias of T.unroll.
tilelang/language/loop.py:315
Function
UnrollLoop
Unroll loops as in Halide pipeline. This pass unrolls loops based on configuration options including: - auto_max_step: Threshold of number of
tilelang/transform/__init__.py:400
Function
VectorizeLoop
VectorizeLoop Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:254
Function
Vectorized
Alias of T.vectorized.
tilelang/language/loop.py:356
Function
VerifyParallelLoop
VerifyParallelLoop Returns ------- fpass : tvm.transform.Pass The result pass
tilelang/transform/__init__.py:113
Function
WarpSpecialize
Tools to construct a warp group frame. Parameters ---------- warp_group_idx : int A integer representing warp group index
tilelang/language/warpgroup.py:17
Function
While
Create a while node. Parameters ---------- condition : PrimExpr The termination condition of the loop. Returns -------
tilelang/language/ast/ir.py:1074
Function
__activemask
Returns a 32-bit integer mask of all currently active threads in the calling warp. PTX: activemask.b32 %mask;
tilelang/contrib/cutedsl/warp.py:31
Method
__add__
(self, offset)
tilelang/contrib/cutedsl/gemm_v2.py:29
Method
__add__
Add byte offset. Like C++ operator+, shifts offset >> 4.
tilelang/contrib/cutedsl/gemm_tcgen05.py:52
Method
__bool__
Allows truthy checks on the stack object itself, e.g., 'if stack: ...'
tilelang/language/kernel.py:68
Method
__bool__
Allows truthy checks on the stack object itself, e.g., 'if stack: ...'
tilelang/language/frame.py:98
Method
__bool__
Return True if the stack is not empty, False otherwise. Returns: bool: Whether the stack contains any items.
tilelang/autotuner/capture.py:71
Method
__call__
(self, x=None)
tilelang/autodd.py:46
Method
__call__
(self, *args: list[PrimExpr])
tilelang/layout/layout.py:250
Method
__call__
(x, y)
tilelang/contrib/cutedsl/reduce.py:139
Method
__call__
(x, y)
tilelang/contrib/cutedsl/reduce.py:147
Method
__call__
(x, y)
tilelang/contrib/cutedsl/reduce.py:155
Method
__call__
(x, y)
tilelang/contrib/cutedsl/reduce.py:164
Method
__call__
(x, y)
tilelang/contrib/cutedsl/reduce.py:172
Method
__call__
(x, y)
tilelang/contrib/cutedsl/reduce.py:180
Method
__call__
GEMM body: both A and B from shared memory
tilelang/contrib/cutedsl/gemm_v1.py:292
Method
__call__
( self, sA_ptr: cute.Pointer, sB_ptr: cute.Pointer, rC_ptr: cute.Pointer,
tilelang/contrib/cutedsl/gemm_v1.py:480
Method
__call__
(self, *args: Any, **kwds: Any)
tilelang/profiler/__init__.py:276
Method
__call__
( self, shape: ShapeType, dtype: DType = "float32", data=None, strides
tilelang/language/proxy.py:95
Method
__call__
(self, shape: ShapeType | PrimExpr | int, dtype: DType = "float32", data=None, scope=None)
tilelang/language/proxy.py:165
Method
__call__
(self, shape: ShapeType, strides: tuple[Any], dtype: DType = "float32", scope=None)
tilelang/language/proxy.py:177
Method
__call__
(self, *args: _P.args, **kwargs: _P.kwargs)
tilelang/language/eager/builder.py:804
Method
__call__
(self, *args, **kwargs)
tilelang/language/eager/builder.py:1229
Method
__call__
( self, shape: ShapeType, dtype: DType = _dtypes.float32, data=None, s
tilelang/language/parser/entry.py:161
Method
__call__
(self, dtype, storage_scope="global")
tilelang/language/parser/entry.py:200
Method
__call__
Make the AutoTuner callable, running the auto-tuning process. Returns: AutotuneResult: Results of the auto-tuning process.
tilelang/autotuner/tuner.py:1198
Method
__call__
(self, *args: _P.args, **kwargs: _P.kwargs)
tilelang/autotuner/tuner.py:1259
Method
__call__
Invokes the compiled function with the given arguments. Parameters ---------- *args : Any Positional arg
tilelang/jit/kernel.py:187
Method
__call__
(self, *args: _P.args, **kwargs: _P.kwargs)
tilelang/jit/__init__.py:448
Method
__call__
(self, *args: Any, **kwds: Any)
tilelang/jit/adapter/base.py:99
Method
__class_getitem__
(cls, key)
tilelang/language/proxy.py:220
Function
__cos
Calculate cos(x) with fast math Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr
tilelang/language/fastmath.py:75
Function
__cos
Calculate cos(x) with fast math Parameters ---------- x : PrimExpr Input argument. Returns ------- y : PrimExpr
tilelang/language/math_intrinsics.py:83
Method
__del__
Cleanup: unload CUDA library when object is destroyed. Critical for resource management - CUDA libraries consume GPU memory. Failure
tilelang/jit/adapter/nvrtc/libgen.py:245
Function
__dtype_as_torch__
Convert TileLang dtype to PyTorch dtype.
tilelang/language/dtypes.py:184
Function
__dtype_bytes__
Return the number of bytes for this dtype.
tilelang/language/dtypes.py:251
Function
__dtype_call__
(self: dtype, *args, is_size_var: bool = False)
tilelang/language/dtypes.py:146
Function
__dtype_is_float4__
Whether this is any FP4 E2M1 logical variant (packed or unpacked storage).
tilelang/language/dtypes.py:284
Function
__dtype_new__
(cls, value: AnyDType)
tilelang/language/dtypes.py:239
Method
__enter__
(self)
examples/bitnet-1.58b/vllm_workspace/conftest.py:398
Method
__enter__
(self)
examples/bitnet-1.58b/vllm_workspace/conftest.py:544
Method
__enter__
(self)
tilelang/autodd.py:40
Method
__enter__
(self)
tilelang/autodd.py:799
Method
__enter__
(self)
tilelang/autodd.py:861
Method
__enter__
(self)
tilelang/profiler/bench.py:19
Method
__enter__
Enters the KernelLaunchFrame scope and pushes this frame onto the stack. Returns one Var if we detect exactly 5 frames (meaning there
tilelang/language/kernel.py:140
Method
__enter__
Enter the let frame scope and process buffer loads. Returns: Var: The variable bound in this frame
tilelang/language/frame.py:139
Method
__enter__
(self)
tilelang/autotuner/capture.py:93
Method
__eq__
(self, other)
tilelang/language/eager/builder.py:816
← previous
next →
3,401–3,500 of 7,254, ranked by callers