mirror of
https://github.com/tile-ai/tilelang.git
synced 2026-10-22 06:50:25 +08:00
Lei Wang
bf8a6fc142
* [Refactor] Improve flash attention example and layout comparison logic - Removed unnecessary annotation for `lse_local_split` in the flash attention example to streamline the code. - Updated the handling of `lse_local_split` to utilize parallel processing for better performance. - Refactored kernel compilation and profiling logic to enhance clarity and maintainability in the flash attention example. - Added a condition in `FragmentNode::IsEqual` to handle broadcast cases, improving the robustness of layout comparisons. * lint fix * [Enhancement] Add support for shared memory scope in Fill operation - Introduced handling for `shared.dyn` and `shared` memory scopes in the Fill operation. - Implemented parallel operation and layout inference for improved performance in shared memory scenarios. - Updated thread loop partitioning and vectorization logic to accommodate new memory scope handling. * [Refactor] Remove deprecated decorator and enhance Cython kernel handling - Removed the deprecated decorator from the main module and added a new implementation in the utils module for better organization. - Introduced a pointer map in the Cython kernel adapter to manage pointer arguments, improving runtime shape resolution. - Updated the Cython kernel wrapper to utilize the new pointer map for handling kernel arguments. - Enhanced error checking in the tensor utility functions to ensure static shapes are enforced. - Added a new proxy module for buffer and tensor handling, streamlining the interface for TIR programs. * [Feature] Add matrix multiplication test and kernel implementation - Introduced a new test file `test_tilelang_language_ptr.py` that implements a matrix multiplication function using TileLang's primitives. - The `matmul_test` function defines a kernel for performing tile-level GEMM operations with customizable block sizes and data types. - Added a `run_matmul` function to compile and execute the kernel, along with a test function to validate the implementation. - Updated the `proxy.py` file to enhance type handling for buffer and tensor proxies, ensuring compatibility with TIR programs. - Minor formatting improvements in `deprecated.py` for better readability. * lint fix * [Refactor] Update tensor creation in matrix multiplication test - Replaced `T.Tensor.from_ptr` with `T.make_tensor` in `matmul_test` for improved clarity and consistency. - Updated imports in `__init__.py` to include `make_tensor`. - Added `make_tensor` function in `proxy.py` to streamline tensor creation from pointers. * [Refactor] Update tensor definitions across multiple files - Replaced instances of `T.Tensor` with updated tensor definitions in various benchmark and example files to enhance consistency and clarity. - Adjusted tensor shapes and types in functions related to matrix multiplication, attention mechanisms, and other operations. - Improved documentation in README and example files to reflect changes in tensor usage. * lint fix * [Refactor] Update tensor types in attention and matrix multiplication examples - Replaced instances of `T.Tensor` with `T.SharedTensor` and `T.FragmentTensor` in various attention and matrix multiplication functions to improve consistency and clarity. - Adjusted tensor definitions in benchmark and example files to align with the new tensor types. - Enhanced the overall structure and readability of the code by standardizing tensor usage across multiple files. * lint fix * [Refactor] Update tensor types in GEMM example and test files - Replaced instances of `T.Tensor` with `T.LocalTensor` and `T.Buffer` in the GEMM example and related test functions to improve consistency and clarity. - Enhanced the overall structure of the code by standardizing tensor usage across multiple files, aligning with recent updates in tensor definitions. * [Refactor] Update tensor usage in customize.py - Replaced instances of `T.Tensor` with `T.Buffer` in the `reshape` and `view` functions to enhance consistency with recent tensor definitions. - Improved code clarity by standardizing buffer usage across the file. * [Refactor] Update tensor types in test_tilelang_transform_annotate_device_regions.py - Replaced instances of `T.Tensor` with `T.Buffer` in the `before` and `expected` methods of the `TestAnnotateThreadExtent` and `TestAnnotateDeviceScope` classes to enhance consistency with recent tensor definitions. - Improved code clarity by standardizing buffer usage across the test file. * [Refactor] Update tensor types to SharedBuffer and FragmentBuffer - Replaced instances of `T.SharedTensor` and `T.FragmentTensor` with `T.SharedBuffer` and `T.FragmentBuffer` across multiple benchmark, example, and test files to enhance consistency with recent tensor definitions. - Improved code clarity and structure by standardizing buffer usage in attention and matrix multiplication functions. * [Refactor] Introduce Tensor alias for Buffer in proxy.py - Added a new alias `Tensor` for `Buffer` in `proxy.py` to facilitate JIT compilation, ensuring that inputs and outputs are mapped with `torch.Tensor`. - This change enhances clarity and consistency in tensor usage across the codebase.
118 lines
3.0 KiB
Python
118 lines
3.0 KiB
Python
import tilelang.testing
|
|
from tilelang.utils.tensor import map_torch_type
|
|
|
|
|
|
def clamp_within_bounds(
|
|
N,
|
|
block_N,
|
|
dtype,
|
|
min_val=None,
|
|
max_val=None,
|
|
):
|
|
import tilelang.language as T
|
|
|
|
@T.prim_func
|
|
def main(
|
|
A: T.Tensor((N,), dtype),
|
|
B: T.Tensor((N,), dtype),
|
|
):
|
|
with T.Kernel(T.ceildiv(N, block_N), threads=block_N) as bx:
|
|
A_shared = T.alloc_shared([block_N], dtype)
|
|
T.copy(A[bx * block_N], A_shared)
|
|
for i in T.Parallel(block_N):
|
|
A_shared[i] = T.clamp(A_shared[i], min_val=min_val, max_val=max_val)
|
|
T.copy(A_shared, B[bx * block_N])
|
|
|
|
return main
|
|
|
|
|
|
def run_clamp(
|
|
N,
|
|
block_N,
|
|
dtype,
|
|
min=None,
|
|
max=None,
|
|
):
|
|
program = clamp_within_bounds(N, block_N, dtype, min, max)
|
|
|
|
kernel = tilelang.compile(program, out_idx=[1])
|
|
profiler = kernel.get_profiler()
|
|
|
|
def ref_program(A):
|
|
import torch
|
|
|
|
output = torch.clamp(A, min, max)
|
|
return output
|
|
|
|
profiler.assert_allclose(ref_program, atol=1e-2, rtol=1e-2)
|
|
|
|
|
|
def clamp_value_range(
|
|
N,
|
|
block_N,
|
|
dtype,
|
|
):
|
|
import tilelang.language as T
|
|
|
|
@T.prim_func
|
|
def main(
|
|
A: T.Tensor((1, N), dtype),
|
|
B: T.Tensor((1, N), dtype),
|
|
):
|
|
with T.Kernel(T.ceildiv(N, block_N), threads=block_N) as bx:
|
|
# A_shared = T.alloc_shared([1, block_N], dtype=dtype)
|
|
A_frag = T.alloc_fragment([1, block_N], dtype=dtype)
|
|
min_frag = T.alloc_fragment([1], dtype=dtype)
|
|
max_frag = T.alloc_fragment([1], dtype=dtype)
|
|
T.copy(A[0, bx * block_N], A_frag)
|
|
T.reduce_min(A_frag, min_frag, dim=1)
|
|
T.reduce_max(A_frag, max_frag, dim=1)
|
|
for i in T.Parallel(block_N):
|
|
# A_frag[0, i] = T.max(A_frag[0, i], min_frag[0] * 0.5)
|
|
# A_frag[0, i] = T.min(A_frag[0, i], max_frag[0] * 0.5)
|
|
A_frag[0, i] = T.clamp(A_frag[0, i], min_frag[0] * 0.5, max_frag[0] * 0.5)
|
|
T.copy(A_frag, B[0, bx * block_N])
|
|
|
|
return main
|
|
|
|
|
|
def run_clamp_value_range(
|
|
N,
|
|
block_N,
|
|
dtype,
|
|
):
|
|
program = clamp_value_range(
|
|
N,
|
|
block_N,
|
|
dtype,
|
|
)
|
|
kernel = tilelang.compile(program, out_idx=[1])
|
|
|
|
import torch
|
|
|
|
# Convert string dtype to torch.dtype
|
|
torch_dtype = map_torch_type(dtype)
|
|
|
|
def ref_program(A):
|
|
min_val = torch.min(A) * 0.5
|
|
max_val = torch.max(A) * 0.5
|
|
output = torch.clamp(A, min_val, max_val)
|
|
return output
|
|
|
|
A = torch.randint(-5, 5, (1, N)).cuda().to(dtype=torch_dtype)
|
|
B = kernel(A)
|
|
ref_b = ref_program(A)
|
|
torch.testing.assert_close(B, ref_b)
|
|
|
|
|
|
def test_clamp():
|
|
# clamp tests for float16 and float32
|
|
run_clamp(1024, 128, "float16", -0.05, 0.05)
|
|
run_clamp(1024, 128, "float32", -0.06, 0.05)
|
|
run_clamp_value_range(1024, 128, "float16")
|
|
run_clamp_value_range(1024, 128, "float32")
|
|
|
|
|
|
if __name__ == "__main__":
|
|
tilelang.testing.main()
|