mirror of
https://github.com/tile-ai/tilelang.git
synced 2026-10-23 07:00:29 +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
4.1 KiB
Python
118 lines
4.1 KiB
Python
import tilelang
|
|
import tilelang.testing
|
|
from tilelang import tvm as tvm
|
|
import tilelang.language as T
|
|
import torch
|
|
|
|
|
|
def matmul(M, N, K, block_M, block_N, block_K, dtype="float16", accum_dtype="float"):
|
|
num_stages = 0
|
|
|
|
@T.prim_func
|
|
def matmul(
|
|
A: T.Tensor((M, K), dtype),
|
|
B: T.Tensor((K, N), dtype),
|
|
C: T.Tensor((M, N), dtype),
|
|
):
|
|
with T.Kernel(T.ceildiv(N, block_N), T.ceildiv(M, block_M), is_cpu=True) as (bx, by):
|
|
A_local = T.alloc_local((block_M, block_K), dtype)
|
|
B_local = T.alloc_local((block_K, block_N), dtype)
|
|
C_local = T.alloc_local((block_M, block_N), accum_dtype)
|
|
|
|
T.clear(C_local)
|
|
|
|
# Apply layout optimizations or define your own layout
|
|
# (Optional).
|
|
# T.annotate_layout(
|
|
# {
|
|
# A_local: make_swizzle_layout(A_local),
|
|
# B_local: make_swizzle_layout(B_local),
|
|
# }
|
|
# )
|
|
|
|
for ko in T.Pipelined(K // block_K, num_stages=num_stages):
|
|
|
|
T.copy(A[by * block_M, ko * block_K], A_local)
|
|
|
|
# Or Copy with Parallel
|
|
for k, j in T.Parallel(block_K, block_N):
|
|
B_local[k, j] = B[ko * block_K + k, by * block_N + j]
|
|
|
|
for i, j, k in T.grid(block_M, block_N, block_K):
|
|
C_local[i, j] += A_local[i, k] * B_local[k, j]
|
|
|
|
T.copy(C_local, C[by * block_M, bx * block_N])
|
|
|
|
return matmul
|
|
|
|
|
|
def assert_matmul_codegen(M=1024, N=1024, K=1024, block_M=128, block_N=128, block_K=32):
|
|
func = matmul(M, N, K, block_M, block_N, block_K)
|
|
|
|
artifact = tilelang.lower(func, target="c")
|
|
|
|
code = artifact.kernel_source
|
|
|
|
assert code is not None, "Code generation failed"
|
|
|
|
|
|
def test_matmul_codegen():
|
|
assert_matmul_codegen(M=1024, N=1024, K=1024, block_M=128, block_N=128, block_K=32)
|
|
|
|
|
|
def test_matmul_compile():
|
|
|
|
def matmul_jit_test(M, N, K, block_M, block_N, block_K, dtype="float16", accum_dtype="float"):
|
|
# a simple kernel just for jit test
|
|
@T.prim_func
|
|
def matmul(
|
|
A: T.Tensor((M, K), dtype),
|
|
B: T.Tensor((K, N), dtype),
|
|
C: T.Tensor((M, N), dtype),
|
|
):
|
|
with T.Kernel(T.ceildiv(N, block_N), T.ceildiv(M, block_M), is_cpu=True) as (bx, by):
|
|
A_local = T.alloc_local((block_M, block_K), dtype)
|
|
B_local = T.alloc_local((block_K, block_N), dtype)
|
|
C_local = T.alloc_local((block_M, block_N), accum_dtype)
|
|
|
|
for p in T.serial(block_M):
|
|
for w in T.serial(block_N):
|
|
C_local[p, w] = 0
|
|
for ko in T.serial(K // block_K):
|
|
for i in T.serial(block_M):
|
|
for k in T.serial(block_K):
|
|
A_local[i, k] = A[by * block_M + i, ko * block_K + k]
|
|
|
|
for k in T.serial(block_K):
|
|
for j in T.serial(block_N):
|
|
B_local[k, j] = B[ko * block_K + k, bx * block_N + j]
|
|
|
|
for i in T.serial(block_M):
|
|
for j in T.serial(block_N):
|
|
for k in T.serial(block_K):
|
|
C_local[i, j] += A_local[i, k] * B_local[k, j]
|
|
|
|
for i in T.serial(block_M):
|
|
for j in T.serial(block_N):
|
|
C[by * block_M + i, bx * block_N + j] = C_local[i, j]
|
|
|
|
return matmul
|
|
|
|
M, N, K = 1024, 512, 512
|
|
block_M, block_N, block_K = M // 4, N // 4, K // 4
|
|
cpu_func = matmul_jit_test(M, N, K, block_M, block_N, block_K)
|
|
complied_fun = tilelang.compile(cpu_func, -1, execution_backend="ctypes", target="c")
|
|
|
|
in_dtype = "float16"
|
|
A = torch.randn(M, K, dtype=torch.__getattribute__(in_dtype))
|
|
B = torch.randn(K, N, dtype=torch.__getattribute__(in_dtype))
|
|
|
|
C = complied_fun(A, B)
|
|
C_torch = torch.matmul(A, B)
|
|
|
|
tilelang.testing.torch_assert_close(C, C_torch, atol=1e-2, rtol=1e-2, max_mismatched_ratio=0.05)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
tilelang.testing.main()
|