mirror of
https://github.com/tile-ai/tilelang.git
synced 2026-10-22 06:50:25 +08:00
Lei Wang
c750fb8a98
* [Enhancement] Update examples and tests for improved type handling and functionality
- Enhanced various example scripts to support new data types and improve compatibility with PyTorch.
- Updated tests across multiple modules to ensure correct functionality with the latest changes in type handling.
- Refactored code in examples to streamline operations and improve clarity, particularly in tensor operations and memory management.
- Added comprehensive tests for new features and fixed existing issues related to type conversions and buffer handling.
* [Refactor] Update accumulation data type to float32 across examples
- Changed accumulation data type from "float" to T.float32 in multiple example scripts to ensure consistency and improve numerical stability.
- This update affects various modules including flash attention, GEMM analysis, convolution, and deepseek MLA examples, enhancing type handling across the board.
* [Refactor] Standardize data type usage across benchmark scripts
- Updated data type definitions in benchmark scripts to use T.float16 and T.float32 consistently, enhancing clarity and type handling.
- Adjusted dtype assignments in matmul functions and configuration setups to align with the new standard.
- Improved overall code consistency and maintainability by ensuring uniform data type usage across various modules.
* [Refactor] Standardize data type usage in templates and scripts
- Updated data type definitions in various templates and scripts to use string representations (e.g., "float16", "int32") instead of T.float16 and T.int32 for improved consistency and clarity.
- Enhanced overall code maintainability by ensuring uniform data type usage across multiple modules, including convolution, elementwise operations, and matrix multiplication templates.
- This change aims to streamline type handling and improve compatibility with existing workflows.
* [Refactor] Standardize data type usage in examples and benchmarks
- Updated data type definitions in various example and benchmark scripts to use T.float16 and T.int32 consistently, enhancing clarity and maintainability.
- Adjusted dtype assignments in kernel functions and configuration setups to align with the new standard.
- Improved overall code consistency by ensuring uniform data type usage across multiple modules, including attention mechanisms, matrix multiplication, and GEMM examples.
* [Refactor] Import dtypes from language.v2 module
- Added import statement for dtypes from the language.v2 module to enhance type handling and maintain consistency across the codebase.
- This change aims to streamline data type management and improve overall code clarity.
* fix
* [Refactor] Standardize data type usage across scripts
- Updated data type definitions in various scripts to use string representations (e.g., "float16", "int8") instead of T.float16 and T.int8 for improved consistency and clarity.
- Adjusted dtype assignments in functions and configuration setups to align with the new standard, enhancing overall code maintainability.
- This change affects multiple modules, including benchmark and attention mechanisms, ensuring uniform data type usage throughout the codebase.
* [Refactor] Update data type handling for consistency and clarity
- Changed string representations of data types in the Hint class to use T.float32 and T.int32 for improved consistency.
- Added new data types "int4" and "int16" to the dtypes module, enhancing type support across the codebase.
- Updated function signatures and assertions in the lop3 and mxfp modules to utilize the new data types, ensuring uniformity in type handling.
- This refactor aims to streamline data type management and improve overall code clarity and maintainability.
* [Enhancement] Improve data type handling and error messaging
- Introduced a mapping for canonical data types to their display strings, enhancing clarity in type representation.
- Updated the dtype creation logic to utilize the new mapping, ensuring more intuitive handling of string inputs.
- Refined error messages in the lop3 module to provide clearer feedback on invalid source formats, improving debugging and user experience.
* [Fix] Correct boolean flag in GEMM SP test case
- Updated the boolean flag in the test_gemm_sp_sm90 function to ensure proper functionality in the test case.
- This change enhances the accuracy of the test and aligns it with expected behavior for the GEMM SP implementation.
* [Refactor] Standardize data type usage across scripts
- Updated data type definitions in various scripts to use T.float16 and T.bfloat16 consistently, enhancing clarity and maintainability.
- Adjusted dtype assignments in function signatures and argument parsing to align with the new standard, ensuring uniform data type usage throughout the codebase.
- This change affects multiple modules, including benchmarks and examples, improving overall code consistency and readability.
* [Refactor] Standardize data type usage in various modules
- Updated data type assignments in multiple scripts to utilize T.float32, T.int8, and T.int32 consistently, enhancing clarity and maintainability.
- Adjusted function signatures and parameter types across benchmarks, examples, and tests to align with the new standard, ensuring uniform data type usage throughout the codebase.
- This change improves overall code consistency and readability, impacting modules related to matrix multiplication, GEMM, and tensor operations.
* [Refactor] Update argument parsing for data types in benchmarks
- Changed argument parsing for data types in benchmark_matmul_intrinsic.py and benchmark_matmul_sp.py to use string representations ("float16", "int8", "float") instead of T.float16 and T.float.
- This update enhances consistency in data type handling across benchmark scripts, improving clarity and maintainability.
* [Refactor] Update data type handling in benchmark and example scripts
- Changed data type arguments in benchmark and example scripts to use string representations ("float16") instead of T.float16 for improved consistency.
- Updated function signatures and argument parsing to align with the new standard, enhancing clarity and maintainability across the codebase.
- This change affects multiple modules related to attention mechanisms and tensor operations, ensuring uniform data type usage throughout the examples.
* [Refactor] Fix data type conversion in multiple scripts
- Corrected the usage of the data type conversion method from dtype..as_torch() to dtype.as_torch() across various benchmark and example scripts.
- This change enhances consistency in data type handling and improves code readability, impacting modules related to attention mechanisms and tensor operations.
* [Refactor] Update float8 data type usage across multiple scripts
- Changed instances of T.float8_e4m3 to T.float8_e4m3fn in various benchmark, example, and test scripts to ensure consistency in data type handling.
- This update enhances clarity and maintainability across the codebase, particularly in modules related to matrix multiplication and tensor operations.
* [Refactor] Enhance float8 data type handling in CUDA code generation
- Updated the handling of float8 data types in the CUDA code generation to include additional float8 variants, improving type conversion logic.
- Adjusted conditions to ensure proper type checks for float8 conversions, enhancing clarity and maintainability in the codebase.
- Modified layout inference to streamline float8 type checks, ensuring consistency across the implementation.
- This change impacts modules related to matrix operations and CUDA code generation, improving overall type handling and conversion accuracy.
* [Refactor] Streamline float8 data type handling in CUDA and related modules
- Enhanced float8 data type handling in CUDA code generation by refining type conversion logic and ensuring consistent type checks.
- Updated layout inference for float8 types to improve clarity and maintainability across the implementation.
- This change impacts modules related to matrix operations and CUDA code generation, improving overall type handling and conversion accuracy.
* [Refactor] Remove unnecessary cache disabling in float8 example script
- Eliminated the call to tilelang.disable_cache() in example_group_per_split_token_cast_to_fp8.py to streamline the code.
- This change enhances clarity and maintainability of the example script without affecting its functionality.
* [Refactor] Update data type usage in debug print tests
- Changed the argument for dtype in the test_debug_print_buffer function from a string representation to the corresponding T.bool type.
- This update enhances consistency in data type handling within the test suite, improving clarity and maintainability.
* lint fix
* Update function parameter types from `str` to `T.dtype` for improved type safety in attention sink and related examples
* Refactor `gemv_alloc_reducer` function signature for improved readability by formatting parameters across multiple lines.
366 lines
12 KiB
Python
366 lines
12 KiB
Python
import tilelang.testing
|
|
import tilelang.language as T
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_add_program(K, M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_add(A: T.Tensor((K, M, N), dtype), B: T.Tensor((M, N), dtype)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), K, threads=32) as (bx, by, bz):
|
|
A_shared = T.alloc_shared((block_M, block_N), dtype)
|
|
|
|
T.copy(A[bz, bx * block_M : (bx + 1) * block_M, by * block_N : (by + 1) * block_N], A_shared)
|
|
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
T.atomic_add(B[bx * block_M + i, by * block_N + j], A_shared[i, j])
|
|
|
|
return atomic_add
|
|
|
|
|
|
def run_atomic_add(K, M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = atomic_add_program(K, M, N, block_M, block_N, dtype=dtype)
|
|
import torch
|
|
|
|
def ref_program(A, B):
|
|
for k in range(K):
|
|
for i in range(M):
|
|
for j in range(N):
|
|
B[i, j] += A[k, i, j]
|
|
|
|
A = torch.randn(K, M, N, dtype=getattr(torch, dtype)).cuda()
|
|
B = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
ref_B = B.clone()
|
|
ref_program(A, ref_B)
|
|
kernel(A, B)
|
|
torch.testing.assert_close(B, ref_B, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
@tilelang.jit
|
|
def tile_atomic_add_program(K, M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_add(A: T.Tensor((K, M, N), dtype), B: T.Tensor((M, N), dtype)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), K, threads=32) as (bx, by, bz):
|
|
A_shared = T.alloc_shared((block_M, block_N), dtype)
|
|
|
|
T.copy(A[bz, bx * block_M : (bx + 1) * block_M, by * block_N : (by + 1) * block_N], A_shared)
|
|
|
|
T.atomic_add(B[bx * block_M, by * block_N], A_shared)
|
|
|
|
return atomic_add
|
|
|
|
|
|
def run_tile_atomic_add(K, M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = tile_atomic_add_program(K, M, N, block_M, block_N, dtype=dtype)
|
|
print(kernel.get_kernel_source())
|
|
import torch
|
|
|
|
def ref_program(A, B):
|
|
for k in range(K):
|
|
for i in range(M):
|
|
for j in range(N):
|
|
B[i, j] += A[k, i, j]
|
|
|
|
A = torch.randn(K, M, N, dtype=getattr(torch, dtype)).cuda()
|
|
B = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
ref_B = B.clone()
|
|
ref_program(A, ref_B)
|
|
kernel(A, B)
|
|
print(B)
|
|
print(ref_B)
|
|
torch.testing.assert_close(B, ref_B, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_max_program(K, M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_max(A: T.Tensor((K, M, N), dtype), B: T.Tensor((M, N), dtype)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), K, threads=32) as (bx, by, bz):
|
|
A_shared = T.alloc_shared((block_M, block_N), dtype)
|
|
|
|
T.copy(A[bz, bx * block_M : (bx + 1) * block_M, by * block_N : (by + 1) * block_N], A_shared)
|
|
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
T.atomic_max(B[bx * block_M + i, by * block_N + j], A_shared[i, j])
|
|
|
|
return atomic_max
|
|
|
|
|
|
def run_atomic_max(K, M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = atomic_max_program(K, M, N, block_M, block_N, dtype=dtype)
|
|
import torch
|
|
|
|
def ref_program(A, B):
|
|
for k in range(K):
|
|
for i in range(M):
|
|
for j in range(N):
|
|
B[i, j] = max(B[i, j], A[k, i, j])
|
|
|
|
A = torch.randn(K, M, N, dtype=getattr(torch, dtype)).cuda()
|
|
B = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
ref_B = B.clone()
|
|
ref_program(A, ref_B)
|
|
kernel(A, B)
|
|
torch.testing.assert_close(B, ref_B, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_min_program(K, M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_min(A: T.Tensor((K, M, N), dtype), B: T.Tensor((M, N), dtype)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), K, threads=32) as (bx, by, bz):
|
|
A_shared = T.alloc_shared((block_M, block_N), dtype)
|
|
|
|
T.copy(A[bz, bx * block_M : (bx + 1) * block_M, by * block_N : (by + 1) * block_N], A_shared)
|
|
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
T.atomic_min(B[bx * block_M + i, by * block_N + j], A_shared[i, j])
|
|
|
|
return atomic_min
|
|
|
|
|
|
def run_atomic_min(K, M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = atomic_min_program(K, M, N, block_M, block_N, dtype=dtype)
|
|
import torch
|
|
|
|
def ref_program(A, B):
|
|
for k in range(K):
|
|
for i in range(M):
|
|
for j in range(N):
|
|
B[i, j] = min(B[i, j], A[k, i, j])
|
|
|
|
A = torch.randn(K, M, N, dtype=getattr(torch, dtype)).cuda()
|
|
B = torch.full((M, N), float("inf"), dtype=getattr(torch, dtype)).cuda()
|
|
ref_B = B.clone()
|
|
ref_program(A, ref_B)
|
|
kernel(A, B)
|
|
torch.testing.assert_close(B, ref_B, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_load_store_program(M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_load_store(A: T.Tensor((M, N), dtype), B: T.Tensor((M, N), dtype)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), threads=32) as (bx, by):
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
idx_i = bx * block_M + i
|
|
idx_j = by * block_N + j
|
|
if idx_i < M and idx_j < N:
|
|
val = T.atomic_load(A[idx_i, idx_j])
|
|
T.atomic_store(B[idx_i, idx_j], val)
|
|
|
|
return atomic_load_store
|
|
|
|
|
|
def run_atomic_load_store(M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = atomic_load_store_program(M, N, block_M, block_N, dtype=dtype)
|
|
import torch
|
|
|
|
A = torch.randn(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
B = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
kernel(A, B)
|
|
torch.testing.assert_close(B, A, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_memory_order_program(K, M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_with_memory_order(A: T.Tensor((K, M, N), dtype), B: T.Tensor((M, N), dtype)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), K, threads=32) as (bx, by, bz):
|
|
A_shared = T.alloc_shared((block_M, block_N), dtype)
|
|
|
|
T.copy(A[bz, bx * block_M : (bx + 1) * block_M, by * block_N : (by + 1) * block_N], A_shared)
|
|
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
T.atomic_add(B[bx * block_M + i, by * block_N + j], A_shared[i, j], memory_order="relaxed")
|
|
|
|
return atomic_with_memory_order
|
|
|
|
|
|
def run_atomic_memory_order(K, M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = atomic_memory_order_program(K, M, N, block_M, block_N, dtype=dtype)
|
|
import torch
|
|
|
|
def ref_program(A, B):
|
|
for k in range(K):
|
|
for i in range(M):
|
|
for j in range(N):
|
|
B[i, j] += A[k, i, j]
|
|
|
|
A = torch.randn(K, M, N, dtype=getattr(torch, dtype)).cuda()
|
|
B = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
ref_B = B.clone()
|
|
ref_program(A, ref_B)
|
|
kernel(A, B)
|
|
torch.testing.assert_close(B, ref_B, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_addx2_program(M, N, block_M, block_N):
|
|
@T.prim_func
|
|
def atomic_addx2(A: T.Tensor((M, N), T.float16), B: T.Tensor((M, N), T.float16)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), threads=32) as (bx, by):
|
|
for i, j in T.Parallel(block_M, block_N // 2):
|
|
idx_i = bx * block_M + i
|
|
idx_j = by * block_N + j * 2
|
|
T.atomic_addx2(B[idx_i, idx_j], A[idx_i, idx_j])
|
|
|
|
return atomic_addx2
|
|
|
|
|
|
def run_atomic_addx2(M, N, block_M, block_N):
|
|
kernel = atomic_addx2_program(M, N, block_M, block_N)
|
|
import torch
|
|
|
|
A = torch.randn(M, N, dtype=torch.float16).cuda()
|
|
B = torch.zeros(M, N, dtype=torch.float16).cuda()
|
|
ref_B = B.clone()
|
|
|
|
for i in range(M):
|
|
for j in range(0, N - 1, 2):
|
|
ref_B[i, j] += A[i, j]
|
|
ref_B[i, j + 1] += A[i, j + 1]
|
|
kernel(A, B)
|
|
torch.testing.assert_close(B, ref_B, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
def test_atomic_add():
|
|
run_atomic_add(8, 128, 128, 32, 32)
|
|
|
|
|
|
def test_atomic_max():
|
|
run_atomic_max(4, 64, 64, 16, 16)
|
|
|
|
|
|
def test_atomic_min():
|
|
run_atomic_min(4, 64, 64, 16, 16)
|
|
|
|
|
|
def test_atomic_load_store():
|
|
run_atomic_load_store(64, 64, 16, 16)
|
|
|
|
|
|
def test_atomic_memory_order():
|
|
run_atomic_memory_order(4, 64, 64, 16, 16)
|
|
|
|
|
|
def test_atomic_addx2():
|
|
run_atomic_addx2(32, 64, 8, 16)
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_different_memory_orders_program(M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_different_orders(
|
|
A: T.Tensor((M, N), dtype), B: T.Tensor((M, N), dtype), C: T.Tensor((M, N), dtype), D: T.Tensor((M, N), dtype)
|
|
):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), threads=32) as (bx, by):
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
idx_i = bx * block_M + i
|
|
idx_j = by * block_N + j
|
|
if idx_i < M and idx_j < N:
|
|
val = A[idx_i, idx_j]
|
|
T.atomic_add(B[idx_i, idx_j], val, memory_order="release")
|
|
T.atomic_max(C[idx_i, idx_j], val, memory_order="relaxed")
|
|
T.atomic_min(D[idx_i, idx_j], val, memory_order="relaxed")
|
|
|
|
return atomic_different_orders
|
|
|
|
|
|
def run_atomic_different_memory_orders(M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = atomic_different_memory_orders_program(M, N, block_M, block_N, dtype=dtype)
|
|
import torch
|
|
|
|
A = torch.randn(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
B = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
C = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
D = torch.full((M, N), float("inf"), dtype=getattr(torch, dtype)).cuda()
|
|
|
|
kernel(A, B, C, D)
|
|
|
|
torch.testing.assert_close(B, A, atol=1e-3, rtol=1e-3)
|
|
torch.testing.assert_close(C, torch.maximum(torch.zeros_like(A), A))
|
|
torch.testing.assert_close(D, torch.minimum(torch.full_like(A, float("inf")), A))
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_addx4_program(M, N, block_M, block_N):
|
|
@T.prim_func
|
|
def atomic_addx4(A: T.Tensor((M, N), T.float32), B: T.Tensor((M, N), T.float32)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), threads=32) as (bx, by):
|
|
for i, j in T.Parallel(block_M, block_N // 4):
|
|
idx_i = bx * block_M + i
|
|
idx_j = by * block_N + j * 4
|
|
T.atomic_addx4(B[idx_i, idx_j], A[idx_i, idx_j])
|
|
|
|
return atomic_addx4
|
|
|
|
|
|
def run_atomic_addx4(M, N, block_M, block_N):
|
|
kernel = atomic_addx4_program(M, N, block_M, block_N)
|
|
import torch
|
|
|
|
A = torch.randn(M, N, dtype=torch.float32).cuda()
|
|
B = torch.zeros(M, N, dtype=torch.float32).cuda()
|
|
ref_B = B.clone()
|
|
|
|
for i in range(M):
|
|
for j in range(0, N - 3, 4):
|
|
ref_B[i, j] += A[i, j]
|
|
ref_B[i, j + 1] += A[i, j + 1]
|
|
ref_B[i, j + 2] += A[i, j + 2]
|
|
ref_B[i, j + 3] += A[i, j + 3]
|
|
|
|
kernel(A, B)
|
|
torch.testing.assert_close(B, ref_B, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
@tilelang.jit
|
|
def atomic_return_prev_program(M, N, block_M, block_N, dtype=T.float32):
|
|
@T.prim_func
|
|
def atomic_with_return_prev(A: T.Tensor((M, N), dtype), B: T.Tensor((M, N), dtype), old_vals: T.Tensor((M, N), dtype)):
|
|
with T.Kernel(T.ceildiv(M, block_M), T.ceildiv(N, block_N), threads=32) as (bx, by):
|
|
for i, j in T.Parallel(block_M, block_N):
|
|
idx_i = bx * block_M + i
|
|
idx_j = by * block_N + j
|
|
if idx_i < M and idx_j < N:
|
|
old_vals[idx_i, idx_j] = T.atomic_add(B[idx_i, idx_j], A[idx_i, idx_j], return_prev=True)
|
|
|
|
return atomic_with_return_prev
|
|
|
|
|
|
def run_atomic_return_prev(M, N, block_M, block_N, dtype=T.float32):
|
|
kernel = atomic_return_prev_program(M, N, block_M, block_N, dtype=dtype)
|
|
import torch
|
|
|
|
A = torch.ones(M, N, dtype=getattr(torch, dtype)).cuda() * 5.0
|
|
B = torch.ones(M, N, dtype=getattr(torch, dtype)).cuda() * 2.0
|
|
old_vals = torch.zeros(M, N, dtype=getattr(torch, dtype)).cuda()
|
|
|
|
initial_B = B.clone()
|
|
kernel(A, B, old_vals)
|
|
|
|
torch.testing.assert_close(old_vals, initial_B, atol=1e-3, rtol=1e-3)
|
|
torch.testing.assert_close(B, initial_B + A, atol=1e-3, rtol=1e-3)
|
|
|
|
|
|
def test_atomic_different_memory_orders():
|
|
run_atomic_different_memory_orders(32, 32, 8, 8, dtype=T.float32)
|
|
run_atomic_different_memory_orders(32, 32, 8, 8, dtype=T.float16)
|
|
run_atomic_different_memory_orders(32, 32, 8, 8, dtype=T.bfloat16)
|
|
|
|
|
|
def test_atomic_addx4():
|
|
run_atomic_addx4(16, 64, 4, 4)
|
|
|
|
|
|
def test_atomic_return_prev():
|
|
run_atomic_return_prev(32, 32, 8, 8)
|
|
|
|
|
|
def test_tile_atomic_add():
|
|
run_tile_atomic_add(8, 128, 128, 32, 32)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
tilelang.testing.main()
|