mirror of
https://github.com/tile-ai/tilelang.git
synced 2026-10-24 07:04:00 +08:00
SepcntandSiriusNEO
c2cabd9b3f
* [Feature] Add Windows support with cross-platform minimal diff * [Feature] Add ccache support for Windows * [Feature] Add parallel autotune support for Windows * [Feature] Add nvrtc support for Windows * [Bugfix] Fix tilelibrary support for Windows * [Bugfix] Fix alignment for Windows * [Feature] Add clang-cl support for Windows * [Bugfix] Fix numeric limit for Windows * [Bugfix] Fix rand for Windows * [Misc] Use dynamic test workload to avoid hang * [Bugfix] Fix TMA alignment for Windows and sm120 * [Bugfix] Add unix-like hex formatter for Windows * [CI] Add Windows wheel job * [Feature] Flexible CUDA loader * fix rebase --------- Co-authored-by: SiriusNEO <chaofan@deepseek.com>
114 lines
3.3 KiB
Python
114 lines
3.3 KiB
Python
import ctypes
|
|
import re
|
|
|
|
import pytest
|
|
|
|
import tilelang
|
|
import tilelang.language as T
|
|
import tilelang.testing
|
|
from tilelang import tvm
|
|
|
|
|
|
CUDA_SM90_TARGET = "cuda -arch=sm_90"
|
|
|
|
|
|
def _compile_tvm_ffi(func, *, target=CUDA_SM90_TARGET, target_host="c", pass_configs=None):
|
|
tilelang.disable_cache()
|
|
try:
|
|
return tilelang.compile(
|
|
func,
|
|
target=target,
|
|
target_host=target_host,
|
|
execution_backend="tvm_ffi",
|
|
pass_configs=pass_configs or {},
|
|
)
|
|
finally:
|
|
tilelang.enable_cache()
|
|
|
|
|
|
def _get_tma_create_tiled():
|
|
func = tvm.ffi.get_global_func("__tvm_tensormap_create_tiled", allow_missing=True)
|
|
if func is None:
|
|
pytest.skip("__tvm_tensormap_create_tiled is unavailable in this build.")
|
|
return func
|
|
|
|
|
|
@tilelang.testing.requires_cuda
|
|
def test_tma_runtime_validation_surfaces_invalid_argument_constraints():
|
|
create_tiled = _get_tma_create_tiled()
|
|
|
|
with pytest.raises(tvm.error.InternalError) as exc_info:
|
|
create_tiled(
|
|
ctypes.c_void_p(0x10),
|
|
7,
|
|
2,
|
|
ctypes.c_void_p(0x20),
|
|
1733,
|
|
4,
|
|
4,
|
|
6932,
|
|
64,
|
|
1,
|
|
1,
|
|
1,
|
|
0,
|
|
0,
|
|
2,
|
|
0,
|
|
)
|
|
|
|
message = str(exc_info.value)
|
|
assert "Invalid TMA descriptor arguments for __tvm_tensormap_create_tiled" in message
|
|
assert "tensorMap address must be 64-byte aligned" in message
|
|
assert "effective cuda globalStrides[0]" in message
|
|
assert "globalStridesRaw [4, 6932]" in message
|
|
assert "cudaGlobalStrides [6932]" in message
|
|
assert "format 7 (CU_TENSOR_MAP_DATA_TYPE_FLOAT32)" in message
|
|
|
|
|
|
@tilelang.testing.requires_cuda
|
|
def test_tma_host_codegen_aligns_tvm_ffi_stack_alloca_for_descriptor():
|
|
m, k = 16, 256
|
|
block_m, block_k = 4, 128
|
|
threads = 32
|
|
|
|
@T.prim_func
|
|
def tma_copy_2d_desc(
|
|
x: T.Tensor((m, k), T.float16),
|
|
y: T.Tensor((m, k), T.float16),
|
|
):
|
|
with T.Kernel(T.ceildiv(m, block_m), T.ceildiv(k, block_k), threads=threads) as (pid_m, pid_k):
|
|
x_shared = T.alloc_shared((block_m, block_k), dtype=T.float16)
|
|
mbar = T.alloc_barrier(1)
|
|
T.tma_copy(
|
|
x[
|
|
pid_m * block_m : (pid_m + 1) * block_m,
|
|
pid_k * block_k : (pid_k + 1) * block_k,
|
|
],
|
|
x_shared,
|
|
barrier=mbar,
|
|
)
|
|
T.barrier_arrive(mbar)
|
|
T.mbarrier_wait_parity(mbar, 0)
|
|
T.copy(
|
|
x_shared,
|
|
y[
|
|
pid_m * block_m : (pid_m + 1) * block_m,
|
|
pid_k * block_k : (pid_k + 1) * block_k,
|
|
],
|
|
)
|
|
|
|
kernel = _compile_tvm_ffi(
|
|
tma_copy_2d_desc,
|
|
pass_configs={tilelang.PassConfigKey.TL_ENABLE_FAST_MATH: False, tilelang.PassConfigKey.TL_DISABLE_WARP_SPECIALIZED: True},
|
|
)
|
|
|
|
source = kernel.get_host_source()
|
|
assert "__tvm_tensormap_create_tiled_packed" in source
|
|
assert re.search(r"TL_ALIGN\(128\) TVMFFIAny stack(_\d+)?\[", source)
|
|
assert not re.search(r"TL_ALIGN\(64\) TVMFFIAny stack(_\d+)?\[", source)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
tilelang.testing.main()
|