wehub-resource-sync
75 lines
2.5 KiB
Python
75 lines
2.5 KiB
Python
# Licensed to the Apache Software Foundation (ASF) under one
|
|
# or more contributor license agreements. See the NOTICE file
|
|
# distributed with this work for additional information
|
|
# regarding copyright ownership. The ASF licenses this file
|
|
# to you under the Apache License, Version 2.0 (the
|
|
# "License"); you may not use this file except in compliance
|
|
# with the License. You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing,
|
|
# software distributed under the License is distributed on an
|
|
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
|
|
# KIND, either express or implied. See the License for the
|
|
# specific language governing permissions and limitations
|
|
# under the License.
|
|
import numpy as np
|
|
import pytest
|
|
|
|
import tvm
|
|
import tvm.testing
|
|
from tvm.script import tirx as T
|
|
from tvm.testing import env
|
|
|
|
|
|
@T.prim_func(s_tir=True)
|
|
def vector_add(A: T.Buffer((16), "float32"), B: T.Buffer((32), "float32")) -> None:
|
|
T.func_attr({"global_symbol": "default_function", "tirx.noalias": True})
|
|
bx = T.env_thread("blockIdx.x")
|
|
tx = T.env_thread("threadIdx.x")
|
|
T.launch_thread(bx, 1)
|
|
T.launch_thread(tx, 32)
|
|
with T.sblock():
|
|
A_local = T.sblock_alloc_buffer((32), "float32", scope="local")
|
|
|
|
with T.sblock():
|
|
T.reads(A[0:16])
|
|
T.writes(A_local[0:32])
|
|
A_local[tx] = T.if_then_else(tx % 2 == 0, A[tx // 2], T.float32(0), dtype="float32")
|
|
B[tx] = A_local[tx] + 1.0
|
|
|
|
|
|
@pytest.mark.gpu
|
|
@pytest.mark.skipif(not env.has_cuda(), reason="need cuda")
|
|
def test_inject_ptx_intrin():
|
|
f = vector_add
|
|
arch = tvm.support.nvcc.get_target_compute_version()
|
|
major, _ = tvm.support.nvcc.parse_compute_version(arch)
|
|
if major < 8:
|
|
# Require at least SM80
|
|
return
|
|
with tvm.transform.PassContext(config={"tirx.ptx.ldg32": True}):
|
|
mod = tvm.compile(f, target="cuda")
|
|
A_np = np.random.rand(16).astype("float32")
|
|
B_np = np.zeros(32).astype("float32")
|
|
C_np = np.zeros(32).astype("float32")
|
|
|
|
for i in range(32):
|
|
if i % 2 == 0:
|
|
C_np[i] = A_np[i // 2]
|
|
C_np[i] += 1.0
|
|
|
|
def run_and_check():
|
|
dev = tvm.cuda(0)
|
|
A_nd = tvm.runtime.tensor(A_np, device=dev)
|
|
B_nd = tvm.runtime.tensor(B_np, device=dev)
|
|
mod(A_nd, B_nd)
|
|
tvm.testing.assert_allclose(B_nd.numpy(), C_np)
|
|
|
|
tvm.testing.run_with_gpu_lock(run_and_check)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
test_inject_ptx_intrin()
|