test_conv2d_hwcn_map.py 3.13 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
# 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.
17 18 19 20 21 22 23
"""Example code to do convolution."""
import os
import numpy as np
import scipy.signal
import tvm
from tvm.contrib import nvcc
import topi
24
from topi.util import get_const_tuple
25 26 27 28 29 30

TASK = "conv2d_hwcn_map"
USE_MANUAL_CODE = False

@tvm.register_func
def tvm_callback_cuda_compile(code):
31
    ptx = nvcc.compile_cuda(code, target="ptx")
32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
    return ptx

def write_code(code, fname):
    with open(fname, "w") as f:
        f.write(code)

@tvm.register_func
def tvm_callback_cuda_postproc(code):
    if not os.path.exists("perf"):
        os.mkdir("perf")
    write_code(code, "perf/%s_generated.cu" % TASK)
    if USE_MANUAL_CODE:
        code = open("perf/%s_manual.cu" % TASK).read()
    return code


def test_conv2d_hwcn_map():
    batch = 64
    in_channel = 128
    in_height = 16
    in_width = 16
    num_filter = 128
    kernel = 3
    stride = 2
    padding = 'SAME'

    A = tvm.placeholder((in_height, in_width, in_channel, batch), name='A')
    W = tvm.placeholder((kernel, kernel, in_channel, num_filter), name='W')
    B = topi.nn.conv2d_hwcn(A, W, stride, padding)
    C = topi.nn.relu(B)
62 63
    s1 = topi.cuda.schedule_conv2d_hwcn([B])
    s2 = topi.cuda.schedule_conv2d_hwcn([C])
64 65 66 67 68 69 70 71 72 73

    a_np = np.random.uniform(size=get_const_tuple(A.shape)).astype(A.dtype)
    w_np = np.random.uniform(size=get_const_tuple(W.shape)).astype(W.dtype)
    b_np = topi.testing.conv2d_hwcn_python(a_np, w_np, stride, padding)
    c_np = np.maximum(b_np, 0)

    def check_device(device):
        if not tvm.module.enabled(device):
            print("Skip because %s is not enabled" % device)
            return
74
        ctx = tvm.context(device, 0)
75 76 77 78
        a = tvm.nd.array(a_np, ctx)
        w = tvm.nd.array(w_np, ctx)
        b = tvm.nd.array(np.zeros(get_const_tuple(B.shape), dtype=B.dtype), ctx)
        c = tvm.nd.array(np.zeros(get_const_tuple(C.shape), dtype=C.dtype), ctx)
79
        with tvm.build_config(auto_unroll_max_step=128,
80
                              unroll_explicit=device == 'rocm'):
81 82
            func1 = tvm.build(s1, [A, W, B], device)
            func1(a, w, b)
83
            tvm.testing.assert_allclose(b.asnumpy(), b_np, rtol=1e-5)
84 85
            func2 = tvm.build(s2, [A, W, C], device)
            func2(a, w, c)
86
            tvm.testing.assert_allclose(c.asnumpy(), c_np, rtol=1e-5)
87

88
    for device in ['cuda', 'opencl', 'rocm']:
89 90 91 92 93
        check_device(device)


if __name__ == "__main__":
    test_conv2d_hwcn_map()