-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathtune_bgemm.py
120 lines (88 loc) · 3.96 KB
/
tune_bgemm.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
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
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
import logging
import numpy as np
import sys
import topi.nn
import tvm
from tvm import autotvm
def bitpack(Anp, dtype="uint32"):
A = tvm.placeholder(Anp.shape, name="A", dtype=str(Anp.dtype))
B = topi.nn.bitpack(A, 1, 1, 1, dtype)
s = tvm.create_schedule(B.op)
fn = tvm.build(s, [A, B], "llvm -mcpu=core-avx2")
ctx = tvm.context("llvm -mcpu=core-avx2")
Bshape = topi.util.get_const_tuple(B.shape)
b = tvm.nd.array(np.zeros(Bshape, dtype=dtype), ctx)
fn(tvm.nd.array(Anp, ctx), b)
return b.asnumpy() # .reshape(b.shape[0], b.shape[2])
@autotvm.template
def bgemm_topi(Y, X, K, dtype="uint64"):
DB = 1
WB = 1
out_dtype = dtype
data_packed = tvm.placeholder((Y, DB, K), dtype=dtype, name="A")
weight_packed = tvm.placeholder((X, WB, K), dtype=dtype, name="B")
oshape = (Y, X)
k = tvm.reduce_axis((0, K), name='k')
db = tvm.reduce_axis((0, DB), name='db')
wb = tvm.reduce_axis((0, WB), name='wb')
matmul = tvm.compute(oshape, lambda i, j: tvm.sum(
tvm.popcount(weight_packed[j, wb, k] & data_packed[i, db, k]).astype(out_dtype)
<< (db+wb).astype(out_dtype), axis=[wb, db, k]), tag='bitserial_dense')
s = tvm.create_schedule(matmul.op)
cfg = autotvm.get_config()
CC = s.cache_write(matmul, "global")
y, x = s[matmul].op.axis
yo, yi = cfg.define_split("tile_y", y, num_outputs=2, filter=lambda x: x.size[-1] <= 8)
xo, xi = cfg.define_split("tile_x", x, num_outputs=2, filter=lambda x: x.size[-1] <= 8)
yo, yi = cfg["tile_y"].apply(s, matmul, y)
xo, xi = cfg["tile_x"].apply(s, matmul, x)
s[matmul].reorder(yo, xo, yi, xi)
cfg.define_knob("compute_at_axis", [0, 1, 2])
if cfg["compute_at_axis"].val == 0:
s[CC].compute_at(s[matmul], xo)
elif cfg["compute_at_axis"].val == 1:
s[CC].compute_at(s[matmul], yi)
elif cfg["compute_at_axis"].val == 2:
s[CC].compute_at(s[matmul], xi)
yc, xc = s[CC].op.axis
wb, db, k = s[CC].op.reduce_axis
cfg.define_reorder("reorder_0", [k, yc, xc], policy="all")
cfg["reorder_0"].apply(s, CC, [k, yc, xc])
cfg.add_flop(2 * Y * X * K * int(dtype[4:]))
return s, [data_packed, weight_packed, matmul]
def bgemm_topi_tuned(Y, X, K, dtype="uint64"):
with autotvm.apply_history_best("matmul_wc.log"):
with tvm.target.create("llvm -mcpu=core-avx2"):
s, arg_bufs = bgemm_topi(Y, X, K, "uint64")
return tvm.build(s, arg_bufs), s, arg_bufs
def test_bgemm(name, bgemm, dtype, M=2, N=2, K=8192, verbose=True):
ctx = tvm.context("llvm -mcpu=core-avx2")
a = np.random.randint(0, 2, (M, K))
b = np.random.randint(0, 2, (N, K))
pa = bitpack(a, dtype=dtype)
pb = bitpack(b, dtype=dtype)
fn, s, bufs = bgemm(pa.shape[0], pb.shape[0], pa.shape[2], dtype)
print(tvm.lower(s, bufs, simple_mode=True))
c = tvm.nd.array(np.zeros((pa.shape[0], pb.shape[0]), dtype=dtype), ctx)
fn(tvm.nd.array(pa, ctx), tvm.nd.array(pb, ctx), c)
tvm.testing.assert_allclose(c.asnumpy(), a.dot(b.T))
evaluator = fn.time_evaluator(fn.entry_name, ctx, number=40)
time = evaluator(tvm.nd.array(pa, ctx), tvm.nd.array(pb, ctx), c).mean
print("{} {}/{}/{} GFLOPS: {}".format(
name,
M, N, K,
2 * a.shape[0] * a.shape[1] * b.shape[0] / time / 1e9))
Y, X, K = 32, 32, 1024//64
if True:
task = autotvm.task.create(bgemm_topi, args=(Y, X, K, "uint64"), target="llvm -mcpu=core-avx2")
print(task.config_space)
logging.getLogger('autotvm').setLevel(logging.DEBUG)
logging.getLogger('autotvm').addHandler(logging.StreamHandler(sys.stdout))
measure_option = autotvm.measure_option(
builder='local',
runner=autotvm.LocalRunner(number=40))
tuner = autotvm.tuner.GridSearchTuner(task)
tuner.tune(n_trial=288,
measure_option=measure_option,
callbacks=[autotvm.callback.log_to_file('matmul_wc.log')])
test_bgemm("bgemm_topi_tuned", bgemm_topi_tuned, "uint64", Y, X, K*64)