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
|
from scipy import optimize
from scipy.sparse import linalg
import numpy
import parse_sdf
import design_pip_classes
import pip_classes
import sys
import math
import json
def solve_pip_delays(ncl, sdf, debug=False):
path_pip_classes, wire_fanout = design_pip_classes.get_equations(ncl)
sdf_data = parse_sdf.parse_sdf_file(sdf)
top_ic = None
for cell in sdf_data.cells:
if len(sdf_data.cells[cell].interconnect) != 0:
top_ic = sdf_data.cells[cell].interconnect
break
assert top_ic is not None
variables = {}
for path in sorted(path_pip_classes.values()):
for wire, pipclass in path:
if (pipclass, "delay") not in variables and not pip_classes.force_zero_delay_pip(pipclass):
vid = len(variables)
variables[pipclass, "delay"] = vid
if not pip_classes.force_zero_fanout_pip(pipclass):
if (pipclass, "fanout") not in variables:
vid = len(variables)
variables[pipclass, "fanout"] = vid
kfid = len(variables)
data = {}
corners = "min", "typ", "max"
i = 0
A = numpy.zeros((len(path_pip_classes), len(variables)))
for arc, path in sorted(path_pip_classes.items()):
for wire, pipclass in path:
if not pip_classes.force_zero_delay_pip(pipclass):
A[i, variables[pipclass, "delay"]] += 1
if not pip_classes.force_zero_fanout_pip(pipclass):
A[i, variables[pipclass, "fanout"]] += wire_fanout[wire, pipclass]
if pipclass not in data:
data[pipclass] = {
"delay": [0, 0, 0],
"fanout": [0, 0, 0],
}
i += 1
for corner in corners:
b = numpy.zeros((len(path_pip_classes), ))
i = 0
for arc, path in sorted(path_pip_classes.items()):
src, dest = arc
srcname = "{}/{}".format(src[0].replace('/', '\\/').replace('.', '\\.'), src[1])
destname = "{}/{}".format(dest[0].replace('/', '\\/').replace('.', '\\.'), dest[1])
b[i] = getattr(top_ic[srcname, destname].rising, corner + "v")
i += 1
print("Starting least squares solver!")
x, rnorm = optimize.nnls(A, b)
for var, j in sorted(variables.items()):
pipclass, vartype = var
data[pipclass][vartype][corners.index(corner)] = x[j]
if debug:
error = numpy.matmul(A, x) - b
i = 0
worst = 0
sqsum = 0
for arc, path in sorted(path_pip_classes.items()):
src, dest = arc
rel_error = error[i] / b[i]
print("error {}.{} -> {}.{} = {:.01f} ps ({:.01f} %)".format(src[0], src[1], dest[0], dest[1], error[i], rel_error * 100))
worst = max(worst, abs(error[i]))
sqsum += error[i]**2
i = i + 1
with open("error.csv", "w") as f:
print("actual, percent", file=f)
for i in range(len(path_pip_classes)):
rel_error = error[i] / b[i]
print("{}, {}".format(b[i], rel_error * 100), file=f)
for var, j in sorted(variables.items()):
print("{}_{} = {:.01f} ps".format(var[0], var[1], x[j]))
print("Error: {:.01f} ps max, {:.01f} ps RMS".format(worst, math.sqrt(sqsum / len(path_pip_classes))))
return data
def main():
data = solve_pip_delays(sys.argv[1], sys.argv[2], debug=True)
with open("out.json", "w") as f:
json.dump(data, f, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
|
