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
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
|
# Dissertation Visualisations
```elixir
Mix.install([
{:tucan, "~> 0.3.0"},
{:kino_vega_lite, "~> 0.1.8"},
{:json, "~> 1.4"},
{:explorer, "~> 0.8.0"},
{:kino_explorer, "~> 0.1.11"}
])
```
## Variables
```elixir
require Explorer.DataFrame
alias Explorer.DataFrame, as: DF
job_id = "1146"
job_dir = Path.expand(~c"./" ++ job_id) |> Path.absname()
sections_dir = Path.join(job_dir, "sections")
criterion_dir = Path.join(job_dir, "criterion")
```
## Read cost model data
```elixir
{:ok, cost_model_files} = File.ls(sections_dir)
cost_model_files =
cost_model_files
|> Enum.filter(fn name -> String.contains?(name, "cost-model") end)
|> Enum.filter(fn name -> String.contains?(name, "vec--Vec") end)
|> Enum.map(fn fname -> Path.join(sections_dir, fname) |> Path.absname() end)
cost_model_files
```
```elixir
defmodule CostModel do
defstruct impl: nil, ops: []
end
defmodule OpCostModel do
defstruct op: nil, x0: nil, x1: nil, x2: nil, x3: nil, nmrse: nil
end
defmodule Parse do
def cost_model_row(row) do
[name, x0, x1, x2, x3, nmrse] = String.split(row, " & ")
[nmrse | _] = String.split(nmrse)
%OpCostModel{
op: name,
x0: String.to_float(x0),
x1: String.to_float(x1),
x2: String.to_float(x2),
x3: String.to_float(x3),
nmrse: String.to_float(nmrse)
}
end
def cost_model_output(fname) do
{:ok, contents} = File.read(fname)
[_, table | _] = String.split(contents, "line\n")
[rows | _] = String.split(table, "\n\\end")
rows = String.split(rows, "\n")
[_, impl] = String.split(fname, "cost-model-")
impl = String.replace(impl, "-", ":")
%CostModel{
impl: impl,
ops: rows |> Enum.map(&cost_model_row/1)
}
end
end
```
<!-- livebook:{"reevaluate_automatically":true} -->
```elixir
cost_models = cost_model_files |> Enum.map(&Parse.cost_model_output/1)
```
## Cost model exploratory plots
```elixir
defmodule PlotCostModel do
@startn 0
@endn 20000
@resolution 100
def gen_ts(ns, %OpCostModel{x0: x0, x1: x1, x2: x2, x3: x3}) do
Enum.map(ns, fn n -> %{n: n, t: x0 + n * x1 + n * n * x2 + n * n * n * x3} end)
end
def points_op(op) do
ns = @startn..@endn//@resolution
gen_ts(ns, op)
|> Enum.map(fn data -> Map.put(data, :name, op.op) end)
end
def points_model(model) do
model.ops
|> Enum.map(fn op ->
points_op(op)
|> Enum.map(fn point -> Map.put(point, :impl, model.impl) end)
end)
|> List.flatten()
end
def points(models) do
models
|> Enum.map(&points_model/1)
|> List.flatten()
end
def plot_op(op) do
points = points_op(op)
Tucan.lineplot(points, "n", "t", title: op.op)
end
def plot_all_ops(%CostModel{impl: impl, ops: ops}) do
ops
|> Enum.map(&plot_op/1)
|> Tucan.concat(columns: 3)
|> Tucan.set_title(impl)
end
end
```
<!-- livebook:{"reevaluate_automatically":true} -->
```elixir
cost_models
|> Enum.map(&PlotCostModel.plot_all_ops/1)
|> Tucan.vconcat()
```
## Read benchmark data
```elixir
benchmarks =
File.ls!(criterion_dir)
|> Enum.map(fn name ->
File.ls!(Path.join(criterion_dir, name))
|> Enum.map(fn p -> %{bench: name, subbench: p} end)
end)
|> List.flatten()
|> Enum.map(fn %{bench: bench, subbench: subbench} ->
File.ls!(Path.join([criterion_dir, bench, subbench]))
|> Enum.filter(fn x -> String.contains?(x, "Mapping2D") end)
|> Enum.map(fn x -> Path.join([criterion_dir, bench, subbench, x]) end)
|> Enum.map(fn dir ->
raw_results =
Path.join(dir, "estimates.json")
|> File.read!()
|> JSON.decode!()
%{
bench: bench,
subbench: subbench,
bench_id: bench <> "/" <> subbench,
using:
Regex.scan(~r/\"(\w*)\", ([\w:]*)/, Path.basename(dir))
|> Enum.map(fn [_, ctn, impl] -> %{ctn: ctn, impl: impl} end),
dir: dir,
lower_bound: raw_results["mean"]["confidence_interval"]["lower_bound"] / 10 ** 9,
mean: raw_results["mean"]["point_estimate"] / 10 ** 9,
upper_bound: raw_results["mean"]["confidence_interval"]["upper_bound"] / 10 ** 9
}
end)
end)
|> List.flatten()
|> DF.new()
```
```elixir
benchmarks
|> DF.distinct([:bench_id])
```
<!-- livebook:{"reevaluate_automatically":true} -->
```elixir
# Difference in execution time between worst and best selection
Tucan.bar(
benchmarks
|> DF.group_by("bench_id")
|> DF.summarise(range: max(mean) - min(mean)),
"bench_id",
"range",
orient: :horizontal,
clip: true
)
|> Tucan.Scale.set_x_domain(0, 10)
```
|
