forked from google/periph
/
benchmark_fast_gpio_support.go
410 lines (388 loc) · 10.3 KB
/
benchmark_fast_gpio_support.go
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
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
// Copyright 2017 The Periph Authors. All rights reserved.
// Use of this source code is governed under the Apache License, Version 2.0
// that can be found in the LICENSE file.
// This file is expected to be copy-pasted in all GPIO benchmark smoke test that
// support FastOut(). The only delta shall be the package name.
package allwinnersmoketest
import (
"testing"
"periph.io/x/periph/conn/gpio"
)
// runFastGPIOBenchmark runs the standardized GPIO benchmark for this specific
// implementation plus the FastOut variants.
func (s *Benchmark) runFastGPIOBenchmark() {
s.runGPIOBenchmark()
if !s.short {
printBench("FastReadNaive ", testing.Benchmark(s.benchmarkFastReadNaive))
printBench("FastReadDiscard ", testing.Benchmark(s.benchmarkFastReadDiscard))
printBench("FastReadSliceLevel ", testing.Benchmark(s.benchmarkFastReadSliceLevel))
}
printBench("FastReadBitsLSBLoop ", testing.Benchmark(s.benchmarkFastReadBitsLSBLoop))
if !s.short {
printBench("FastReadBitsMSBLoop ", testing.Benchmark(s.benchmarkFastReadBitsMSBLoop))
}
printBench("FastReadBitsLSBUnrol", testing.Benchmark(s.benchmarkFastReadBitsLSBUnroll))
if !s.short {
printBench("FastReadBitsMSBUnrol", testing.Benchmark(s.benchmarkFastReadBitsMSBUnroll))
}
printBench("FastOutClock ", testing.Benchmark(s.benchmarkFastOutClock))
if !s.short {
printBench("FastOutSliceLevel ", testing.Benchmark(s.benchmarkFastOutSliceLevel))
}
printBench("FastOutBitsLSBLoop ", testing.Benchmark(s.benchmarkFastOutBitsLSBLoop))
if !s.short {
printBench("FastOutBitsMSBLoop ", testing.Benchmark(s.benchmarkFastOutBitsMSBLoop))
}
printBench("FastOutBitsLSBUnroll", testing.Benchmark(s.benchmarkFastOutBitsLSBUnroll))
if !s.short {
printBench("FastOutBitsMSBUnroll", testing.Benchmark(s.benchmarkFastOutBitsMSBUnroll))
printBench("FastOutInterface ", testing.Benchmark(s.benchmarkFastOutInterface))
printBench("FastOutMemberVariabl", testing.Benchmark(s.benchmarkFastOutMemberVariabl))
}
}
// FastRead
// benchmarkFastInNaive reads but ignores the data.
//
// This is an intentionally naive benchmark.
func (s *Benchmark) benchmarkFastReadNaive(b *testing.B) {
p := s.p
if err := p.In(s.pull, gpio.NoEdge); err != nil {
b.Fatal(err)
}
b.ResetTimer()
for i := 0; i < b.N; i++ {
p.FastRead()
}
b.StopTimer()
}
// benchmarkFastReadDiscard reads but discards the data except for the last
// value.
//
// It measures the maximum raw read speed, at least in theory.
func (s *Benchmark) benchmarkFastReadDiscard(b *testing.B) {
p := s.p
if err := p.In(s.pull, gpio.NoEdge); err != nil {
b.Fatal(err)
}
l := gpio.Low
b.ResetTimer()
for i := 0; i < b.N; i++ {
l = p.FastRead()
}
b.StopTimer()
b.Log(l)
}
// benchmarkFastReadSliceLevel reads into a []gpio.Level.
//
// This is 8x less space efficient that using bits packing, it measures if this
// has any performance impact versus bit packing.
func (s *Benchmark) benchmarkFastReadSliceLevel(b *testing.B) {
p := s.p
if err := p.In(s.pull, gpio.NoEdge); err != nil {
b.Fatal(err)
}
buf := make([]gpio.Level, b.N)
b.ResetTimer()
for i := range buf {
buf[i] = p.FastRead()
}
b.StopTimer()
}
// benchmarkFastReadBitsLSBLoop reads into a []byte using LSBF using a loop to
// iterate over the bits.
func (s *Benchmark) benchmarkFastReadBitsLSBLoop(b *testing.B) {
p := s.p
if err := p.In(s.pull, gpio.NoEdge); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
b.ResetTimer()
for i := 0; i < b.N; i++ {
if p.FastRead() {
mask := byte(1) << uint(i&7)
buf[i/8] |= mask
}
}
b.StopTimer()
}
// benchmarkFastReadBitsMSBLoop reads into a []byte using MSBF using a loop to
// iterate over the bits.
func (s *Benchmark) benchmarkFastReadBitsMSBLoop(b *testing.B) {
p := s.p
if err := p.In(s.pull, gpio.NoEdge); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
b.ResetTimer()
for i := 0; i < b.N; i++ {
if p.FastRead() {
mask := byte(1) << uint(7-(i&7))
buf[i/8] |= mask
}
}
b.StopTimer()
}
// benchmarkFastReadBitsLSBUnroll reads into a []byte using LSBF using an
// unrolled loop to iterate over the bits.
//
// It is expected to be slightly faster than benchmarkFastReadBitsLSBLoop.
func (s *Benchmark) benchmarkFastReadBitsLSBUnroll(b *testing.B) {
p := s.p
if err := p.In(s.pull, gpio.NoEdge); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
b.ResetTimer()
for i := range buf {
l := byte(0)
if p.FastRead() {
l |= 0x01
}
if p.FastRead() {
l |= 0x02
}
if p.FastRead() {
l |= 0x04
}
if p.FastRead() {
l |= 0x08
}
if p.FastRead() {
l |= 0x10
}
if p.FastRead() {
l |= 0x20
}
if p.FastRead() {
l |= 0x40
}
if p.FastRead() {
l |= 0x80
}
buf[i] = l
}
b.StopTimer()
}
// benchmarkFastReadBitsMSBUnroll reads into a []byte using MSBF using an
// unrolled loop to iterate over the bits.
//
// It is expected to be slightly faster than benchmarkFastReadBitsMSBLoop.
func (s *Benchmark) benchmarkFastReadBitsMSBUnroll(b *testing.B) {
p := s.p
if err := p.In(s.pull, gpio.NoEdge); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
b.ResetTimer()
for i := range buf {
l := byte(0)
if p.FastRead() {
l |= 0x80
}
if p.FastRead() {
l |= 0x40
}
if p.FastRead() {
l |= 0x20
}
if p.FastRead() {
l |= 0x10
}
if p.FastRead() {
l |= 0x08
}
if p.FastRead() {
l |= 0x04
}
if p.FastRead() {
l |= 0x02
}
if p.FastRead() {
l |= 0x01
}
buf[i] = l
}
b.StopTimer()
}
// FastOut
// benchmarkFastOutClock outputs an hardcoded clock.
//
// It measures maximum raw output performance when the bitstream is hardcoded.
func (s *Benchmark) benchmarkFastOutClock(b *testing.B) {
p := s.p
if err := p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
n := (b.N + 1) / 2
b.ResetTimer()
for i := 0; i < n; i++ {
p.FastOut(gpio.High)
p.FastOut(gpio.Low)
}
b.StopTimer()
}
// benchmarkFastOutSliceLevel writes into a []gpio.Level.
//
// This is 8x less space efficient that using bits packing, it measures if this
// has any performance impact versus bit packing.
func (s *Benchmark) benchmarkFastOutSliceLevel(b *testing.B) {
p := s.p
if err := p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
buf := make([]gpio.Level, b.N)
for i := 0; i < len(buf); i += 2 {
buf[i] = gpio.High
}
b.ResetTimer()
for _, l := range buf {
p.FastOut(l)
}
b.StopTimer()
}
// benchmarkFastOutBitsLSBLoop writes into a []byte using LSBF using a loop to
// iterate over the bits.
func (s *Benchmark) benchmarkFastOutBitsLSBLoop(b *testing.B) {
p := s.p
if err := p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
for i := 0; i < len(buf); i += 2 {
buf[i] = 0x55
}
b.ResetTimer()
for _, l := range buf {
for i := 0; i < 8; i++ {
mask := byte(1) << uint(i)
p.FastOut(gpio.Level(l&mask != 0))
}
}
b.StopTimer()
}
// benchmarkFastOutBitsMSBLoop writes into a []byte using MSBF using a loop to
// iterate over the bits.
func (s *Benchmark) benchmarkFastOutBitsMSBLoop(b *testing.B) {
p := s.p
if err := p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
for i := 0; i < len(buf); i += 2 {
buf[i] = 0xAA
}
b.ResetTimer()
for _, l := range buf {
for i := 7; i >= 0; i-- {
mask := byte(1) << uint(i)
p.FastOut(gpio.Level(l&mask != 0))
}
}
b.StopTimer()
}
// benchmarkFastOutBitsLSBUnroll writes into a []byte using LSBF using an
// unrolled loop to iterate over the bits.
//
// It is expected to be slightly faster than benchmarkFastOutBitsLSBLoop.
func (s *Benchmark) benchmarkFastOutBitsLSBUnroll(b *testing.B) {
p := s.p
if err := p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
for i := 0; i < len(buf); i += 2 {
buf[i] = 0x55
}
b.ResetTimer()
for _, l := range buf {
p.FastOut(gpio.Level(l&0x01 != 0))
p.FastOut(gpio.Level(l&0x02 != 0))
p.FastOut(gpio.Level(l&0x04 != 0))
p.FastOut(gpio.Level(l&0x08 != 0))
p.FastOut(gpio.Level(l&0x10 != 0))
p.FastOut(gpio.Level(l&0x20 != 0))
p.FastOut(gpio.Level(l&0x40 != 0))
p.FastOut(gpio.Level(l&0x80 != 0))
}
b.StopTimer()
}
// benchmarkFastOutBitsMSBUnroll writes into a []byte using MSBF using an
// unrolled loop to iterate over the bits.
//
// It is expected to be slightly faster than benchmarkFastOutBitsMSBLoop.
func (s *Benchmark) benchmarkFastOutBitsMSBUnroll(b *testing.B) {
p := s.p
if err := p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
for i := 0; i < len(buf); i += 2 {
buf[i] = 0xAA
}
b.ResetTimer()
for _, l := range buf {
p.FastOut(gpio.Level(l&0x80 != 0))
p.FastOut(gpio.Level(l&0x40 != 0))
p.FastOut(gpio.Level(l&0x20 != 0))
p.FastOut(gpio.Level(l&0x10 != 0))
p.FastOut(gpio.Level(l&0x08 != 0))
p.FastOut(gpio.Level(l&0x04 != 0))
p.FastOut(gpio.Level(l&0x02 != 0))
p.FastOut(gpio.Level(l&0x01 != 0))
}
b.StopTimer()
}
// benchmarkFastOutInterface is an anti-pattern where an interface is used.
//
// It is otherwise the same as benchmarkFastOutBitsMSBUnroll.
func (s *Benchmark) benchmarkFastOutInterface(b *testing.B) {
type fastOuter interface {
Out(l gpio.Level) error
FastOut(l gpio.Level)
}
var p fastOuter = s.p
if err := p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
for i := 0; i < len(buf); i += 2 {
buf[i] = 0xAA
}
b.ResetTimer()
for _, l := range buf {
p.FastOut(gpio.Level(l&0x80 != 0))
p.FastOut(gpio.Level(l&0x40 != 0))
p.FastOut(gpio.Level(l&0x20 != 0))
p.FastOut(gpio.Level(l&0x10 != 0))
p.FastOut(gpio.Level(l&0x08 != 0))
p.FastOut(gpio.Level(l&0x04 != 0))
p.FastOut(gpio.Level(l&0x02 != 0))
p.FastOut(gpio.Level(l&0x01 != 0))
}
b.StopTimer()
}
// benchmarkFastOutMemberVariabl is an anti-pattern where the struct member
// variable is used.
//
// It is otherwise the same as benchmarkFastOutBitsMSBUnroll.
func (s *Benchmark) benchmarkFastOutMemberVariabl(b *testing.B) {
if err := s.p.Out(gpio.Low); err != nil {
b.Fatal(err)
}
buf := make([]byte, (b.N+7)/8)
for i := 0; i < len(buf); i += 2 {
buf[i] = 0xAA
}
b.ResetTimer()
for _, l := range buf {
s.p.FastOut(gpio.Level(l&0x80 != 0))
s.p.FastOut(gpio.Level(l&0x40 != 0))
s.p.FastOut(gpio.Level(l&0x20 != 0))
s.p.FastOut(gpio.Level(l&0x10 != 0))
s.p.FastOut(gpio.Level(l&0x08 != 0))
s.p.FastOut(gpio.Level(l&0x04 != 0))
s.p.FastOut(gpio.Level(l&0x02 != 0))
s.p.FastOut(gpio.Level(l&0x01 != 0))
}
b.StopTimer()
}