-
Notifications
You must be signed in to change notification settings - Fork 87
Expand file tree
/
Copy pathBENCHMARKS.md
More file actions
259 lines (211 loc) · 16.7 KB
/
Copy pathBENCHMARKS.md
File metadata and controls
259 lines (211 loc) · 16.7 KB
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
# Benchmark: pinky
This benchmark is a cycle-accurate NES emulator, running a real, homebrew NES game. ([source code](https://github.com/koute/polkavm/blob/af63824a1929f8b1ae2c3e790c2a4807a9a17144/guest-programs/bench-pinky/src/main.rs))
## Oneshot execution (for pinky)
These benchmarks measure the end-to-end time that it takes to run the program a single time, including compilation and initialization.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| (bare metal) | 39.003ms ± 0.052ms | 1.00x |
| **PolkaVM (recompiler)** | 58.425ms ± 0.184ms | 1.50x |
| PolkaVM (recompiler, async gas) | 63.575ms ± 0.177ms | 1.63x |
| PolkaVM (recompiler, sync gas) | 73.333ms ± 0.136ms | 1.88x |
| Wasmtime (winch) | 122.738ms ± 0.899ms | 3.15x |
| Wasmtime (cranelift) | 127.711ms ± 0.293ms | 3.27x |
| wazero | 134.183ms ± 5.796ms | 3.44x |
| Wasmtime (cranelift, epoch interruption) | 153.187ms ± 1.146ms | 3.93x |
| Wasmer (singlepass) | 156.030ms ± 1.030ms | 4.00x |
| Wasmtime (cranelift, fuel metering) | 176.693ms ± 0.301ms | 4.53x |
| Solana RBPF | 776.309ms ± 3.500ms | 19.90x |
| Wasm3 | 1.150 s ± 0.001s | 29.50x |
| Wasmi (eager) | 1.179 s ± 0.002s | 30.23x |
| Wasmi (lazy translation) | 1.181 s ± 0.001s | 30.29x |
| Wasmi (lazy) | 1.183 s ± 0.002s | 30.33x |
| CKB VM (ASM) | 1.538 s ± 0.001s | 39.44x |
| PolkaVM (interpreter) | 2.230 s ± 0.000s | 57.18x |
| CKB VM (non-ASM) | 11.956 s ± 0.007s | 306.54x |
## Execution time (for pinky)
These benchmarks measure the execution time of the benchmark, *without* the time it takes to compile or initialize it.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| (bare metal) | 4.295ms ± 0.003ms | 1.00x |
| **PolkaVM (recompiler)** | 6.522ms ± 0.013ms | 1.52x |
| Wasmtime (cranelift) | 7.023ms ± 0.003ms | 1.64x |
| PolkaVM (recompiler, async gas) | 7.184ms ± 0.013ms | 1.67x |
| PolkaVM (recompiler, sync gas) | 8.192ms ± 0.171ms | 1.91x |
| Wasmtime (cranelift, epoch interruption) | 8.374ms ± 0.002ms | 1.95x |
| Wasmtime (cranelift, fuel metering) | 9.866ms ± 0.034ms | 2.30x |
| Wasmtime (winch) | 12.484ms ± 0.281ms | 2.91x |
| wazero | 13.429ms ± 0.021ms | 3.13x |
| Wasmer (singlepass) | 16.301ms ± 0.229ms | 3.80x |
| Solana RBPF | 80.072ms ± 0.424ms | 18.64x |
| Wasm3 | 123.743ms ± 0.063ms | 28.81x |
| Wasmi (lazy translation) | 127.722ms ± 0.216ms | 29.74x |
| Wasmi (lazy) | 130.821ms ± 0.306ms | 30.46x |
| Wasmi (eager) | 130.947ms ± 0.342ms | 30.49x |
| CKB VM (ASM) | 164.443ms ± 0.041ms | 38.29x |
| PolkaVM (interpreter) | 241.175ms ± 0.072ms | 56.15x |
| CKB VM (non-ASM) | 1.302 s ± 0.002s | 303.02x |
## Compilation time (for pinky)
These benchmarks measure the time it takes to compile a given program by the VM.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| Wasmi (lazy) | 97.929µs ± 0.344µs | 1.00x |
| PolkaVM (interpreter) | 246.039µs ± 0.549µs | 2.51x |
| **PolkaVM (recompiler)** | 335.052µs ± 0.171µs | 3.42x |
| PolkaVM (recompiler, async gas) | 357.722µs ± 0.823µs | 3.65x |
| PolkaVM (recompiler, sync gas) | 377.115µs ± 0.510µs | 3.85x |
| Wasmi (lazy translation) | 395.634µs ± 0.627µs | 4.04x |
| Solana RBPF | 598.010µs ± 0.346µs | 6.11x |
| Wasmi (eager) | 975.353µs ± 1.714µs | 9.96x |
| wazero | 1.539ms ± 0.110ms | 15.72x |
| Wasmer (singlepass) | 3.736ms ± 0.004ms | 38.15x |
| Wasmtime (winch) | 6.852ms ± 0.006ms | 69.97x |
| Wasmtime (cranelift) | 62.933ms ± 0.368ms | 642.64x |
| Wasmtime (cranelift, epoch interruption) | 74.570ms ± 0.059ms | 761.47x |
| Wasmtime (cranelift, fuel metering) | 84.926ms ± 0.035ms | 867.23x |
# Benchmark: prime-sieve
This benchmark is a prime sieve, searching for subsequent prime numbers. ([source code](https://github.com/koute/polkavm/tree/af63824a1929f8b1ae2c3e790c2a4807a9a17144/guest-programs/bench-prime-sieve))
## Oneshot execution (for prime-sieve)
These benchmarks measure the end-to-end time that it takes to run the program a single time, including compilation and initialization.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| (bare metal) | 10.787ms ± 0.006ms | 1.00x |
| **PolkaVM (recompiler)** | 19.103ms ± 0.043ms | 1.77x |
| PolkaVM (recompiler, async gas) | 21.965ms ± 0.019ms | 2.04x |
| PolkaVM (recompiler, sync gas) | 28.087ms ± 0.061ms | 2.60x |
| wazero | 45.305ms ± 5.194ms | 4.20x |
| Wasmer (singlepass) | 52.152ms ± 0.049ms | 4.83x |
| Wasmtime (winch) | 56.976ms ± 0.098ms | 5.28x |
| Wasmtime (cranelift) | 108.193ms ± 0.226ms | 10.03x |
| Wasmtime (cranelift, epoch interruption) | 130.627ms ± 0.319ms | 12.11x |
| Wasmtime (cranelift, fuel metering) | 169.620ms ± 0.203ms | 15.72x |
| Wasm3 | 247.270ms ± 0.883ms | 22.92x |
| CKB VM (ASM) | 294.185ms ± 2.110ms | 27.27x |
| Wasmi (lazy translation) | 320.970ms ± 2.782ms | 29.76x |
| Wasmi (eager) | 321.276ms ± 0.636ms | 29.78x |
| Wasmi (lazy) | 321.391ms ± 0.709ms | 29.79x |
| Solana RBPF | 407.462ms ± 0.314ms | 37.77x |
| PolkaVM (interpreter) | 638.592ms ± 0.600ms | 59.20x |
| CKB VM (non-ASM) | 2.477 s ± 0.003s | 229.58x |
## Execution time (for prime-sieve)
These benchmarks measure the execution time of the benchmark, *without* the time it takes to compile or initialize it.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| (bare metal) | 1.584ms ± 0.002ms | 1.00x |
| Wasmtime (cranelift) | 2.160ms ± 0.002ms | 1.36x |
| **PolkaVM (recompiler)** | 2.364ms ± 0.006ms | 1.49x |
| PolkaVM (recompiler, async gas) | 2.463ms ± 0.003ms | 1.55x |
| Wasmtime (cranelift, fuel metering) | 2.472ms ± 0.003ms | 1.56x |
| PolkaVM (recompiler, sync gas) | 2.498ms ± 0.011ms | 1.58x |
| Wasmtime (cranelift, epoch interruption) | 2.673ms ± 0.004ms | 1.69x |
| wazero | 4.735ms ± 0.013ms | 2.99x |
| Wasmtime (winch) | 5.330ms ± 0.003ms | 3.36x |
| Wasmer (singlepass) | 6.095ms ± 0.010ms | 3.85x |
| CKB VM (ASM) | 26.500ms ± 0.016ms | 16.73x |
| Wasm3 | 28.609ms ± 0.018ms | 18.06x |
| Wasmi (eager) | 33.076ms ± 0.137ms | 20.88x |
| Wasmi (lazy) | 33.301ms ± 0.048ms | 21.02x |
| Wasmi (lazy translation) | 33.705ms ± 0.723ms | 21.27x |
| PolkaVM (interpreter) | 84.675ms ± 0.110ms | 53.45x |
| Solana RBPF | 162.503ms ± 0.130ms | 102.57x |
| CKB VM (non-ASM) | 366.227ms ± 0.470ms | 231.16x |
## Compilation time (for prime-sieve)
These benchmarks measure the time it takes to compile a given program by the VM.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| Wasmi (lazy) | 87.280µs ± 1.615µs | 1.00x |
| Wasmi (lazy translation) | 669.850µs ± 1.897µs | 7.67x |
| PolkaVM (interpreter) | 677.561µs ± 0.339µs | 7.76x |
| **PolkaVM (recompiler)** | 851.639µs ± 3.264µs | 9.76x |
| PolkaVM (recompiler, async gas) | 914.962µs ± 5.154µs | 10.48x |
| PolkaVM (recompiler, sync gas) | 955.179µs ± 6.275µs | 10.94x |
| Solana RBPF | 1.449ms ± 0.001ms | 16.60x |
| Wasmi (eager) | 2.085ms ± 0.002ms | 23.89x |
| wazero | 4.371ms ± 0.776ms | 50.08x |
| Wasmer (singlepass) | 9.650ms ± 0.038ms | 110.57x |
| Wasmtime (winch) | 10.085ms ± 0.009ms | 115.55x |
| Wasmtime (cranelift) | 93.418ms ± 0.097ms | 1070.32x |
| Wasmtime (cranelift, epoch interruption) | 106.631ms ± 0.332ms | 1221.70x |
| Wasmtime (cranelift, fuel metering) | 144.623ms ± 0.349ms | 1656.99x |
# Benchmark: minimal
This benchmark is a tiny, minimal program which doesn't do much work; it just increments a global variable and returns immediately. It is a good test case for measuring constant-time overhead. ([source code](https://github.com/koute/polkavm/blob/af63824a1929f8b1ae2c3e790c2a4807a9a17144/guest-programs/bench-minimal/src/main.rs))
## Oneshot execution (for minimal)
These benchmarks measure the end-to-end time that it takes to run the program a single time, including compilation and initialization.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| PolkaVM (interpreter) | 2.444µs ± 0.002µs | 1.00x |
| CKB VM (non-ASM) | 7.081µs ± 0.003µs | 2.90x |
| Solana RBPF | 16.800µs ± 0.064µs | 6.87x |
| Wasm3 | 24.265µs ± 0.030µs | 9.93x |
| (bare metal) | 28.157µs ± 0.012µs | 11.52x |
| Wasmi (lazy) | 28.426µs ± 0.232µs | 11.63x |
| Wasmi (lazy translation) | 29.003µs ± 0.066µs | 11.87x |
| Wasmi (eager) | 29.438µs ± 0.386µs | 12.05x |
| CKB VM (ASM) | 66.107µs ± 0.441µs | 27.05x |
| **PolkaVM (recompiler)** | 95.033µs ± 0.165µs | 38.88x |
| wazero | 95.100µs ± 20.554µs | 38.91x |
| PolkaVM (recompiler, sync gas) | 95.173µs ± 0.263µs | 38.94x |
| PolkaVM (recompiler, async gas) | 95.314µs ± 0.213µs | 39.00x |
| Wasmer (singlepass) | 104.873µs ± 2.653µs | 42.91x |
| Wasmtime (winch) | 184.367µs ± 0.508µs | 75.44x |
| Wasmtime (cranelift) | 707.082µs ± 1.228µs | 289.32x |
| Wasmtime (cranelift, epoch interruption) | 854.033µs ± 9.982µs | 349.45x |
| Wasmtime (cranelift, fuel metering) | 888.159µs ± 1.711µs | 363.41x |
## Execution time (for minimal)
These benchmarks measure the execution time of the benchmark, *without* the time it takes to compile or initialize it.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| (bare metal) | 0.044µs ± 0.000µs | 1.00x |
| Wasm3 | 0.138µs ± 0.000µs | 3.15x |
| Wasmer (singlepass) | 0.140µs ± 0.002µs | 3.21x |
| Wasmtime (cranelift) | 0.158µs ± 0.001µs | 3.63x |
| Wasmtime (cranelift, epoch interruption) | 0.159µs ± 0.001µs | 3.63x |
| Wasmtime (winch) | 0.160µs ± 0.002µs | 3.66x |
| Wasmtime (cranelift, fuel metering) | 0.160µs ± 0.001µs | 3.66x |
| Solana RBPF | 0.166µs ± 0.001µs | 3.80x |
| PolkaVM (interpreter) | 0.171µs ± 0.000µs | 3.92x |
| Wasmi (lazy translation) | 0.183µs ± 0.002µs | 4.18x |
| Wasmi (eager) | 0.183µs ± 0.000µs | 4.19x |
| Wasmi (lazy) | 0.185µs ± 0.000µs | 4.24x |
| wazero | 0.225µs ± 0.004µs | 5.16x |
| CKB VM (ASM) | 2.704µs ± 0.002µs | 61.92x |
| CKB VM (non-ASM) | 3.514µs ± 0.002µs | 80.48x |
| PolkaVM (recompiler, async gas) | 6.906µs ± 0.020µs | 158.17x |
| **PolkaVM (recompiler)** | 6.916µs ± 0.017µs | 158.40x |
| PolkaVM (recompiler, sync gas) | 6.975µs ± 0.043µs | 159.75x |
## Compilation time (for minimal)
These benchmarks measure the time it takes to compile a given program by the VM.
| VM | Time | vs fastest |
|------------------------------------------|----------------------|------------|
| PolkaVM (interpreter) | 1.339µs ± 0.001µs | 1.00x |
| Wasmi (lazy) | 8.577µs ± 0.117µs | 6.40x |
| Wasmi (lazy translation) | 9.818µs ± 0.021µs | 7.33x |
| Wasmi (eager) | 11.628µs ± 0.134µs | 8.68x |
| **PolkaVM (recompiler)** | 14.028µs ± 0.058µs | 10.48x |
| PolkaVM (recompiler, sync gas) | 14.375µs ± 0.157µs | 10.73x |
| PolkaVM (recompiler, async gas) | 14.386µs ± 0.206µs | 10.74x |
| Solana RBPF | 15.478µs ± 0.013µs | 11.56x |
| wazero | 20.861µs ± 0.045µs | 15.58x |
| Wasmtime (winch) | 135.206µs ± 0.131µs | 100.97x |
| Wasmtime (cranelift) | 649.177µs ± 0.343µs | 484.78x |
| Wasmtime (cranelift, epoch interruption) | 800.618µs ± 1.132µs | 597.87x |
| Wasmtime (cranelift, fuel metering) | 843.816µs ± 1.618µs | 630.13x |
---------------------------------------------------------------------------
# Supplemental information
CPU: AMD Ryzen Threadripper 3970X 32-Core Processor
Platform: x86_64-linux
Commit: [af63824a1929f8b1ae2c3e790c2a4807a9a17144](https://github.com/koute/polkavm/tree/af63824a1929f8b1ae2c3e790c2a4807a9a17144)
Timestamp: 2024-03-24 13:12:58 UTC
---------------------------------------------------------------------------
# Replication
You can replicate these benchmarks as follows:
```
$ git clone https://github.com/koute/polkavm.git
$ cd polkavm
$ git checkout af63824a1929f8b1ae2c3e790c2a4807a9a17144
$ cd tools/benchtool
$ ./01-build-benchmarks.sh
$ ./02-run-benchmarks.rb
$ ./03-analyze-benchmarks.rb
```
Only running the benchmarks on Linux is officially supported.
WARNING: The `02-run-benchmarks.rb` script uses a couple of system-level tricks to make benchmarking more consistent and requires 'sudo' and 'schedtool' to be installed. If you're uncomfortable with that or if you're running a non-Linux OS you can also run the benchmarks with `cargo run --release` instead.