1 Billion Row Challenge in Rust Environment for the benchmark
48 Intel vCPUs / 96 GB Memory / 600 GB Disk, dedicated CPU-optimized DigitalOcean instance with Premium Intel
CPU, c-48-intel
Ubuntu 24.04 LTS (GNU/Linux 6.8.0-31-generic x86_64)
Rust compiler rustc 1.78.0 (9b00956e5 2024-04-29), x86_64-unknown-linux-gnu, LLVM version: 18.1.2
Java 21.0.3 2024-04-16 LTS, Java HotSpot(TM) 64-Bit Server VM Oracle GraalVM 21.0.3+7.1 (build
21.0.3+7-LTS-jvmci-23.1-b37, mixed mode, sharing)
File with 1 billion measurements that is used as an input is stored in tmpfs
The results obtained by running scripts/run_benchmark_original.sh . It slightly
modifies the original code to allow passing number of cores and also makes sure the project is built with native CPU
support to get max performance.
Number of threads
Mean [s]
Min [s]
Max [s]
1
42.898 ± 0.068
42.830
42.992
2
21.718 ± 0.019
21.703
21.745
4
10.955 ± 0.046
10.904
11.017
8
5.686 ± 0.036
5.652
5.724
16
3.024 ± 0.019
3.006
3.050
24
2.209 ± 0.029
2.168
2.231
32
2.185 ± 0.013
2.176
2.205
48
1.722 ± 0.007
1.711
1.726
The results obtained by running scripts/run_benchmark.sh .
Implementation
Number of threads
Mean [s]
Min [s]
Max [s]
parse_large_chunks_as_i64_v2
1
19.797 ± 0.031
19.764
19.830
parse_large_chunks_as_i64_as_java
1
16.848 ± 0.050
16.784
16.900
parse_large_chunks_as_i64_v2
2
10.010 ± 0.016
9.990
10.030
parse_large_chunks_as_i64_as_java
2
8.570 ± 0.010
8.558
8.579
parse_large_chunks_as_i64_v2
4
5.077 ± 0.010
5.065
5.090
parse_large_chunks_as_i64_as_java
4
4.373 ± 0.005
4.368
4.380
parse_large_chunks_as_i64_v2
8
2.650 ± 0.010
2.639
2.661
parse_large_chunks_as_i64_as_java
8
2.299 ± 0.009
2.286
2.306
parse_large_chunks_as_i64_v2
16
1.496 ± 0.014
1.478
1.507
parse_large_chunks_as_i64_as_java
16
1.321 ± 0.006
1.312
1.326
parse_large_chunks_as_i64_v2
24
1.148 ± 0.009
1.137
1.156
parse_large_chunks_as_i64_as_java
24
1.057 ± 0.058
1.002
1.138
parse_large_chunks_as_i64_v2
32
1.232 ± 0.019
1.214
1.259
parse_large_chunks_as_i64_as_java
32
1.127 ± 0.009
1.116
1.138
parse_large_chunks_as_i64_v2
48
1.223 ± 0.012
1.210
1.238
parse_large_chunks_as_i64_as_java
48
1.174 ± 0.008
1.163
1.180
Single-thread dummy implementations to understand how fast it can be
Implementation
Mean [s]
Min [s]
Max [s]
naive_line_by_line_dummy
44.508 ± 0.370
44.191
45.042
parse_large_chunks_as_bytes_dummy
25.021 ± 0.059
24.941
25.081
parse_large_chunks_as_i64_dummy
16.163 ± 0.156
16.021
16.328
parse_large_chunks_simd_dummy
10.931 ± 0.044
10.896
10.995
All implementations in single thread
Command
Mean [s]
Min [s]
Max [s]
Relative
naive_line_by_line
91.523 ± 0.482
90.955
92.110
5.43 ± 0.03
naive_line_by_line_v2
58.189 ± 0.306
57.847
58.493
3.45 ± 0.02
parse_large_chunks_as_bytes
33.714 ± 0.096
33.623
33.848
2.00 ± 0.01
parse_large_chunks_as_i64
26.451 ± 0.478
26.122
27.162
1.57 ± 0.03
parse_large_chunks_as_i64_v2
19.797 ± 0.031
19.764
19.830
1.18 ± 0.00
parse_large_chunks_as_i64_unsafe
27.408 ± 0.390
27.058
27.780
1.63 ± 0.02
parse_large_chunks_as_i64_as_java
16.848 ± 0.050
16.784
16.900
1.00
parse_large_chunks_simd
31.520 ± 0.082
31.437
31.632
1.87 ± 0.01
parse_large_chunks_simd_v1
30.221 ± 0.339
29.961
30.689
1.79 ± 0.02
Type of run
Number of threads
Mean [s]
Min [s]
Max [s]
JVM
1
9.420 ± 0.060
9.327
9.504
Native Image
1
8.647 ± 0.038
8.612
8.704
JVM
48
2.133 ± 0.236
1.857
2.586
Native Image
48
0.4121 ± 0.0051
0.4019
0.4163
hyperfine --warmup 4 --runs 10 --export-markdown java_thomaswue.md " java --enable-preview --class-path /root/code/github/gunnarmorling/1brc/target/average-1.0.0-SNAPSHOT.jar dev.morling.onebrc.CalculateAverage_thomaswue" hyperfine --warmup 4 --runs 10 --export-markdown java_native_thomaswue.md /root/code/github/gunnarmorling/1brc/target/CalculateAverage_thomaswue_image