curve25519-dalek并行运算性能

export RUSTFLAGS="-C target_cpu=native"
cargo bench --no-default-features --features "std avx2_backend"

与之前博客Rust curve25519-dalek密码学库性能解析对比可发现，采用avx2并行计算，性能提升明显。（测试环境为1CPU，4G内存Ubuntu16.04虚拟机）。若增加虚拟机内CPU核数和内存，性能提升更明显。

https://gist.github.com/hdevalence/d118f270ac1177fd0c357640433ab8aa 中笔记内容为：

git clone https://github.com/dalek-cryptography/curve25519-dalek.git
cd curve25519-dalek
git checkout 0.18.0
echo "nightly-2018-06-19" > rust-toolchain
export RUSTFLAGS="-C target_cpu=znver1"
cargo bench --features "nightly u64_backend" "aA+bB"
cargo bench --features "nightly avx2_backend" "aA+bB"

     Running target/release/deps/dalek_benchmarks-7f3aff4c8fe1105b
Scalar inversion        time:   [11.758 us 11.868 us 12.002 us]
                        change: [-9.7451% +5.2931% +21.841%] (p = 0.53 > 0.05)
                        No change in performance detected.
Found 20 outliers among 100 measurements (20.00%)
  1 (1.00%) high mild
  19 (19.00%) high severe

Batch scalar inversion/1
                        time:   [12.075 us 12.249 us 12.472 us]
                        change: [-9.8322% +3.8519% +17.908%] (p = 0.62 > 0.05)
                        No change in performance detected.
Found 18 outliers among 100 measurements (18.00%)
  1 (1.00%) high mild
  17 (17.00%) high severe
Batch scalar inversion/2
                        time:   [12.274 us 12.414 us 12.591 us]
                        change: [-12.058% +2.3723% +16.892%] (p = 0.76 > 0.05)
                        No change in performance detected.
Found 19 outliers among 100 measurements (19.00%)
  19 (19.00%) high severe
Batch scalar inversion/4
                        time:   [12.822 us 13.012 us 13.249 us]
                        change: [-5.5844% +7.8832% +23.099%] (p = 0.28 > 0.05)
                        No change in performance detected.
Found 22 outliers among 100 measurements (22.00%)
  3 (3.00%) high mild
  19 (19.00%) high severe
Batch scalar inversion/8
                        time:   [17.317 us 19.470 us 21.890 us]
                        change: [+10.266% +33.815% +60.232%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 5 outliers among 100 measurements (5.00%)
  4 (4.00%) high mild
  1 (1.00%) high severe
Batch scalar inversion/16
                        time:   [16.590 us 17.927 us 19.554 us]
                        change: [-16.972% -4.2389% +8.6643%] (p = 0.58 > 0.05)
                        No change in performance detected.
Found 16 outliers among 100 measurements (16.00%)
  3 (3.00%) high mild
  13 (13.00%) high severe

Montgomery pseudomultiplication
                        time:   [101.61 us 111.27 us 121.26 us]
                        change: [+13.822% +34.800% +57.787%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 3 outliers among 100 measurements (3.00%)
  3 (3.00%) high severe

RistrettoPoint compression
                        time:   [5.7802 us 6.3079 us 6.9770 us]
                        change: [-15.929% -5.1237% +5.4630%] (p = 0.41 > 0.05)
                        No change in performance detected.
Found 10 outliers among 100 measurements (10.00%)
  3 (3.00%) high mild
  7 (7.00%) high severe

                                                                            ed 5.0530 s (414k iterations)RistrettoPoint decompression
                        time:   [14.147 us 15.325 us 16.708 us]
                        change: [+90.200% +114.76% +142.62%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 20 outliers among 100 measurements (20.00%)
  4 (4.00%) low mild
  7 (7.00%) high mild
  9 (9.00%) high severe

Benchmarking Batch Ristretto double-and-encode/1: Collecting 100 samples inBatch Ristretto double-and-encode/1
                        time:   [10.898 us 11.364 us 11.916 us]
                        change: [+83.047% +108.29% +135.57%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 18 outliers among 100 measurements (18.00%)
  3 (3.00%) low mild
  6 (6.00%) high mild
  9 (9.00%) high severe
Benchmarking Batch Ristretto double-and-encode/2: Collecting 100 samples inBatch Ristretto double-and-encode/2
                        time:   [11.994 us 12.189 us 12.414 us]
                        change: [+66.956% +91.582% +120.20%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 17 outliers among 100 measurements (17.00%)
  5 (5.00%) low severe
  3 (3.00%) high mild
  9 (9.00%) high severe
Benchmarking Batch Ristretto double-and-encode/4: Collecting 100 samples inBatch Ristretto double-and-encode/4
                        time:   [25.688 us 28.581 us 31.533 us]
                        change: [+173.14% +218.59% +266.92%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 4 outliers among 100 measurements (4.00%)
  2 (2.00%) high mild
  2 (2.00%) high severe
Benchmarking Batch Ristretto double-and-encode/8: Collecting 100 samples inBatch Ristretto double-and-encode/8
                        time:   [10.729 us 10.824 us 10.950 us]
                        change: [-18.022% -9.8761% -2.6149%] (p = 0.02 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  6 (6.00%) high mild
  3 (3.00%) high severe
Benchmarking Batch Ristretto double-and-encode/16: Collecting 100 samples inBatch Ristretto double-and-encode/16
                        time:   [17.181 us 17.308 us 17.450 us]
                        change: [-20.514% -10.813% -1.6923%] (p = 0.04 < 0.05)
                        Performance has improved.
Found 7 outliers among 100 measurements (7.00%)
  5 (5.00%) high mild
  2 (2.00%) high severe

Benchmarking EdwardsPoint compression: Collecting 100 samples in estimated 5EdwardsPoint compression
                        time:   [3.9237 us 4.0441 us 4.1985 us]
                        change: [-14.820% -4.1252% +5.8990%] (p = 0.52 > 0.05)
                        No change in performance detected.
Found 15 outliers among 100 measurements (15.00%)
  4 (4.00%) high mild
  11 (11.00%) high severe

Benchmarking EdwardsPoint decompression: Collecting 100 samples in estimatedEdwardsPoint decompression
                        time:   [5.2305 us 5.2683 us 5.3148 us]
                        change: [-18.087% -7.4637% +1.1875%] (p = 0.20 > 0.05)
                        No change in performance detected.
Found 6 outliers among 100 measurements (6.00%)
  1 (1.00%) high mild
  5 (5.00%) high severe

Constant-time fixed-base scalar mul
                        time:   [27.322 us 28.771 us 30.642 us]
                        change: [+19.979% +42.670% +67.794%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 6 outliers among 100 measurements (6.00%)
  5 (5.00%) high mild
  1 (1.00%) high severe

Constant-time variable-base scalar mul
                        time:   [38.103 us 38.479 us 38.919 us]
                        change: [-55.822% -48.864% -41.636%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 16 outliers among 100 measurements (16.00%)
  1 (1.00%) high mild
  15 (15.00%) high severe

Variable-time aA+bB, A variable, B fixed
                        time:   [34.221 us 34.633 us 35.201 us]
                        change: [-56.445% -48.551% -39.649%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 19 outliers among 100 measurements (19.00%)
  2 (2.00%) high mild
  17 (17.00%) high severe

Constant-time variable-base multiscalar multiplication/1
                        time:   [44.556 us 45.450 us 46.735 us]
                        change: [-58.017% -47.328% -33.101%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Constant-time variable-base multiscalar multiplication/2
                        time:   [66.014 us 66.636 us 67.768 us]
                        change: [-62.468% -51.743% -39.304%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Constant-time variable-base multiscalar multiplication/4
                        time:   [112.00 us 118.73 us 131.32 us]
                        change: [-56.389% -44.312% -29.491%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Constant-time variable-base multiscalar multiplication/8
                        time:   [196.46 us 198.13 us 201.25 us]
                        change: [-49.214% -37.044% -21.080%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Constant-time variable-base multiscalar multiplication/16
                        time:   [370.09 us 374.42 us 381.66 us]
                        change: [-47.703% -33.950% -16.772%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Constant-time variable-base multiscalar multiplication/32
                        time:   [715.54 us 721.08 us 732.16 us]
                        change: [-50.158% -37.001% -21.577%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Constant-time variable-base multiscalar multiplication/64
                        time:   [1.4372 ms 1.5083 ms 1.6573 ms]
                        change: [-40.057% -19.325% +3.0995%] (p = 0.12 > 0.05)
                        No change in performance detected.
Constant-time variable-base multiscalar multiplication/128
                        time:   [2.8152 ms 2.8412 ms 2.8901 ms]
                        change: [-48.691% -35.667% -20.808%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Constant-time variable-base multiscalar multiplication/256
                        time:   [5.6019 ms 5.6829 ms 5.7984 ms]
                        change: [-62.533% -52.773% -38.776%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Constant-time variable-base multiscalar multiplication/384
                        time:   [8.3317 ms 8.4071 ms 8.5425 ms]
                        change: [-46.978% -33.976% -17.981%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Constant-time variable-base multiscalar multiplication/512
                        time:   [11.157 ms 11.279 ms 11.453 ms]
                        change: [-45.586% -31.253% -11.079%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Constant-time variable-base multiscalar multiplication/768
                        time:   [16.668 ms 16.833 ms 17.052 ms]
                        change: [-45.989% -35.297% -21.922%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Constant-time variable-base multiscalar multiplication/1024
                        time:   [22.436 ms 22.603 ms 22.901 ms]
                        change: [-41.414% -32.018% -20.808%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe

Variable-time variable-base multiscalar multiplication/1
                        time:   [32.009 us 32.362 us 32.981 us]
                        change: [-61.484% -50.752% -37.332%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/2
                        time:   [39.440 us 39.801 us 40.455 us]
                        change: [-61.587% -51.092% -38.454%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 15 measurements (26.67%)
  1 (6.67%) low mild
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/4
                        time:   [52.718 us 53.413 us 54.343 us]
                        change: [-64.767% -54.260% -41.566%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/8
                        time:   [82.409 us 83.957 us 85.930 us]
                        change: [-63.983% -53.520% -40.592%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/16
                        time:   [137.99 us 139.18 us 141.70 us]
                        change: [-74.390% -66.543% -55.934%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time variable-base multiscalar multiplication/32
                        time:   [247.69 us 249.69 us 253.99 us]
                        change: [-76.048% -69.737% -61.975%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 15 measurements (26.67%)
  1 (6.67%) low mild
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/64
                        time:   [465.25 us 470.62 us 480.29 us]
                        change: [-65.252% -56.554% -44.948%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/128
                        time:   [902.96 us 913.69 us 934.12 us]
                        change: [-66.331% -57.095% -45.527%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/256
                        time:   [1.6413 ms 1.6555 ms 1.6836 ms]
                        change: [-68.650% -59.655% -47.471%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time variable-base multiscalar multiplication/384
                        time:   [2.2856 ms 2.3174 ms 2.3753 ms]
                        change: [-66.139% -57.503% -46.852%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time variable-base multiscalar multiplication/512
                        time:   [2.8687 ms 2.9335 ms 3.0195 ms]
                        change: [-65.941% -57.506% -47.267%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time variable-base multiscalar multiplication/768
                        time:   [3.9503 ms 4.0154 ms 4.0957 ms]
                        change: [-67.592% -59.632% -48.351%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time variable-base multiscalar multiplication/1024
                        time:   [5.1086 ms 5.1795 ms 5.2720 ms]
                        change: [-65.685% -58.300% -48.651%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe

Variable-time fixed-base multiscalar multiplication/1
                        time:   [31.347 us 31.774 us 32.496 us]
                        change: [-64.228% -53.773% -41.007%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time fixed-base multiscalar multiplication/2
                        time:   [37.801 us 38.247 us 38.953 us]
                        change: [-61.552% -51.240% -39.098%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time fixed-base multiscalar multiplication/4
                        time:   [49.251 us 50.144 us 51.624 us]
                        change: [-62.432% -51.077% -36.512%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time fixed-base multiscalar multiplication/8
                        time:   [72.680 us 73.607 us 75.015 us]
                        change: [-59.781% -49.568% -37.008%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time fixed-base multiscalar multiplication/16
                        time:   [121.68 us 123.98 us 127.90 us]
                        change: [-59.817% -48.515% -33.847%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time fixed-base multiscalar multiplication/32
                        time:   [214.47 us 216.94 us 220.67 us]
                        change: [-58.793% -48.412% -36.223%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time fixed-base multiscalar multiplication/64
                        time:   [403.04 us 407.96 us 416.94 us]
                        change: [-57.240% -47.620% -36.330%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time fixed-base multiscalar multiplication/128
                        time:   [773.60 us 785.60 us 804.77 us]
                        change: [-57.177% -47.069% -34.742%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time fixed-base multiscalar multiplication/256
                        time:   [1.5352 ms 1.5561 ms 1.5875 ms]
                        change: [-56.234% -45.338% -32.129%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time fixed-base multiscalar multiplication/384
                        time:   [2.3122 ms 2.3557 ms 2.4153 ms]
                        change: [-53.630% -43.824% -32.583%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time fixed-base multiscalar multiplication/512
                        time:   [3.1217 ms 3.1856 ms 3.2986 ms]
                        change: [-72.720% -66.712% -58.409%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time fixed-base multiscalar multiplication/768
                        time:   [4.6818 ms 4.8525 ms 5.0627 ms]
                        change: [-49.745% -41.409% -30.072%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time fixed-base multiscalar multiplication/1024
                        time:   [6.3437 ms 6.4028 ms 6.4924 ms]
                        change: [-49.096% -40.208% -27.238%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe

Variable-time mixed-base multiscalar multiplication (0pct dyn)/1
                        time:   [31.049 us 31.484 us 32.158 us]
                        change: [-65.267% -53.766% -39.284%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/2
                        time:   [36.541 us 36.886 us 37.593 us]
                        change: [-66.401% -55.452% -40.706%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/4
                        time:   [48.253 us 48.950 us 50.019 us]
                        change: [-61.536% -50.855% -35.873%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/8
                        time:   [71.186 us 72.079 us 73.834 us]
                        change: [-61.147% -49.275% -34.459%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/16
                        time:   [122.20 us 124.03 us 126.19 us]
                        change: [-60.455% -49.671% -35.338%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/32
                        time:   [213.87 us 216.62 us 220.97 us]
                        change: [-62.166% -52.026% -40.443%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/64
                        time:   [399.35 us 402.87 us 410.04 us]
                        change: [-58.633% -48.409% -35.624%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/128
                        time:   [775.31 us 789.83 us 808.15 us]
                        change: [-57.471% -46.597% -31.880%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/256
                        time:   [1.5270 ms 1.5364 ms 1.5553 ms]
                        change: [-57.771% -46.462% -32.774%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 15 measurements (26.67%)
  1 (6.67%) low mild
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/384
                        time:   [2.3059 ms 2.3297 ms 2.3679 ms]
                        change: [-55.509% -44.175% -30.690%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/512
                        time:   [3.0876 ms 3.1384 ms 3.2045 ms]
                        change: [-55.343% -44.852% -32.114%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/768
                        time:   [4.6441 ms 4.6872 ms 4.7589 ms]
                        change: [-48.930% -39.051% -24.919%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (0pct dyn)/1024
                        time:   [6.3576 ms 6.4380 ms 6.5171 ms]
                        change: [-49.108% -42.518% -35.767%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe

Variable-time mixed-base multiscalar multiplication (20pct dyn)/1
                        time:   [30.854 us 31.355 us 32.049 us]
                        change: [-62.091% -51.738% -39.410%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/2
                        time:   [36.635 us 37.062 us 37.832 us]
                        change: [-63.863% -54.066% -41.430%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/4
                        time:   [48.938 us 49.900 us 50.923 us]
                        change: [-61.493% -51.682% -37.781%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/8
                        time:   [74.475 us 75.454 us 77.029 us]
                        change: [-61.011% -50.648% -36.597%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/16
                        time:   [130.68 us 133.68 us 138.13 us]
                        change: [-61.641% -50.355% -36.078%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/32
                        time:   [223.37 us 226.26 us 230.39 us]
                        change: [-64.771% -54.611% -41.629%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/64
                        time:   [422.99 us 430.07 us 439.75 us]
                        change: [-59.202% -47.202% -31.547%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/128
                        time:   [811.05 us 817.68 us 831.99 us]
                        change: [-61.892% -50.745% -37.388%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/256
                        time:   [1.5990 ms 1.6227 ms 1.6569 ms]
                        change: [-57.979% -47.886% -36.500%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/384
                        time:   [2.3941 ms 2.4231 ms 2.4629 ms]
                        change: [-57.890% -46.382% -32.828%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/512
                        time:   [4.3678 ms 5.6680 ms 6.4096 ms]
                        change: [-41.770% -28.533% -12.680%] (p = 0.00 < 0.05)
                        Performance has improved.
Variable-time mixed-base multiscalar multiplication (20pct dyn)/768
                        time:   [4.8405 ms 4.8906 ms 4.9637 ms]
                        change: [-53.932% -45.604% -34.033%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (20pct dyn)/1024
                        time:   [6.6044 ms 6.6966 ms 6.8190 ms]
                        change: [-50.498% -42.790% -31.922%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe

Variable-time mixed-base multiscalar multiplication (50pct dyn)/1
                        time:   [33.326 us 34.518 us 36.299 us]
                        change: [-59.575% -48.255% -32.768%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/2
                        time:   [40.477 us 41.291 us 42.375 us]
                        change: [-63.104% -51.908% -36.532%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/4
                        time:   [54.725 us 55.666 us 56.946 us]
                        change: [-63.210% -51.949% -36.572%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/8
                        time:   [81.637 us 83.847 us 86.776 us]
                        change: [-58.902% -47.557% -33.438%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/16
                        time:   [136.95 us 139.26 us 142.10 us]
                        change: [-61.792% -52.150% -40.257%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/32
                        time:   [240.51 us 244.52 us 248.65 us]
                        change: [-64.056% -54.555% -42.640%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/64
                        time:   [448.48 us 458.40 us 469.47 us]
                        change: [-60.510% -49.722% -35.914%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/128
                        time:   [861.79 us 873.04 us 888.20 us]
                        change: [-61.740% -51.493% -38.396%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/256
                        time:   [1.6788 ms 1.6971 ms 1.7233 ms]
                        change: [-61.690% -52.895% -40.635%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  3 (20.00%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/384
                        time:   [3.0623 ms 3.4853 ms 3.8362 ms]
                        change: [-53.111% -42.374% -30.227%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 1 outliers among 15 measurements (6.67%)
  1 (6.67%) high mild
Variable-time mixed-base multiscalar multiplication (50pct dyn)/512
                        time:   [3.3801 ms 3.4386 ms 3.5381 ms]
                        change: [-57.869% -46.880% -32.766%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/768
                        time:   [5.1305 ms 5.2413 ms 5.3797 ms]
                        change: [-57.582% -48.441% -37.380%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 3 outliers among 15 measurements (20.00%)
  1 (6.67%) high mild
  2 (13.33%) high severe
Variable-time mixed-base multiscalar multiplication (50pct dyn)/1024
                        time:   [7.1194 ms 7.3318 ms 7.6753 ms]
                        change: [-53.294% -44.064% -33.131%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 2 outliers among 15 measurements (13.33%)
  2 (13.33%) high severe

参考资料：
[1] https://gist.github.com/hdevalence/d118f270ac1177fd0c357640433ab8aa