uint: div_mod_word use intrinsic u128 division(#406) #478

frostRed · 2020-12-11T05:34:05Z

The fuzz test is ok, but I don't see any div_mod_word bench test because it's a private function.

ordian

please report the results of the division benchmarks in uint:

cd uint
cargo bench -- div

ordian · 2020-12-11T10:52:24Z

uint/fuzz/fuzz_targets/div_mod_word.rs

+	let x = (u128::from(hi) << 64) + u128::from(lo);
+	let y = u128::from(y);
+	((x / y) as u64, (x % y) as u64)


please revert this change. we're comparing two implementations here, they should produce identical results, it doesn't make sense to compare it with itself.

No problem.

ordian · 2020-12-11T11:19:58Z

@frostRed to be clear, I didn't mean the PR should be closed, just the change in fuzz tests should be reverted.

frostRed · 2020-12-11T11:25:45Z

Sorry for misunderstand, I will just revert fuzz tests.

frostRed · 2020-12-11T11:42:47Z

Hardware：CPU 8 cores, Intel(R) Core(TM) i7-8550U CPU @ 1.80GHz, 16G Memory
OS：Linux version 5.4.80-2-MANJARO
Rust：stable-x86_64-unknown-linux-gnu (default)

cargo bench -- div result:

running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out

     Running /home/paysonl/Work/TurboGeth/parity-common/target/release/deps/bigint-955fd06249dc3a95
Gnuplot not found, using plotters backend
u256_div                time:   [79.113 ns 80.665 ns 82.545 ns]                     
                        change: [-6.8722% -4.0399% -1.3118%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  9 (9.00%) high severe

u512_div_mod//(1340780792994259709957402499820584612747936582059239337772356144372176403007354697680...                                                                            
                        time:   [94.804 ns 97.305 ns 100.24 ns]
                        change: [-8.3888% -5.2581% -2.1371%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 10 outliers among 100 measurements (10.00%)
  6 (6.00%) high mild
  4 (4.00%) high severe
u512_div_mod//(18446744073709551615, 4294967295)                                                                            
                        time:   [80.631 ns 85.097 ns 89.689 ns]
                        change: [+21.595% +27.837% +34.164%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 7 outliers among 100 measurements (7.00%)
  7 (7.00%) high mild
u512_div_mod//(18446744073709551615, 18446744073709551614)                                                                            
                        time:   [93.279 ns 95.371 ns 98.026 ns]
                        change: [-1.3642% +0.6383% +3.0009%] (p = 0.57 > 0.05)
                        No change in performance detected.
Found 12 outliers among 100 measurements (12.00%)
  5 (5.00%) high mild
  7 (7.00%) high severe
u512_div_mod//(18446744073709551615, 18446744073709551614) #2                                                                            
                        time:   [98.748 ns 102.06 ns 105.45 ns]
                        change: [+2.2879% +5.0513% +8.0426%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 13 outliers among 100 measurements (13.00%)
  6 (6.00%) high mild
  7 (7.00%) high severe
u512_div_mod//(3759751734479964094783137206182536765532905409829204647089173492, 2167484464668298946...                                                                            
                        time:   [133.25 ns 137.63 ns 142.49 ns]
                        change: [+18.953% +22.845% +26.439%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 1 outliers among 100 measurements (1.00%)
  1 (1.00%) high mild
u512_div_mod//(1942668892225729070919461906823518906642406839052139521251812409738904285205208498175...                                                                            
                        time:   [105.52 ns 107.42 ns 109.72 ns]
                        change: [+5.6022% +8.5035% +11.329%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 12 outliers among 100 measurements (12.00%)
  5 (5.00%) high mild
  7 (7.00%) high severe
u512_div_mod//(2045869129935088668758243560517249470135401278776915493427057105060083622743875339830...                                                                            
                        time:   [375.91 ns 384.56 ns 394.69 ns]
                        change: [+0.0208% +2.8799% +5.8761%] (p = 0.06 > 0.05)
                        No change in performance detected.
Found 6 outliers among 100 measurements (6.00%)
  5 (5.00%) high mild
  1 (1.00%) high severe

u512_div                time:   [109.21 ns 110.54 ns 112.29 ns]                     
                        change: [-13.816% -11.509% -9.1624%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 100 measurements (4.00%)
  4 (4.00%) high severe

u128_div//(0, 18446744073709551615, 100)                                                                             
                        time:   [6.4493 ns 6.6085 ns 6.8338 ns]
                        change: [-0.2883% +2.3483% +5.4248%] (p = 0.12 > 0.05)
                        No change in performance detected.
Found 5 outliers among 100 measurements (5.00%)
  1 (1.00%) high mild
  4 (4.00%) high severe
u128_div//(18446744073709551615, 18446744073709551615, 99)                                                                            
                        time:   [306.36 ns 309.53 ns 313.88 ns]
                        change: [-52.692% -49.737% -46.654%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 9 outliers among 100 measurements (9.00%)
  5 (5.00%) high mild
  4 (4.00%) high severe
u128_div//(42, 42, 100500)                                                                            
                        time:   [245.77 ns 249.00 ns 253.12 ns]
                        change: [-22.851% -20.064% -16.626%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 15 outliers among 100 measurements (15.00%)
  1 (1.00%) high mild
  14 (14.00%) high severe

ordian · 2020-12-25T21:02:37Z

Unfortunately, the benchmarks are inconclusive. There are some severe regressions, would need to investigate further. This will likely not go into the next release, but we can publish it later as a patch release.

ordian · 2022-02-04T09:19:06Z

Getting good results on my laptop with rust 1.58 and CARGO_PROFILE_RELEASE_LTO=true cargo bench -- div:

u256_div                time:   [64.181 ns 64.274 ns 64.369 ns]                     
                        change: [-2.8874% -1.0136% +0.4653%] (p = 0.30 > 0.05)
                        No change in performance detected.
Found 3 outliers among 100 measurements (3.00%)
  1 (1.00%) low mild
  2 (2.00%) high mild

u512_div_mod//(1340780792994259709957402499820584612747936582059239337772356144372176403007354697680...                                                                            
                        time:   [76.674 ns 76.833 ns 77.016 ns]
                        change: [-59.160% -58.841% -58.557%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 11 outliers among 100 measurements (11.00%)
  2 (2.00%) low mild
  3 (3.00%) high mild
  6 (6.00%) high severe
u512_div_mod//(18446744073709551615, 4294967295)                                                                            
                        time:   [73.206 ns 73.451 ns 73.761 ns]
                        change: [-48.940% -48.548% -47.952%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 5 outliers among 100 measurements (5.00%)
  2 (2.00%) high mild
  3 (3.00%) high severe
u512_div_mod//(18446744073709551615, 18446744073709551614)                                                                            
                        time:   [79.133 ns 79.251 ns 79.376 ns]
                        change: [-43.538% -43.420% -43.307%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 8 outliers among 100 measurements (8.00%)
  4 (4.00%) low mild
  4 (4.00%) high mild
u512_div_mod//(18446744073709551615, 18446744073709551614) #2                                                                            
                        time:   [79.176 ns 79.251 ns 79.327 ns]
                        change: [-43.839% -43.687% -43.540%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 6 outliers among 100 measurements (6.00%)
  2 (2.00%) low severe
  1 (1.00%) low mild
  2 (2.00%) high mild
  1 (1.00%) high severe
u512_div_mod//(3759751734479964094783137206182536765532905409829204647089173492, 2167484464668298946...                                                                            
                        time:   [89.617 ns 89.767 ns 89.941 ns]
                        change: [+1.5136% +1.8538% +2.2394%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 13 outliers among 100 measurements (13.00%)
  6 (6.00%) low mild
  2 (2.00%) high mild
  5 (5.00%) high severe
u512_div_mod//(1942668892225729070919461906823518906642406839052139521251812409738904285205208498175...                                                                            
                        time:   [75.859 ns 75.938 ns 76.024 ns]
                        change: [-11.705% -11.482% -11.263%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 10 outliers among 100 measurements (10.00%)
  4 (4.00%) low mild
  5 (5.00%) high mild
  1 (1.00%) high severe
u512_div_mod//(2045869129935088668758243560517249470135401278776915493427057105060083622743875339830...                                                                            
                        time:   [190.73 ns 191.19 ns 191.71 ns]
                        change: [-3.4927% -3.2306% -2.9578%] (p = 0.00 < 0.05)
                        Performance has improved.
Found 4 outliers among 100 measurements (4.00%)
  4 (4.00%) high severe

u512_div                time:   [86.340 ns 87.108 ns 88.241 ns]                     
                        change: [+3.1944% +4.0582% +5.2751%] (p = 0.00 < 0.05)
                        Performance has regressed.
Found 9 outliers among 100 measurements (9.00%)
  2 (2.00%) low mild
  3 (3.00%) high mild
  4 (4.00%) high severe

ordian reviewed Dec 11, 2020

View reviewed changes

frostRed closed this Dec 11, 2020

frostRed deleted the div-mod-word branch December 11, 2020 11:13

frostRed restored the div-mod-word branch December 11, 2020 11:25

frostRed reopened this Dec 11, 2020

uint: div_mod_word use intrinsic u128 division(paritytech#406)

b1b72a9

frostRed force-pushed the div-mod-word branch from 7a77934 to b1b72a9 Compare December 11, 2020 11:31

ordian mentioned this pull request Jul 2, 2021

Removed a workaround for 128-bit division #406 #565

Closed

ordian linked an issue Jul 11, 2021 that may be closed by this pull request

[uint] remove a workaround for 128-bit division #406

Closed

ordian mentioned this pull request Dec 12, 2021

U256 div measurement #608

Closed

ordian added the changelog Needs to be added to the changelog label Feb 4, 2022

ordian merged commit 2d7f7e2 into paritytech:master Feb 4, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

uint: div_mod_word use intrinsic u128 division(#406) #478

uint: div_mod_word use intrinsic u128 division(#406) #478

frostRed commented Dec 11, 2020

ordian left a comment

ordian Dec 11, 2020

frostRed Dec 11, 2020

ordian commented Dec 11, 2020

frostRed commented Dec 11, 2020

frostRed commented Dec 11, 2020

ordian commented Dec 25, 2020

ordian commented Feb 4, 2022

uint: div_mod_word use intrinsic u128 division(#406) #478

uint: div_mod_word use intrinsic u128 division(#406) #478

Conversation

frostRed commented Dec 11, 2020

ordian left a comment

Choose a reason for hiding this comment

ordian Dec 11, 2020

Choose a reason for hiding this comment

frostRed Dec 11, 2020

Choose a reason for hiding this comment

ordian commented Dec 11, 2020

frostRed commented Dec 11, 2020

frostRed commented Dec 11, 2020

ordian commented Dec 25, 2020

ordian commented Feb 4, 2022