refactor: Speed up Base58 encoding/decoding by 400%/200% via preliminary byte packing #29473
+113
−84
Commits on May 29, 2024
-
-
Encode by iterating forward and dividing in-place
Instead of collecting the values to and empty b58, we're cloning the input without leading zeros and dividing it in-place, while doing the quadratic division from the most significant, non-zero digit, forward. make && ./src/bench/bench_bitcoin --filter=Base58Encode --min-time=1000 Before: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 59.34 | 16,851,250.88 | 0.2% | 1.10 | `Base58Encode` | 59.20 | 16,892,479.61 | 0.2% | 1.10 | `Base58Encode` | 58.97 | 16,958,226.76 | 0.2% | 1.10 | `Base58Encode` ``` After: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 40.80 | 24,508,402.06 | 0.1% | 1.10 | `Base58Encode` | 40.83 | 24,489,762.49 | 0.2% | 1.10 | `Base58Encode` | 39.71 | 25,182,310.62 | 0.2% | 1.10 | `Base58Encode` ``` and make && ./src/bench/bench_bitcoin --filter=Base58CheckEncode --min-time=1000 Before: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 79.93 | 12,511,576.75 | 0.4% | 1.10 | `Base58CheckEncode` | 79.49 | 12,580,136.21 | 0.1% | 1.10 | `Base58CheckEncode` | 79.65 | 12,554,785.16 | 0.1% | 1.10 | `Base58CheckEncode` ``` After: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 55.27 | 18,094,561.35 | 0.1% | 1.10 | `Base58CheckEncode` | 55.41 | 18,046,778.32 | 0.1% | 1.10 | `Base58CheckEncode` | 55.32 | 18,075,763.16 | 0.1% | 1.10 | `Base58CheckEncode` ```
-
Speed up Base58 encoding with 64-bit packing
To mitigate the quadratic complexity inherent in the sequential byte division of the Base58 encoding process, this update aggregates characters into groups of 7, fitting them into 64-bit long integers. This refinement utilizes the inherent efficiency of native division and modulo operations on 64-bit architectures, significantly optimizing computational overhead. The benchmarks indicate a 4x speedup for `Base58CheckEncode` and `Base58Encode`: make && ./src/bench/bench_bitcoin --filter=Base58Encode --min-time=1000 After: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 14.32 | 69,831,031.44 | 0.1% | 1.10 | `Base58Encode` | 14.32 | 69,811,995.18 | 0.1% | 1.10 | `Base58Encode` | 14.35 | 69,662,527.88 | 0.5% | 1.10 | `Base58Encode` ``` make && ./src/bench/bench_bitcoin --filter=Base58CheckEncode --min-time=1000 After: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 19.15 | 52,231,968.11 | 0.1% | 1.06 | `Base58CheckEncode` | 19.13 | 52,269,345.54 | 0.4% | 1.05 | `Base58CheckEncode` | 19.14 | 52,244,117.61 | 0.3% | 1.06 | `Base58CheckEncode` ```
-
Decode by iterating forward and dividing in-place
> make && ./src/bench/bench_bitcoin --filter=Base58Decode --min-time=1000 Before: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 17.50 | 57,141,622.65 | 0.1% | 1.10 | `Base58Decode` | 17.42 | 57,392,223.96 | 0.0% | 1.10 | `Base58Decode` | 17.43 | 57,358,655.44 | 0.2% | 1.10 | `Base58Decode` ``` After: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 16.37 | 61,093,842.90 | 0.2% | 1.10 | `Base58Decode` | 16.37 | 61,100,514.64 | 0.1% | 1.10 | `Base58Decode` | 16.38 | 61,046,275.93 | 0.1% | 1.10 | `Base58Decode` ```
-
Speed up Base58 decoding with 64-bit packing
> make && ./src/bench/bench_bitcoin --filter=Base58Decode --min-time=1000 After: ``` | ns/byte | byte/s | err% | total | benchmark |--------------------:|--------------------:|--------:|----------:|:---------- | 8.79 | 113,767,685.06 | 0.1% | 1.10 | `Base58Decode` | 8.78 | 113,831,528.33 | 0.0% | 1.10 | `Base58Decode` | 8.80 | 113,607,470.35 | 0.2% | 1.10 | `Base58Decode` ```
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.