Fix fast 1-weso parameter approximation, for massive speedup #205
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The "fast" 1-Wesolowski prover inadvertently mixed up parameters in the
call to ApproximateParameter.
ApproximateParameter's signature has L followed by K:
void ApproximateParameters(uint64_t T, uint32_t& L, uint32_t& K)
However, the 1-Weso calls used K followed by L:
ApproximateParameters(num_iterations, k, l);
While this typo does not affect the correctness of the created 1-Weso
proof, this causes 1-Weso proving to be _much_ slower than necessary. In
isolated testing, the 1-Weso proof generation is up to 10x slower.
Funnily enough, the "slow" VDF impl (in `src/provers_slow.h`) uses a
separate 1-Weso prover implementation which _doesn't_ suffer from this
bug, thus making the "slow" VDF impl faster than the "fast" overall.
To be clear: the fixed bug _only_ affects the _performance_ (not the
correctness!) of computing 1-Weso proofs (thus, blueboxing) using the
"fast" implementation. Regular software timelords (N-Weso), hardware
(ASIC) timelords, and the "slow" bluebox implementation are unaffected.
The speedup can be observed/tested by running `src/1weso_test.cpp`:
$ hyperfine --warmup 5 "./1weso_test.orig" "./1weso_test.fixed"
Benchmark 1: ./1weso_test.orig
Time (mean ± σ): 17.250 s ± 1.153 s [User: 23.946 s, System: 0.037 s]
Range (min … max): 15.582 s … 18.372 s 10 runs
Benchmark 2: ./1weso_test.fixed
Time (mean ± σ): 8.746 s ± 0.010 s [User: 14.651 s, System: 0.029 s]
Range (min … max): 8.733 s … 8.765 s 10 runs
Summary
./1weso_test.fixed ran
1.97 ± 0.13 times faster than ./1weso_test.orig
|
merged here because it needed additional unrelated changes for CI but will be sure to credit you |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
The "fast" 1-Wesolowski prover inadvertently mixed up parameters in the call to ApproximateParameter.
ApproximateParameter's signature has L followed by K:
However, the 1-Weso calls used K followed by L:
While this typo does not affect the correctness of the created 1-Weso proof, this causes 1-Weso proving to be much slower than necessary. In isolated testing, the 1-Weso proof generation is up to 10x slower.
Funnily enough, the "slow" VDF impl (in
src/provers_slow.h) uses a separate 1-Weso prover implementation which doesn't suffer from this bug, thus making the "slow" VDF impl faster than the "fast" overall.To be clear: the fixed bug only affects the performance (not the correctness!) of computing 1-Weso proofs (thus, blueboxing) using the "fast" implementation. Regular software timelords (N-Weso), hardware (ASIC) timelords, and the "slow" bluebox implementation are unaffected.
The speedup can be observed/tested by running
src/1weso_test.cpp:1weso_testalso confirms that 1-Weso proof correctness is unaffected by this change. The prover still generates the exact same proofs, it just does so much faster.