Optimize Crypto::Bcrypt performance #3880
Merged
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
lribeiro
reviewed
Jan 13, 2017
| 0x01c36ae4, 0xd6ebe1f9, 0x90d4f869, 0xa65cdea0, 0x3f09252d, 0xc208e69f, | ||
| 0xb74e6132, 0xce77e25b, 0x578fdfe3, 0x3ac372e6, | ||
| ], | ||
| private S = StaticArray[ |
There was a problem hiding this comment.
Why not UInt32.static_array(...) instead?
Won't it spare us all the _u32?
There was a problem hiding this comment.
Thanks. I wasn't aware of this initializer. I wish Int[1, 2, 3] would return a StaticArray instead of an Array...
ysbaddaden
force-pushed
the
enhancement-bcrypt-performance
branch
2 times, most recently
from
093eaf7
to
8e36580
Compare
|
replace f with, gives also 5% performance: @[AlwaysInline]
private def f(x)
d = x.to_u8
c = (x >> 8).to_u8
b = (x >> 16).to_u8
a = (x >> 24).to_u8
((@s.to_unsafe[a] + @s.to_unsafe[256 + b]) ^ @s.to_unsafe[512 + c]) + @s.to_unsafe[768 + d]
end |
ysbaddaden
force-pushed
the
enhancement-bcrypt-performance
branch
from
8e36580
to
8eef6af
Compare
Optimizes the Blowfish and Bcrypt algorithms by using stack allocated objects (StaticArray) instead of a HEAP allocation (Array), and raw pointer accesses. The implementation is still not on par with alternative C implementations (e.g. as used by the Ruby extension).
ysbaddaden
force-pushed
the
enhancement-bcrypt-performance
branch
from
8eef6af
to
535c50f
Compare
|
@kostya I didn't notice an improvement (above benchmark, LLVM 3.9.1). I got a 6% improvement, by memoizing pointers to It's now down to 2.81s vs 2.46s (ruby). |
mverzilli
approved these changes
Jan 20, 2017
|
Excellent as usual, Julien! 🙇 Thank you so much, this is an amazing contribution 💯 |
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 current
Crypto::BCryptimplementation is roughly ~68% slower than the Ruby extension.Changing the HEAP allocated Array for the stack allocated StaticArray and flattening
Pimmediately reduces the difference to ~32% slower. Using raw pointers closes the gap further by going down to ~19% slower. Raw pointers are unsafe, but the algorithm has control over the pointer's range, and it never goes out of bounds. It's still not on par with the Ruby implementation (that uses C) but it's getting closer.Running this benchmark, generating 10 hashes at cost 12 got a
71%41% performance increase, down from 4.15s to 2.94s (vs 2.46s for Ruby).refs #1309