streamline/speed up addc/subc by using in/out carry parameter #267
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
|
I believe this version is significantly faster! |
|
I benchmarked GCC 11. In general we would need to split these into smaller changes (I can do this myself) for better analysis.
We should also consult Clang. It can optimize add/sub perfectly so as long it stays this way we can go ahead for changes making it easier for GCC. |
|
Thanks for looking into it @chfast ! I'm not sure how to read the numbers you posted, but I am surprised that you found that add/sub got slower. In my tests it was faster. I noticed that sometimes running the same code twice gave different numbers though! Benchmarks are hard. |
chfast
reviewed
Apr 5, 2022
| @@ -182,6 +182,11 @@ inline constexpr uint64_t addc( | |||
| } | |||
| #endif | |||
|
|
|||
| if (((x | y) & (uint64_t(1) << 63)) == 0) { | |||
There was a problem hiding this comment.
This is ineffective for compilers we care about. GCC and Clang should be using builtins.
There was a problem hiding this comment.
So if we use the builtin how do you explain that my version is slower in the benchmark?
There was a problem hiding this comment.
Actually I had removed this change. Are you looking at my latest version?
see https://github.com/greg7mdp/intx/blob/master/include/intx/intx.hpp
|
Also why diff between add and inline_add? |
This is comparison between master and your changes. Numbers are:
This comes from a tool comparing outputs of |
Because the benchmark run in a loop the "inline_add" can be vectorized. This is not relevant for EVM use case. I probably could review all benchmark cases and remove half of these. |
|
Thanks!
vectorized or inlined? Why shouldn't it be relevant for EVM? |
|
The main difference is that the |
|
@chfast if you are using gcc 11 and therefore the builtin, I am puzzled that my version would be slower. |
chfast
reviewed
Apr 12, 2022
| return subc(x, y).carry; | ||
| for (size_t i = uint<N>::num_words; i-- > 1; ) { | ||
| if (x[i] != y[i]) | ||
| return x[i] < y[i]; |
There was a problem hiding this comment.
This single change looks interesting. Can you submit it as a separate PR for easier verification?
|
Kudos, SonarCloud Quality Gate passed!
|
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.
No description provided.