Optimize comparison operators #269
Open
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
|
@chfast, let me know what you find out please :-) |
|
Preliminary benchmarks: So this looks solid for compare operators directly and seems to help the shifts as well, although this may be code layout effect. |
chfast
changed the title
|
Thanks, glad to hear :-) |
|
Confirmed, this does not affect assembly of shift operators. Proof: https://godbolt.org/z/4b6cT4rje. |
|
Yes, makes sense. When I played with this repo a while back, I noticed that the benchmarks had variability from one run to the next, so I'm not sure how reliable the comparison truly is. |
Codecov Report
@@ Coverage Diff @@
## master #269 +/- ##
=========================================
Coverage 100.00% 100.00%
=========================================
Files 10 10
Lines 1941 1937 -4
=========================================
- Hits 1941 1937 -4
Flags with carried forward coverage won't be shown. Click here to find out more.
|
|
@chfast should this be merged? |
|
On the latest compilers I don't see much difference so I'm not very much in favor because this implementation has 4 branches while the original one is branchless. But people are welcome to drop their own benchmarking results (use "compare" filter). |
|
Why does it matter that it is branchless if it is slower? |
|
I will try to clean up the situation.
Here you can see the assembly output for each: https://godbolt.org/z/WY3qqzrET AnalysisIn the worst case every variant need to load all 8 words. Additionally, "sub" executes 4 subtract instructions (independent of input). The "ne" executes 4 BenchmarksI added all variants of implementations in #277 and benchmarked them on two CPUs. There are 5 benchmark cases:
HaswellSkylakeZen3This is noisy machine with AMD CPU. For this case I used Xeon Platinum 8272CLFrom the results we should rather consider switching back to "sub" implementation. I may do some other benchmarks if I have time. |
|
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.