Merge new find_with_marked intp perf v2 #7385
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
finnschiermer
approved these changes
Feb 27, 2024
…:realm/realm-core into fsa/experimental-find-first-optimization
nicola-cab
commented
Feb 27, 2024
| @@ -791,7 +790,11 @@ constexpr uint32_t inverse_width[65] = { | |||
|
|
|||
| inline int first_field_marked(int width, uint64_t vector) | |||
| { | |||
| #if REALM_WINDOWS | |||
| int lz = (int)_tzcnt_u64(vector); // TODO: not clear if this is ok on all platforms | |||
There was a problem hiding this comment.
@finnschiermer this is a tmp fix, just to please the builders.
|
Some small bug to address in object store for frozen objects, but we can merge this and I will fix it later. Perf wise we are closer to master (worst path ~+50% , best path -20%), but we are not using a valid data set and neither we are implementing any heuristic. |
nicola-cab
added a commit
that referenced
this pull request
Mar 1, 2024
* idea: subword parallel search * better subword search * better naming * new methods for reading unaligned word from array of bitfields * perf work on array with find based on parallel values comparison * major cleanup of bitfield scanning * de-templatified bit field search * more tests and code generalization * more tests * new iterator optimized for linear scan * eliminated last use of templates in subword parallel search * optimization of some subword search methods * working EQ cmp with parallel subword check * fix in all_fields_NE * make populate handle negative values * commented out bypass which disabled subword search * fix in fix of populate() * bugfix and direct methods for signed GT and GE * fix for GT condition * enabled array perf tests (outside debug mode) * fixed inner search loop * made some perf tests non concurrent and silenced warnings * moved call to match() into inner loop in subword parallel search * Perf v2, find_with_marked for packed interger arrays (#7385) * made find_first_marked() branch free * various optimizations of find_first_marked, best one selected * for some reason this is much bettergit add . * no warnings * made search method selection more explicit and clear * bunch of fixes.. * restore subword loop * fix object store tests + use subword cmp always (which is faster on my machine) --------- Co-authored-by: Finn Schiermer Andersen <[email protected]> * Perf work for array flex (still missing timestamps) (#7397) * WIP perf work for array flex * more small stuff, nothing important * parallel subword for eq and neq * move find parallel inside loop for eq and neq * LT parallel subword cmp * GT find for array flex * Int equality as good as Packed * code review --------- Co-authored-by: Finn Schiermer Andersen <[email protected]> Co-authored-by: Finn Schiermer Andersen <[email protected]>
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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.