Bq with u8 alignment support #25
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
IvanPleshkov
commented
May 30, 2024
| EXPORT uint64_t impl_xor_popcnt_neon( | ||
| const uint64_t* query_ptr, | ||
| const uint64_t* vector_ptr, | ||
| EXPORT uint32_t impl_xor_popcnt_neon_uint128( |
There was a problem hiding this comment.
just simplification. not much difference in benches
IvanPleshkov
commented
May 30, 2024
|
|
||
| return (uint64_t)vaddvq_u32(result); | ||
| EXPORT uint32_t impl_xor_popcnt_neon_uint64( |
There was a problem hiding this comment.
There are no xor_popcnt for u32 in neon, because neon doesn't have intrinsic for 32bit popcnt
IvanPleshkov
commented
May 30, 2024
| stop_condition: impl Fn() -> bool, | ||
| ) -> Result<Self, EncodingError> { | ||
| debug_assert!(validate_vector_parameters(orig_data.clone(), vector_parameters).is_ok()); | ||
| pub trait BitsStoreType: |
There was a problem hiding this comment.
To separate dense and multivector behaviour, I defined here a generic type which is u128 for dense and u8 for multivector
IvanPleshkov
commented
May 30, 2024
| #[cfg(any(target_arch = "x86", target_arch = "x86_64"))] | ||
| if is_x86_feature_detected!("sse4.2") { | ||
| unsafe { | ||
| if v1.len() > 16 { |
There was a problem hiding this comment.
For large-dim multivector use more efficient simd implementation
IvanPleshkov
commented
May 30, 2024
|
|
||
| fn encode_vector(vector: &[f32]) -> EncodedBinVector<TBitsStoreType> { | ||
| let mut encoded_vector = | ||
| vec![Default::default(); TBitsStoreType::get_storage_size(vector.len())]; |
There was a problem hiding this comment.
Just realised that for u128 we may have the wrong alignment here. Need additional investigation but it's not a part of this PR
generall
approved these changes
May 30, 2024
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.
For multivector 128-dim alignment is too large. This PR introduces generic BQ storage with
u128andu8alignment.In
u128case there are no any changes for backward compatibility. Inu8case there is a flexible alignment: ifdim >= 128, alignment is 128 and simd is optimized for 128-bit alignment (the same for 64 and 32).