MOD-13451 Add pre-computed sum to L2 query blob in QuantPreprocessor

[meiravgri]

Extended the L2 query blob to include both y_sum (sum of query values) and y_sum_squares (sum of squared query values) to support asymmetric L2 distance computation.

Motivation

L2 squared distance is computed as:

||x - y||² = Σx_i² - 2 * IP(x, y) + Σy_i²

Where the inner product with a quantized storage vector is:

IP(x, y) = min * y_sum + delta * Σ(q_i * y_i)

The y_sum term is required to compute IP(x, y), which in turn is needed for the L2 distance formula. Previously only y_sum_squares was stored in the query blob.

Changes

Query blob layout update for L2:

• Before: | query_values[dim] | y_sum_squares |
• After: | query_values[dim] | y_sum | y_sum_squares |

Query blob size for L2:

• Before: (dim + 1) * sizeof(DataType)
• After: (dim + 2) * sizeof(DataType)

Code improvements:

• Merged sum_fast() and sum_squares_fast() into a single assign_query_metadata() function
• Single pass over the query vector instead of two passes for L2
• Uses 4-accumulator loop unrolling for performance
• Added sq8::SUM_QUERY and sq8::SUM_SQUARES_QUERY metadata indices

Files Modified

• src/VecSim/spaces/computer/preprocessors.h
• src/VecSim/types/sq8.h

Notes

• IP/Cosine query blob layout remains unchanged: | query_values[dim] | y_sum |

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.13%. Comparing base (51d5247) to head (165d6bf).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #883   +/-   ##
=======================================
  Coverage   97.13%   97.13%           
=======================================
  Files         129      129           
  Lines        7611     7614    +3     
=======================================
+ Hits         7393     7396    +3     
  Misses        218      218           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.