perf: Branchless Parquet Prefiltering

[coastalwhite]

Change the Parquet decoders to use branchless filtering for the prefiltered methods.

This rewrites most of the decoders.

[orlp]

I think this loop can be written more efficiently.

let num_valid = validity.count_ones();
let mut num_written = 0;

while filter != 0 {
    let offset = filter.trailing_zeros();
    validity >>= offset;
    
    let idx = num_valid - validity.count_ones();
    let v = if validity & 1 != 0 {
        dfn.decode(unsafe { values.get_unchecked(idx) })
    } else {
        T::zeroed()
    };

    *out.add(num_written) = v;

    num_written += 1;
    filter >>= offset + 1;
    validity >>= 1;
}

(num_valid, num_written)

[orlp]

Even better I believe:

let num_available = validity.count_ones();
let mut num_written = 0;

while filter != 0 {
    let offset = filter.trailing_zeros();
    let idx_to_read = num_available - (validity >> offset).count_ones();
    let v = if validity & 1 != 0 {
        dfn.decode(unsafe { values.get_unchecked(idx_to_read) })
    } else {
        T::zeroed()
    };

    *out.add(num_written) = v;
    num_written += 1;
    filter &= filter - 1; // Clear least significant bit.
}

(num_available, num_written)

[orlp]

Idem.

[orlp]

Idem.

[codecov]

Codecov Report

Attention: Patch coverage is 79.79107% with 561 lines in your changes missing coverage. Please review.

Project coverage is 80.16%. Comparing base (69c9c3a) to head (e395c93).
Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
...et/src/arrow/read/deserialize/fixed_size_binary.rs	69.86%	91 Missing ⚠️
...lars-parquet/src/arrow/read/deserialize/boolean.rs	62.50%	87 Missing ⚠️
...quet/src/arrow/read/deserialize/primitive/plain.rs	76.47%	80 Missing ⚠️
...lars-parquet/src/arrow/read/deserialize/binview.rs	79.58%	78 Missing ⚠️
...t/src/arrow/read/deserialize/utils/dict_encoded.rs	90.35%	68 Missing ⚠️
...rs-parquet/src/arrow/read/deserialize/utils/mod.rs	68.15%	50 Missing ⚠️
crates/polars-arrow/src/types/aligned_bytes.rs	0.00%	28 Missing ⚠️
...parquet/src/arrow/read/deserialize/nested_utils.rs	90.30%	22 Missing ⚠️
crates/polars-arrow/src/bitmap/bitmask.rs	46.15%	14 Missing ⚠️
...et/src/arrow/read/deserialize/primitive/integer.rs	82.89%	13 Missing ⚠️
... and 7 more

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #19190      +/-   ##
==========================================
+ Coverage   80.03%   80.16%   +0.12%     
==========================================
  Files        1532     1531       -1     
  Lines      210749   210193     -556     
  Branches     2442     2442              
==========================================
- Hits       168676   168494     -182     
+ Misses      41518    41144     -374     
  Partials      555      555              

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