chloro-pn commented on code in PR #6178:
URL: https://github.com/apache/arrow-rs/pull/6178#discussion_r1709725449


##########
arrow-select/src/filter.rs:
##########
@@ -707,6 +710,62 @@ fn filter_byte_view<T: ByteViewType>(
     GenericByteViewArray::from(unsafe { builder.build_unchecked() })
 }
 
+fn filter_fixed_size_binary(
+    array: &FixedSizeBinaryArray,
+    predicate: &FilterPredicate,
+) -> FixedSizeBinaryArray {
+    let values: &[u8] = array.values();
+    let value_length = array.value_length() as usize;
+    let calcualte_offset_from_index = |index: usize| index * value_length;
+    let buffer = match &predicate.strategy {
+        IterationStrategy::SlicesIterator => {
+            let mut buffer = MutableBuffer::with_capacity(predicate.count * 
value_length);
+            for (start, end) in SlicesIterator::new(&predicate.filter) {
+                buffer.extend_from_slice(
+                    
&values[calcualte_offset_from_index(start)..calcualte_offset_from_index(end)],
+                );
+            }
+            buffer
+        }
+        IterationStrategy::Slices(slices) => {
+            let mut buffer = MutableBuffer::with_capacity(predicate.count * 
value_length);
+            for (start, end) in slices {
+                buffer.extend_from_slice(
+                    
&values[calcualte_offset_from_index(*start)..calcualte_offset_from_index(*end)],
+                );
+            }
+            buffer
+        }
+        IterationStrategy::IndexIterator => {
+            let iter = IndexIterator::new(&predicate.filter, 
predicate.count).map(|x| {
+                
&values[calcualte_offset_from_index(x)..calcualte_offset_from_index(x + 1)]
+            });
+
+            // SAFETY: IndexIterator is trusted length
+            unsafe { MutableBuffer::from_trusted_len_iter_slice_u8(iter, 
value_length) }

Review Comment:
   However, for a single memory copy operation, the compiler can copy multiple 
bytes simultaneously by generating SIMD instructions.
   For example, if you call memcpy once and copy four bytes, it is always 
better or equivalent to calling memcpy four times and copying one byte each 
time. Because the first scenario may be optimized by SIMD while the second 
scenario is not.
   
   I am not sure if Rust's compiler has implemented this optimization just like 
C/C++ libc.



-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to