[ARROW-5221] Improvement the performance of class SegmentsUtil

• Type: Improvement
• Status: Closed
• Priority: Minor
• Resolution: Invalid
• External issue URL: https://github.com/apache/arrow/issues/21695
• Assignee: Liya Fan
• Reporter: Liya Fan
• Created: 26/Apr/19 09:36
• Updated: 11/Jan/23 07:39
• Resolved: 26/Apr/19 09:38

Description

Improve the performance of class SegmentsUtil from two points:

1. In method allocateReuseBytes, the generated byte array should be cached for reuse, if the size does not exceed MAX_BYTES_LENGTH. However, the array is not cached if bytes.length < length, and this will lead to performance overhead:

 

if (bytes == null) {
  if (length <= MAX_BYTES_LENGTH)

{     bytes = new byte[MAX_BYTES_LENGTH];     BYTES_LOCAL.set(bytes);   }

else

{     bytes = new byte[length];   }

} else if (bytes.length < length)

{   bytes = new byte[length]; }

 

2. To evaluate the offset, an integer is bitand with a mask to clear to low bits, and then shift right. The bitand is useless:

 

((index & BIT_BYTE_POSITION_MASK) >>> 3)