[HIVE-10235] Loop optimization for SIMD in ColumnDivideColumn.txt

• Type: Sub-task
• Status: Closed
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 1.1.0
• Fix Version/s: 1.3.0, 2.0.0
• Component/s: Vectorization
• Assignee: Chengxiang Li
• Reporter: Chengxiang Li
• Created: 07/Apr/15 06:14
• Updated: 16/Feb/16 23:51
• Resolved: 30/Apr/15 02:08

Description

Found two loop which could be optimized for packed instruction set during execution.
1. hasDivBy0 depends on the result of last loop, which prevent the loop be executed vectorized.

for(int i = 0; i != n; i++) {
      <OperandType2> denom = vector2[i];
      outputVector[i] = vector1[0] <OperatorSymbol> denom;
      hasDivBy0 = hasDivBy0 || (denom == 0);
}

2. same as HIVE-10180, vector2[0] reference provent JVM optimizing loop into packed instruction set.

for(int i = 0; i != n; i++) {
      outputVector[i] = vector1[i] <OperatorSymbol> vector2[0];
}

Issue Links

• links to: https://reviews.apache.org/r/33119