An automatic query optimizer-compiler for Sequential and Parallel LINQ. LinqOptimizer compiles declarative LINQ queries into fast loop-based imperative code. The compiled code has fewer virtual calls and heap allocations, better data locality and speedups of up to 15x (Check the [Performance] (https://github.com/nessos/LinqOptimizer/wiki/Performance) page).
The main idea is that we lift query sources into the world of Expression trees and after various transformations-optimizations we compile them into IL for efficient execution.
var query = (from num in nums.AsQueryExpr() // lift
where num % 2 == 0
select num * num).Sum();
Console.WriteLine("Result: {0}", query.Run()); // compile and execute
For F# we support functional pipelines and support for F# style LINQ queries is in development.
let query = nums
|> Query.ofSeq
|> Query.filter (fun num -> num % 2 = 0)
|> Query.map (fun num -> num * num)
|> Query.sum
printfn "Result: %d" <| Query.run query // compile and execute
Install-Package LinqOptimizer.CSharp
Install-Package LinqOptimizer.FSharp
- Lambda inlining
- Loop fusion
- Nested loop generation
- Anonymous Types-Tuples elimination
- Specialized strategies and algorithms
The expression
var query = (from num in nums.AsQueryExpr()
where num % 2 == 0
select num * num).Sum();
will compile to
int sum = 0;
for (int index = 0; index < nums.Length; index++)
{
int num = nums[index];
if (num % 2 == 0)
sum += num * num;
}
and for the parallel case
var query = (from num in nums.AsParallelQueryExpr()
where num % 2 == 0
select num * num).Sum();
will compile to a reduce-combine style straregy
Parallel.ReduceCombine(nums, 0,
(acc, num) => {
if (num % 2 == 0)
return acc + num * num;
else
return acc;
}, (left, right) => left + right);
- Many missing operators
- New specialized operators
- Even more optimizations
LinqOptimizer draws heavy inspiration from