Code Duplication in the Graal Compiler

Transcript

1. CODE DUPLICATION IN THE GRAAL COMPILER Virtual Machine Meetup (VMM), Prague 2017 David Leopoldseder, JKU Linz, Austria Lukas Stadler, Oracle Labs, Linz - Austria Thomas Würthinger, Oracle Labs, Zurich - CH 1
2. 2 DUPLICATION IN GRAAL Duplication
3. 3 DUPLICATION IN GRAAL Inlining Duplication
4. 4 DUPLICATION IN GRAAL Inlining Duplication Loop Optimizations Loop Unrolling Loop Unswitching Loop Peeling Loop Inversion
5. 5 DUPLICATION IN GRAAL Duplication Simulation Duplication Inlining Loop Optimizations Loop Unrolling Loop Unswitching Loop Peeling Loop Inversion
6. 6 DUPLICATION IN GRAAL Duplication Inlining Loop Optimizations Loop Unrolling Loop Unswitching Loop Peeling Loop Inversion Duplication Simulation
7. 7 MOTIVATING EXAMPLE int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; }
8. int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; } 8 MOTIVATING EXAMPLE Fastpath Slowpath
9. int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; } 9 MOTIVATING EXAMPLE Fastpath Slowpath Expensive operation
10. int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; } 10 MOTIVATING EXAMPLE Fastpath Slowpath Expensive operation p=2 in the false branch
11. int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; } 11 MOTIVATING EXAMPLE Fastpath Slowpath Expensive operation p=2 in the false branch Strength Reduction if we know p == 2 and x.length >= 0 x.length / 2  x.length >> 1
12. int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; } 12 MOTIVATING EXAMPLE Fastpath Slowpath Expensive operation p=2 in the false branch Merge is an optimization boundary Strength Reduction if we know p == 2 and x.length >= 0 x.length / 2  x.length >> 1
13. int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; } 13 MOTIVATING EXAMPLE Fastpath Slowpath Expensive operation p=2 in the false branch Merge is an optimization boundary Only allowed if isPowerOf2(a) == true Strength Reduction if we know p == 2 and x.length >= 0 x.length / 2  x.length >> 1
14. int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; } 14 MOTIVATING EXAMPLE Fastpath Slowpath Expensive operation p=2 in the false branch Merge is an optimization boundary Solution  Duplication Strength Reduction if we know p == 2 and x.length >= 0 x.length / 2  x.length >> 1
15. 15 MOTIVATING EXAMPLE – WHAT WE WANT IN THE END int f(int a, int b, int[] x) { if (a > b) { return x.length / a; } else { return x.length >> 1; } }
16. 16 MOTIVATING EXAMPLE – WHAT WE WANT IN THE END int f(int a, int b, int[] x) { if (a > b) { return x.length / a; } else { return x.length >> 1; } } Slowpath Fastpath Approx. 15-90x faster on Nehalem
17. 17 ENABLED OPTIMIZATIONS Read Eliminations Escape Analysis Opportunities Canonicalizations Conditional Eliminations Check Eliminations (Devirtualization) (Partial Redundancy Eliminations)
18. 18 PROBLEMS TO SOLVE
19. 19 PROBLEMS TO SOLVE P1 We need to determine which duplications will increase peak performance
20. 20 PROBLEMS TO SOLVE P1 We need to determine which duplications will increase peak performance P2 We need to know which optimizations are enabled by a certain duplication
21. 21 PROBLEMS TO SOLVE P1 We need to determine which duplications will increase peak performance P2 We need to know which optimizations are enabled by a certain duplication P3 Finding those optimization opportunities after duplication is compile time intensive, therefore, we need to find a way to perform this kind of analysis in acceptable time in a JIT compiler
22. 22 APPROACH Simulate a duplication per merge Constant Folding Trade-off every possible duplication Sort by • Benefit • Cost • Probability Duplicate & Optimize Conditional Elimination PEA & Scalar Replacement Strength Reduction Duplicate Optimize Initial IR Optimization Potential For each Duplication Beneficial Duplications Optimized IR Decide if duplication is beneficial
23. 23 APPROACH Simulate a duplication per merge Constant Folding Trade-off every possible duplication Sort by • Benefit • Cost • Probability Duplicate & Optimize Conditional Elimination PEA & Scalar Replacement Strength Reduction Duplicate Optimize Initial IR Optimization Potential For each Duplication Beneficial Duplications Optimized IR Decide if duplication is beneficial
24. 24 APPROACH Simulate a duplication per merge Constant Folding Trade-off every possible duplication Sort by • Benefit • Cost • Probability Duplicate & Optimize Conditional Elimination PEA & Scalar Replacement Strength Reduction Duplicate Optimize Initial IR Optimization Potential For each Duplication Beneficial Duplications Optimized IR Decide if duplication is beneficial
25. 25 APPROACH Simulate a duplication per merge Constant Folding Trade-off every possible duplication Sort by • Benefit • Cost • Probability Duplicate & Optimize Conditional Elimination PEA & Scalar Replacement Strength Reduction Duplicate Optimize Initial IR Optimization Potential For each Duplication Beneficial Duplications Optimized IR Decide if duplication is beneficial
26. 26 APPROACH Simulate a duplication per merge Constant Folding Trade-off every possible duplication Sort by • Benefit • Cost • Probability Duplicate & Optimize Simulation Tier Trade-off Tier Optimization Tier Conditional Elimination PEA & Scalar Replacement Strength Reduction Duplicate Optimize Initial IR Optimization Potential For each Duplication Beneficial Duplications Optimized IR Decide if duplication is beneficial
27. 27 RECAP MOTIVATING EXAMPLE int f(int a, int b, int[] x) { int p; if (a > b) { p = a; } else { p = 2; } return x.length / p; }
28. 28 APPROACH int p; if (a > b) p = a; p = 2; return x.length / p;
29. 29 DOMINATOR TREE int p; if (a > b) p = a; p = 2; return x.length / p; Merge Block
30. 30 DUPLICATION SIMULATION int p; if (a > b) p = a; p = 2; return x.length / p; Copy Merge Block return x.length / p; return x.length / p;
31. 31 DUPLICATION SIMULATION int p; if (a > b) p = a; p = 2; return x.length / p; Simulate Dominance Relation return x.length / p; return x.length / p;
32. int p; if (a > b) p = a; p = 2; return x.length / p; Copy Propagation return x.length / p; return x.length / p; a 2 32 DUPLICATION SIMULATION
33. 33 DUPLICATION SIMULATION int p; if (a > b) return x.length / p; return x.length / 2; return x.length / a; Apply Optimizations
34. 34 DUPLICATION SIMULATION int p; if (a > b) return x.length / p; return x.length / 2; return x.length / a; Strength Reduction x.length >> 1
35. 35 BENEFIT / COST -> TRADE OFF int p; if (a > b) return x.length / p; return x.length >> 1; return x.length / a;
36. 36 BENEFIT / COST -> TRADE OFF int p; if (a > b) return x.length / p; return x.length >> 1; return x.length / a; Benefit  (Latency(Div) - Latency(Shift)) * Probability = 31 * 0.9 = 27.9
37. int p; if (a > b) return x.length / p; return x.length >> 1; return x.length / a; 37 BENEFIT / COST -> TRADE OFF Benefit  (Latency(Div) - Latency(Shift)) * Probability = 31 * 0.9 = 27.9 Cost  1 Additional Return (+ 4 Instructions) + 1 Additional Shift + 1 Additional Read = 6 6 – 1 Jump from branch = 5 Instructions
38. EVALUATION 38
39. SCALA DACAPO 39
40. 40 GRAAL JS OCTANE
41. THANK YOU ! – Q/A 41