Js compilation falsy values slides
- 1. JS compilation hot or not? Zbigniew Braniecki (gandalf@mozilla.com)
- 2. let’s talk about benchmarks
- 3. 5 categories of benchmarks
- 4. 5 categories of benchmarks • Real applications
- 5. 5 categories of benchmarks • Real applications • Modified applications (eg. with I/O removed to make it CPU-bound)
- 6. 5 categories of benchmarks • Real applications • Modified applications (eg. with I/O removed to make it CPU-bound) • Kernels (key fragments of real applications)
- 7. 5 categories of benchmarks • Real applications • Modified applications (eg. with I/O removed to make it CPU-bound) • Kernels (key fragments of real applications) • Toy benchmarks (eg. sieve of Erastosthenes)
- 8. 5 categories of benchmarks • Real applications • Modified applications (eg. with I/O removed to make it CPU-bound) • Kernels (key fragments of real applications) • Toy benchmarks (eg. sieve of Erastosthenes) • Synthetic benchmarks (code created artificially to fit a profile of particular operations, e.g Dhrystone)
- 9. statistics
- 10. statistics • How common is average?
- 11. statistics • How common is average? • Biases (sampling, memory etc.)
- 12. statistics • How common is average? • Biases (sampling, memory etc.) • Measuring the right thing is hard
- 13. statistics • How common is average? • Biases (sampling, memory etc.) • Measuring the right thing is hard • Reducing complexity
- 14. History How did browser JS performance improve over time
- 15. V8
- 16. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 17. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 18. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 19. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 20. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 21. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 22. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 23. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 24. V8 5000 3750 2500 1250 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 25. SunSpider
- 26. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 27. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 28. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 29. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 30. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 31. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 32. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 33. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 34. SunSpider 20000 15000 10000 5000 0 1.0 1.5 2.0 3.0 3.5 3.6 4 Nightly-JS
- 35. SunSpider
- 36. SunSpider 3000 2250 1500 750 0 3.0 3.5 3.6 4.0 Nightly-JS
- 37. SunSpider 3000 2250 1500 750 0 3.0 3.5 3.6 4.0 Nightly-JS
- 38. SunSpider 3000 2250 1500 750 0 3.0 3.5 3.6 4.0 Nightly-JS
- 39. SunSpider 3000 2250 1500 750 0 3.0 3.5 3.6 4.0 Nightly-JS
- 40. SunSpider 3000 2250 1500 750 0 3.0 3.5 3.6 4.0 Nightly-JS
- 41. SunSpider 3000 2250 1500 750 0 3.0 3.5 3.6 4.0 Nightly-JS
- 42. Kraken
- 43. Kraken 50000 37500 25000 12500 0 3.0 3.5 3.6 4.0 Nightly-JS
- 44. Kraken 50000 37500 25000 12500 0 3.0 3.5 3.6 4.0 Nightly-JS
- 45. Kraken 50000 37500 25000 12500 0 3.0 3.5 3.6 4.0 Nightly-JS
- 46. Kraken 50000 37500 25000 12500 0 3.0 3.5 3.6 4.0 Nightly-JS
- 47. Kraken 50000 37500 25000 12500 0 3.0 3.5 3.6 4.0 Nightly-JS
- 48. Kraken 50000 37500 25000 12500 0 3.0 3.5 3.6 4.0 Nightly-JS
- 49. Types of JIT method vs. tracing
- 50. method JIT heap 10 9 8 7 6 5 4 3 2 1 0 a b
- 51. method JIT heap 10 9 8 7 6 5 4 3 2 1 0 a b
- 52. method JIT heap 10 9 8 7 6 5 4 3 2 1 0 a b
- 53. method JIT heap 10 9 8 7 6 5 4 3 2 1 0 a b
- 54. method JIT heap 10 9 8 7 6 5 4 3 2 1 0 a b
- 55. method JIT heap 10 9 8 7 6 5 4 3 2 1 0 a b
- 56. method JIT heap hot loop! 10 9 8 7 6 5 4 3 2 1 0 a b
- 57. tracing JIT heap a 10 9 8 7 6 5 4 3 2 1 0 a
- 58. tracing JIT heap b 10 9 8 7 6 5 4 3 2 1 0 a b
- 59. tracing JIT heap b 10 9 8 7 6 5 4 3 2 1 0 hot trace! a b
- 60. How browsers compile stuff • Chakra (IE) - method • Carakan (Opera) - method • Nitro (Safari) - method • V8 (Chrome) - method • JägerMonkey (Firefox) - trace + method
- 62. How to help JIT hot loop! b hot trace!
- 63. What to avoid • competing heaps • variable type changing • evals • looong methods
- 64. Future • Dead code elimination • Type Interface • Function inlining • CrankShaft (Chrome) and IonMonkey (Firefox)
- 65. Questions?
- 66. Thank You!
Editor's Notes
- #2: \n
- #3: \n
- #4: \n
- #5: \n
- #6: \n
- #7: \n
- #8: \n
- #9: \n
- #10: \n
- #11: \n
- #12: \n
- #13: \n
- #14: \n
- #15: \n
- #16: \n
- #17: \n
- #18: \n
- #19: \n
- #20: \n
- #21: \n
- #22: \n
- #23: \n
- #24: \n
- #25: \n
- #26: \n
- #27: \n
- #28: \n
- #29: \n
- #30: \n
- #31: \n
- #32: \n
- #33: \n
- #34: \n
- #35: \n
- #36: \n
- #37: \n
- #38: \n
- #39: \n
- #40: \n
- #41: \n
- #42: \n
- #43: \n
- #44: \n
- #45: \n
- #46: \n
- #47: \n
- #48: \n
- #49: \n
- #50: \n
- #51: \n
- #52: \n
- #53: \n
- #54: - heaps cost memory\n - compilation takes time\n
- #55: - method - well known, tested\n - tracing - very new, Mozilla, LuaJIT and PyPy the first ones to try it out\n
- #56: - fine tuning, compilation vs. time, compression vs. memory etc.\n
- #57: - small functions\n - reduce branching\n - keep var types (int vs. float!)\n
- #58: \n
- #59: \n
- #60: \n
- #61: \n