It Doesn't Have to Be Hard: How to Fix Your Performance Woes
1 like5,675 views
The document outlines the agenda for a presentation given at the Game Developers Conference, focusing on graphics performance optimization using Intel's tools. It covers profiling workflows, common bottlenecks in graphics rendering, and a case study on optimizing Unity's 3D Game Kit. Additionally, the presentation emphasizes effective scaling of graphics performance and the importance of performance awareness among developers.
![@IntelSoftware @IntelGraphics
Legal Notices and Disclaimers
No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.
Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well
as any warranty arising from course of performance, course of dealing, or usage in trade.
You may not use or facilitate the use of this document in connection with any infringement or other legal analysis concerning Intel products described herein. You agree to grant Intel a
non-exclusive, royalty-free license to any patent claim thereafter drafted which includes subject matter disclosed herein.
The products and services described may contain defects or errors known as errata which may cause deviations from published specifications. Current characterized errata are
available on request.
Intel technologies’ features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system
configuration. No computer system can be absolutely secure. Check with your system manufacturer or retailer or learn more at [intel.com].
Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are
measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult
other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other
products. For more complete information visit www.intel.com/benchmarks.
Optimization Notice: Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These
optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on
microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not
specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific
instruction sets covered by this notice.
Results have been estimated or simulated using internal Intel analysis or architecture simulation or modeling, and provided to you for informational purposes. Any differences in your
system hardware, software or configuration may affect your actual performance.
Intel, Core and the Intel logo are trademarks of Intel Corporation in the U.S. and/or other countries.
*Other names and brands may be claimed as the property of others
© Intel Corporation.](https://image.slidesharecdn.com/05-it-doesn-not-have-to-be-hard-how-to-fix-your-performance-woes-190405185658/85/It-Doesn-t-Have-to-Be-Hard-How-to-Fix-Your-Performance-Woes-65-320.jpg)
It Doesn't Have to Be Hard: How to Fix Your Performance Woes
- 1. GameDevelopers CONFERENCE Carlos A Dominguez Caballero (Intel®)
- 2. @IntelSoftware @IntelGraphics 2 Agenda • Introduction • Game profiling workflow overview • Step through profiling workflow • Common bottleneck identification • Case Study: Optimizing Unity’s 3D Game Kit
- 3. @IntelSoftware @IntelGraphics 3 Scaling Graphics Performance Effectively My super awesome game 30 FPS
- 4. @IntelSoftware @IntelGraphics 4 Scaling Graphics Performance Effectively My super awesome game 30 FPS 10 FPS
- 5. @IntelSoftware @IntelGraphics 5 Scaling Graphics Performance Effectively My super awesome game 30 FPS 10 FPS 20 FPS
- 6. @IntelSoftware @IntelGraphics 6 Scaling Graphics Performance Effectively My super awesome game 30 FPS 10 FPS 20 FPS 30 FPS
- 7. @IntelSoftware @IntelGraphics 7 Scaling Graphics Performance Effectively My super awesome game 30 FPS 10 FPS 30 FPS+ 60 FPS+
- 8. @IntelSoftware @IntelGraphics 8 Intel® Graphics Performance Analyzers (Intel GPA) System Analyzer Graphics Frame Analyzer Graphics Trace Analyzer
- 9. @IntelSoftware @IntelGraphics GPU bound 9 Profiling Workflow Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots Identify Hotspots via instrumentation CPU bound
- 10. @IntelSoftware @IntelGraphics GPU bound 10 Profiling Workflow Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 11. @IntelSoftware @IntelGraphics GPU bound 11 Profiling Workflow Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 12. @IntelSoftware @IntelGraphics 12 System Analyzer Overview
- 13. @IntelSoftware @IntelGraphics 13 In Application Live Analysis
- 14. @IntelSoftware @IntelGraphics GPU bound 14 Profiling Workflow Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 15. @IntelSoftware @IntelGraphics 15 • New UI with improved collection mechanism • Collect data from various sources correlated in one timeline • ETW events • Graphics API calls • ITT user markers • Sampled metrics Graphics Trace Analyzer
- 16. @IntelSoftware @IntelGraphics Hardware Queue Graphics API calls Driver Queue 16 Lifetime of Graphics Commands
- 17. @IntelSoftware @IntelGraphics Hardware Queue Graphics API calls Driver Queue 17 Lifetime of Graphics Commands Present call in queue until finally executed by GPU
- 18. @IntelSoftware @IntelGraphics Hardware Queue Graphics API calls Driver Queue 18 Lifetime of Graphics Commands No gaps in queue
- 19. @IntelSoftware @IntelGraphics Hardware Queue Graphics API calls Driver Queue 19 Not GPU bound
- 20. @IntelSoftware @IntelGraphics Hardware Queue Graphics API calls Driver Queue 20 Not GPU bound Gaps in queue
- 21. @IntelSoftware @IntelGraphics 21 VSync Bound
- 22. @IntelSoftware @IntelGraphics 22 GPU bound vs CPU bound comparison Consistent stream in GPU queue? Frame Time < VSync? Consistent gaps in GPU queue? Yes No GPU bound Yes VSync bound Yes CPU bound No CombinationNo Not GPU Bound GPU Bound
- 23. @IntelSoftware @IntelGraphics 23 Differentiate GPU Usage per Application
- 24. @IntelSoftware @IntelGraphics GPU bound 24 Profiling Workflow Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 25. @IntelSoftware @IntelGraphics GPU bound 25 CPU Bound Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 26. @IntelSoftware @IntelGraphics 26 Instrumentation Instrumentation Alternatives • ITT • ETW • PIX/Graphics API markers Hotspots from selected region Dive into region occurrences
- 27. @IntelSoftware @IntelGraphics GPU bound 27 Profiling Workflow Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 28. @IntelSoftware @IntelGraphics GPU bound 28 GPU Bound Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 29. @IntelSoftware @IntelGraphics 29 Graphics Frame Analyzer
- 30. @IntelSoftware @IntelGraphics 30 Multiframe Capture Technical Preview New in Intel® GPA 2019 R1 Release • Difficult problems for single frame analysis • Debug intermittent glitches and frame hitches • Profile multi-frame algorithms • Single frame capture workflow still available
- 31. @IntelSoftware @IntelGraphics 31 Multiframe Capture Technical Preview New in Intel® GPA 2019 R1 Release • Difficult problems for single frame analysis • Debug intermittent glitches and frame hitches • Profile multi-frame algorithms • Single frame capture workflow still available
- 32. @IntelSoftware @IntelGraphics GPU bound 32 GPU Bound • Frame Analysis to identify most expensive calls • Easy hotspot analysis for automatic bottleneck root cause analyzer Identify HotspotsFrame Analysis
- 33. @IntelSoftware @IntelGraphics 33 Find most expensive calls Bottom up approach
- 34. @IntelSoftware @IntelGraphics 34 Find most expensive calls Top down approach
- 35. @IntelSoftware @IntelGraphics 35 Hotspot Analysis Overview
- 36. @IntelSoftware @IntelGraphics 36 • Bottleneck in Geometry Transformation • Things to look out for and mitigation: • Dense geometry rendered to small area • More aggressive LODs or decimation needed • Icebergs (5% in view, 95% out) • Slice geometry for better culling • Stragglers (not in view but not culled) • Debug why is object not being culled Geometry Transformation
- 37. @IntelSoftware @IntelGraphics • X Axis: Pixels Rendered • Y Axis: Primitive Count • Tall skinny bars, big objects rendered to small amount of pixels 37 Geometry Transformation
- 38. @IntelSoftware @IntelGraphics 38 Geometry Transformation Iceberg Stragglers Dense/Iceberg
- 39. @IntelSoftware @IntelGraphics • New Python plugin system • =metric(“GPU Duration”, “>”, 200) • Preloaded with several plugins • Create your own! 39 Geometry Transformation
- 40. @IntelSoftware @IntelGraphics GPU bound 40 Profiling Workflow Frame Analysis Identify Scene + + CPU vs GPU bound? Identify Hotspots CPU bound Identify Hotspots via instrumentation
- 41. @IntelSoftware @IntelGraphics 41 Case Study: Unity 3D Game Kit Overview Tutorial project in Unity asset store 3 built-in quality settings ‘Performance’, ‘Balanced’ and ‘Fantastic’ Our Goal: ‘Performance’ mode FPS on ‘Fantastic’ mode visual quality
- 42. @IntelSoftware @IntelGraphics 42 Case Study: Unity 3D Game Kit Bridge the gap What makes ‘Fantastic’ look much better than ‘Performance’?
- 43. @IntelSoftware @IntelGraphics 43 Case Study: Unity 3D Game Kit Set your performance goals Draw distance Water reflections Shadows What makes ‘Fantastic’ look much better than ‘Performance’?
- 44. @IntelSoftware @IntelGraphics 44 Case Study: Unity 3D Game Kit Spoiler Draw distance Water reflections Shadows What makes ‘Fantastic’ look much better than ‘Performance’?
- 45. @IntelSoftware @IntelGraphics 45 Case Study: Unity 3D Game Kit How did we get there? Tug of war, make some room then bring stuff back in
- 46. @IntelSoftware @IntelGraphics 46 Case Study: Unity 3D Game Kit Understand the game’s anatomy
- 47. @IntelSoftware @IntelGraphics 47 Case Study: Unity 3D Game Kit Understand the game’s anatomy Frame Budget (30 FPS)
- 48. @IntelSoftware @IntelGraphics 48 Case Study: Unity 3D Game Kit Skybox takes 3ms? ~10% of frame budget Frame Budget (30 FPS)
- 49. @IntelSoftware @IntelGraphics 49 Case Study: Unity 3D Game Kit Skybox camera set as deferred ~10% of frame budget Frame Budget (30 FPS)
- 50. @IntelSoftware @IntelGraphics 50 Case Study: Unity 3D Game Kit Skybox camera set as forward 0.8ms 2.4% of frame budget down from 10% Frame Budget (30 FPS)
- 51. @IntelSoftware @IntelGraphics 51 Case Study: Unity 3D Game Kit Keep peeling the onion! Frame Budget (30 FPS)
- 52. @IntelSoftware @IntelGraphics 52 Case Study: Unity 3D Game Kit Reflections take 4.8ms 14.5% of frame budget Frame Budget (30 FPS)
- 53. @IntelSoftware @IntelGraphics 53 Case Study: Unity 3D Game Kit 32 Reflection probes? Can we get by with less? 14.5% of frame budget Frame Budget (30 FPS)
- 54. @IntelSoftware @IntelGraphics 54 Case Study: Unity 3D Game Kit 32 to 3 reflection probes 2.1ms 6.4% of frame budget down from 14.5% Frame Budget (30 FPS)
- 55. @IntelSoftware @IntelGraphics 55 Case Study: Unity 3D Game Kit Keep peeling the onion! Frame Budget (30 FPS)
- 56. @IntelSoftware @IntelGraphics 56 Case Study: Unity 3D Game Kit UI rendering takes 1.8ms 8% of frame budget Frame Budget (30 FPS)
- 57. @IntelSoftware @IntelGraphics 57 Case Study: Unity 3D Game Kit Blur text is expensive, by how much? 8% of frame budget Frame Budget (30 FPS)
- 58. @IntelSoftware @IntelGraphics 58 Case Study: Unity 3D Game Kit UI Rendering without blur 0.2ms 0.6% of frame budget down from 8% Frame Budget (30 FPS)
- 59. @IntelSoftware @IntelGraphics 59 Case Study: Unity 3D Game Kit Level check: Running at 38 FPS now! With all the optimizations we’re now running at 38 FPS Time to start adding things back! Lets start big: Culling distanceFrame Budget (30 FPS)
- 60. @IntelSoftware @IntelGraphics 60 Case Study: Unity 3D Game Kit Restore cull distance Before 38FPS After 33FPS Still some room left Lets add it all! Frame Budget (30 FPS)
- 61. @IntelSoftware @IntelGraphics 61 Case Study: Unity 3D Game Kit Lets add it all! Before 33FPS After 28FPS 8% above budget but we got water reflections and shadows! What else can we cheaply do to stay within budget? Frame Budget (30 FPS)
- 62. @IntelSoftware @IntelGraphics 62 Case Study: Unity 3D Game Kit Downscale deferred render target Downscale deferred render target to a fraction of it’s size Upscale in final pass and render UI at full resolution Back within Frame Budget! Frame Budget (30 FPS)
- 63. @IntelSoftware @IntelGraphics 63 Summary • Scale graphics performance effectively • Performance awareness is everyone’s responsibility • Grab Intel® GPA for free @ https://software.intel.com/en-us/gpa • Want to learn more? Visit us at the Intel® booth • Try it yourself! • Twitter: @carlosadc
- 64. Questions?
- 65. @IntelSoftware @IntelGraphics Legal Notices and Disclaimers No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document. Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade. You may not use or facilitate the use of this document in connection with any infringement or other legal analysis concerning Intel products described herein. You agree to grant Intel a non-exclusive, royalty-free license to any patent claim thereafter drafted which includes subject matter disclosed herein. The products and services described may contain defects or errors known as errata which may cause deviations from published specifications. Current characterized errata are available on request. Intel technologies’ features and benefits depend on system configuration and may require enabled hardware, software or service activation. Performance varies depending on system configuration. No computer system can be absolutely secure. Check with your system manufacturer or retailer or learn more at [intel.com]. Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more complete information visit www.intel.com/benchmarks. Optimization Notice: Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice. Results have been estimated or simulated using internal Intel analysis or architecture simulation or modeling, and provided to you for informational purposes. Any differences in your system hardware, software or configuration may affect your actual performance. Intel, Core and the Intel logo are trademarks of Intel Corporation in the U.S. and/or other countries. *Other names and brands may be claimed as the property of others © Intel Corporation.
- 66. @IntelSoftware @IntelGraphics 66 • Bottleneck in Shader Execution • GPU Execution Units (EU’s) very active • Reduce shader complexity • Reduce GPRs used in shader to avoid register spilling Shader Execution
- 67. @IntelSoftware @IntelGraphics 67 • Simple Fragment Shader Experiment • Quick check for ROI on optimizing shader • Reduce shader optimization iteration time with GPA’s Live Shader Analysis • Modify shader • Replay scene • Recalculate metrics Shader Execution