5035-Pipeline-Optimization-Techniques.pdf
- 1. Overview Locating the bottleneck Performance measurements Optimizations Balancing the pipeline Other optimizations: multi-processing, parallel processing ITCS 4010/5010:Game Engine Design 1 Pipeline Optimization Pipeline Optimization “We should forget about small efficiencies, say about 97% of the time: premature optimization is the root of all evil” – Donald Knuth Make it run first, then optimize But only optimize where it makes any difference Pipeline Optimization: Process to maximize the rendering speed, then allow stages that are not bottlenecks to consume as much time as the bottleneck. ITCS 4010/5010:Game Engine Design 2 Pipeline Optimization Pipeline Optimization Stages execute in parallel Always the slowest stage is the bottleneck of the pipeline The bottleneck determines throughput (i.e., maximum speed) The bottleneck is the average bottleneck over a frame Cannot measure intra-frame bottlenecks easily Bottlenecks can change over a frame Most important: find bottleneck, then optimize that stage! ITCS 4010/5010:Game Engine Design 3 Pipeline Optimization Locating the Bottleneck Two bottleneck location techniques: Technique 1: ◦ Make a certain stage work less ◦ If performance is the better, then that stage is the bottleneck Technique 2: ◦ Make the other two stages work less or (better) not at all ◦ If performance is the same, then the stages not included above is the bottleneck Complication: the bus between CPU and graphics card may be bot- tleneck (not a typical stage) ITCS 4010/5010:Game Engine Design 4 Pipeline Optimization
- 2. Application (CPU) Stage the Bottleneck? Use top, osview command on Unix, TaskManager on Windows. If app uses (near) 100% of CPU time, then very likely application is the bottleneck Using a code profiler is safer. Make CPU do less work (e.g., turn off collision-detection) Replace glVertex and glNormal with glColor Makes the geometry and rasterizer do almost nothing No vertices to transform, no normals to compute lighting for, no tri- angles to rasterize If performance does not change, program is CPU-bound, or CPU- limited ITCS 4010/5010:Game Engine Design 5 Pipeline Optimization Geometry Stage the Bottleneck? Trickiest stage to test Why? Change in geometry workload usually changes application and rasterizer workload. Number of light sources only affects geometry stage: ◦ Disable light sources (vertex shaders can make this simple). ◦ If performance goes up, then geometry is bottleneck, and pro- gram transform-limited Alternately, enable all light sources; if performance stays the same, geometry stage NOT the bottleneck Alternately, test CPU and rasterizer instead ITCS 4010/5010:Game Engine Design 6 Pipeline Optimization Rasterizer Stage the Bottleneck? The easiest, and fastest to test Simply, decrease the size of the window you render to ◦ Does not change app. or geometry workload ◦ But rasterizer needs to fill fewer pixels ◦ If the performance goes up, then program is “fill-limited” or “fill- bound” Make rasterizer work less: Turn of texturing, fog, blending, depth buffering etc (if your architecture have performance penalties for these) ITCS 4010/5010:Game Engine Design 7 Pipeline Optimization Optimization Optimize the bottleneck stage Only put enough effort, so that the bottleneck stage moves Did you get enough performance? ◦ Yes! Quit optimizing ◦ NO! Continute optimizing the (possibly new) bottleneck If close to maximum speed of system, might need to turn to acceler- ation techniques (spatial data structures, occlusion culling, etc) ITCS 4010/5010:Game Engine Design 8 Pipeline Optimization
- 3. Illustrating Optimization Height of bar: time it takes for that stage for one frame Highest bar is bottleneck After optimization: bottleneck has moved to APP No use in optimizing GEOM, turn to optimizing APP instead ITCS 4010/5010:Game Engine Design 9 Pipeline Optimization Application Stage Optimization Initial Steps: ◦ Turn on optimiziation flags in compiler ◦ Use code profilers, shows places where majority of time is spent ◦ This is time consuming stuff Strategy 1: Efficient code ◦ Use fewer instructions ◦ Use more efficient instructions ◦ Recode algorithmically Strategy 2: Efficient memory access ITCS 4010/5010:Game Engine Design 10 Pipeline Optimization Appliction:Code Optimization Tricks SIMD intstructions sets perfect for vector ops ◦ 2-4 operations in parallell ◦ SSE, SSE2, 3DNow! are examples Division is an expensive operation ◦ Between 4-39 times slower than most other instructions ◦ Good usage Example: vector normalization: Instead of v = (vx/d, vy/d, vz/d) Do d = v · v, f = 1/d, v = v ∗ f On some CPUs there are low-precision versions of (1/x) and square root reciprocal (1/ √ x) ITCS 4010/5010:Game Engine Design 11 Pipeline Optimization Code Optimization Tricks (contd) Conditional branches are generally expensive; ◦ Avoid if-then-else if possible ◦ Sometimes branch prediction on CPUs works remarkably well Math functions (sin, cos, tan, sqrt, exp, etc.) are expensive ◦ Rough approximation might be sufficient ◦ Can use first few terms in Taylor series Inline code is good (avoids function calls) float (32 bits) is faster than double (64 bits); less data is sent down the pipeline ITCS 4010/5010:Game Engine Design 12 Pipeline Optimization
- 4. Code Optimization Tricks (contd) Compiler optimization: Hard to predict: –counter vs. counter– Use const in C and C++ to help to compiler with optimization Following often incur overhead: ◦ Dynamic casting (C++) ◦ Virtual methods ◦ Inherited constructors ◦ Passing structs by value ITCS 4010/5010:Game Engine Design 13 Pipeline Optimization Memory Optimization Memory hierarchies (caches) in modern computers - primary, sec- ondary caches. Bad memory access pattern can ruin performance Not really about using less memory, though that can help ITCS 4010/5010:Game Engine Design 14 Pipeline Optimization Memory Optimization Tricks Sequential access: Store data in order in memory: ◦ Tex Coords #0, Position #0, Tex Coords #1, Position #1, Tex coords #2, Position #2, etc. Cache prefetching is good, but hard to control malloc() and free() may be slow: Consider using a custom storage allocator - allocate memory to a pool at startup ITCS 4010/5010:Game Engine Design 15 Pipeline Optimization Memory Optimization Tricks (contd) Align data with size of cache line ◦ Example: on most Pentiums, the cache line size if 32 bytes ◦ Now, assume that it takes 30 bytes to store a vertex ◦ Padding with another 2 bytes to 32 bytes will likely perform bet- ter. Following pointers (linked list) is expensive (if memory is allocated arbitrarily) ◦ Does not use coherence well that cache usually exploits ◦ That is, the address after the one we just used is likely to be used soon ◦ Paper by Smits on ray tracing shows this. ITCS 4010/5010:Game Engine Design 16 Pipeline Optimization
- 5. Geometry Stage: Optimization Geometry stage does per-vertex ops ◦ Best way to optimize: Use Triangle strips!!! Lighting optimization: ◦ Spot lights expensive, point light cheaper, directional light cheapest ◦ Disable lighting if possible ◦ Use as few light sources as possible ◦ If you use 1/d2 fallof, then if d 10 (example), disable light ITCS 4010/5010:Game Engine Design 17 Pipeline Optimization Geometry Stage: Optimization Normals must be normalized to get correct lighting ◦ Normalize them as a preprocess, and disable normalizing if pos- sible Lighting can be computed for both sides of a triangle; disable if not needed. If light sources are static with respect to geometry, and material is only diffuse ◦ Precompute lighting on CPU ◦ Send only precomputed colors (not normals) ITCS 4010/5010:Game Engine Design 18 Pipeline Optimization Raster Stage: Optimization Rasterizer stage does per-pixel ops Simple Optimization: turn on backface culling if possible Turn off Z-buffering if possible: ◦ Example: after screen clear, draw large background polygon ◦ Using polygon-aligned BSP trees Draw in front-to-back order Try disable features: texture filtering mode, fog, blending, multisam- pling ITCS 4010/5010:Game Engine Design 19 Pipeline Optimization Raster Stage: Optimization To make rasterization faster, need to rasterize fewer (or cheaper) pixels: ◦ Make window smaller ◦ Render to a smaller texture, and then enlarge texture onto screen Depth complexity is number of times a pixel has been written to ◦ Good for understanding behaviour of application ITCS 4010/5010:Game Engine Design 20 Pipeline Optimization
- 6. Depth Complexity ITCS 4010/5010:Game Engine Design 21 Pipeline Optimization Overall Optimization: General Techniques Reduce number of primitives, eg. using polygon simplification algo- rithms Preprocess geometry and data for the particular architecture Turn off features not in use such as: ◦ Depth buffering, Blending, Fog, Texturing ITCS 4010/5010:Game Engine Design 22 Pipeline Optimization Overall Optimization (contd) Minimize state changes by grouping objects ◦ Example: objects with the same texture should be rendered to- gether If all pixels are always drawn, avoid color buffer clear Frame buffer reads are expensive Display lists may work faster Precompile a list of primitives for faster rendering OpenGL API supports this ITCS 4010/5010:Game Engine Design 23 Pipeline Optimization Balancing the Pipeline The bottleneck stage sets the frame rate The other two stages will be idle for some time Also, to sync with monitor, there might be idle time for all stages Exploit this time to make quality of images better if possible ITCS 4010/5010:Game Engine Design 24 Pipeline Optimization
- 7. Balancing the Pipeline Increase number of triangles (affects all stages) More lights, more expensive (geometry) More realistic animation, more accurate collision detection (applica- tion) More expensive texture filtering, blending, etc. (rasterizer) If not fill-limited, increase window size Note: there are FIFOs between stages (and at many other places too) to smooth out idleness of stages More techniques in text. ITCS 4010/5010:Game Engine Design 25 Pipeline Optimization Multiprocessing Use this if application is bottleneck, and is affordable Two major ways: (1) Multiprocessor pipelining, (2) Parallel process- ing ITCS 4010/5010:Game Engine Design 26 Pipeline Optimization Summary Pipeline optimization is no substitute for good algorithms! Do optimization as a last step. Primarily for products that should be shipped Most often good to use triangle strips! ITCS 4010/5010:Game Engine Design 27 Pipeline Optimization