DirectCompute Accelerated Separable Filtering

2. DirectCompute Accelerated Separable Filtering AMD‘s Favorite Effects

3. Separable Filters • Much faster than executing a box filter • Classically performed by the Pixel Shader • Consists of a horizontal and vertical pass • Source image over-sampling increases with kernel size • Shader is usually TEX instruction limited AMD‘s Favorite Effects

4. Separable? – Who Cares  • In many cases developers use this technique even though the filter may not actually be separable • Results are often still acceptable • Much faster than performing a real box filter • Accelerates many bilateral cases AMD‘s Favorite Effects

5. Typical Pipeline Steps Source RT Intermediate RT Destination RT Horizontal Pass Vertical Pass AMD‘s Favorite Effects

6. Use Bilinear HW filtering? • Bilinear filter HW can halve the number of ALU and TEX instructions • Just need to compute the correct sampling offsets • Not possible with more advanced filters • Usually because weighting is a dynamic operation • Think about bilateral cases... AMD‘s Favorite Effects

7. Where to start with DirectCompute • Is the Pixel Shader version TEX or ALU limited? • You need to know what to optimize for! • Use IHV tools to establish this • Achieving peak performance is not easy – so write a highly configurable kernel • Will allow you to easily experiment and fine tune AMD‘s Favorite Effects

8. Thread Group Shared Memory (TGSM) • TGSM can be used to reduce TEX ops • TGSM can also be used to cache results • Thus saving ALU ops too • Load a sensible run length – base this on HW wavefront/warp size (AMD = 64, NVIDIA = 32) • Choose a good common factor (multiples of 64) AMD‘s Favorite Effects

9. Kernel #1 128 threads load 128 texels • Redundant compute threads  ........... Kernel Radius 128 – ( Kernel Radius * 2 ) threads compute results AMD‘s Favorite Effects

10. Avoid Redundant Threads • Should ensure that all threads in a group have useful work to do – wherever possible • Redundant threads will not be reassigned work from another group • This would involve alot of redundancy for a large kernel diameter AMD‘s Favorite Effects

11. Kernel #2 Kernel Radius * 2 threads load 1 extra texel each 128 threads load 128 texels • No redundant compute threads  ........... Kernel Radius 128 threads compute results AMD‘s Favorite Effects

12. Multiple Pixels per Thread • Allows for natural vectorization • 4 works well on AMD HW • Doesn‘t hurt performance on scalar HW • Possible to cache TGSM reads on General Purpose Registers (GPRs) • Quartering TGSM reads - absolute winner!! AMD‘s Favorite Effects

13. Kernel #3 Kernel Radius * 2 threads load 1 extra texel each 32 threads load 128 texels • Compute threads not a multiple of 64  ........... Kernel Radius 32 threads compute 128 results AMD‘s Favorite Effects

14. Multiple Lines per Thread Group • Process multiple lines per thread group • Better than one long line • 2 or 4 works well • Improved texture cache efficiency • Compute threads back to a multiple of 64 AMD‘s Favorite Effects

15. Kernel #4 Kernel Radius * 4 threads load 1 extra texel each 64 threads load 256 texels ........... ........... Kernel Radius 64 threads compute 256 results AMD‘s Favorite Effects

16. Kernel Diameter • Kernel diameter needs to be > 7 to see a DirectCompute win • Otherwise the overhead cancels out the advantage • The larger the kernel diameter the greater the win AMD‘s Favorite Effects

17. Use Packing in TGSM • Use packing to reduce storage space required in TGSM • Only have 32k per SIMD • Reduces reads/writes from TGSM • Often a uint is sufficient for color filtering • Use SM5.0 instructions f32tof16(), f16tof32() AMD‘s Favorite Effects

18. High Definition Ambient Occlusion Depth + Normals = * HDAO buffer Original Scene Final Scene AMD‘s Favorite Effects

19. Perform at Half Resolution • HDAO at full resolution is expensive • Running at half resolution captures more occlusion – and is obviously much faster • Problem: Artifacts are introduced when combined with the full resolution scene AMD‘s Favorite Effects

20. Bilateral Dilate & Blur HDAO buffer doesn‘t match with scene A bilateral dilate & blur fixes the issue AMD‘s Favorite Effects

21. New Pipeline... ½ Res Still much faster than performing at full res! Horizontal Pass Vertical Pass Bilinear Upsample Intermediate UAV Dilated & Blurred AMD‘s Favorite Effects

22. Pixel Shader vs DirectCompute *Tested on a range of AMD and NVIDIA DX11 HW, DirectCompute is between ~2.53x to ~3.17x faster than the Pixel Shader AMD‘s Favorite Effects

23. Depth of Field • Many techniques exist to solve this problem • A common technique is to figure out how blurry a pixel should be • Often called the Cirle of Confusion (CoC) • A Gaussian blur weighted by CoC is a pretty efficient way to implement this effect AMD‘s Favorite Effects

24. The Pipeline... Vertical Pass Horizontal Pass Intermediate UAV CoC AMD‘s Favorite Effects

25. Shogun 2: DoF OFF AMD‘s Favorite Effects

26. Shogun 2: DoF ON AMD‘s Favorite Effects

27. Pixel Shader vs DirectCompute *Tested on a range of AMD and NVIDIA DX11 HW, DirectCompute is between ~1.48x to ~1.86x faster than the Pixel Shader AMD‘s Favorite Effects

28. Summary • DirectCompute greatly accelerates larger kernel diameter filters • Allows for filtering at full resolution • For access to source code: • HDAO11: jon.story@amd.com • DoF11: nicolas.thibieroz@amd.com AMD‘s Favorite Effects

29. Questions?takahiro.harada@amd.comholger.gruen@amd.comjon.story@amd.comPlease fill in the feedback forms! AMD‘s Favorite Effects