SlideShare una empresa de Scribd logo
1 de 59
Descargar para leer sin conexión
Tiago Sousa, R&D Principal Graphics Engineer
GRAPHICS GEMS
Advances in Real-Time Rendering course, Siggraph 2013 2
AGENDA
Anti-aliasing
 Practical Deferred MSAA
 Temporal Antialiasing: SMAA 1TX
Camera Post-Processing
 Depth of Field
 Motion Blur
Sharing results from ongoing research
 Results not used in a shipped game yet 
Advances in Real-Time Rendering course, Siggraph 2013 3
ANTIALIASINGDEFERRED MSAA REVIEW
The problem: Multiple passes + r/w from Multisampled RTs
 DX 10.1 introduced SV_SampleIndex / SV_Coverage system value semantics.
 Allows to solve via multipass for pixel/sample frequency passes [Thibieroz11]
SV_SampleIndex
 Forces pixel shader execution for each sub-sample and provides index of the sub-sample currently executed
 Index can be used to fetch sub-sample from a Multisampled RT. E.g. FooMS.Load( UnnormScreenCoord, nSampleIndex)
SV_Coverage
 Indicates to pixel shader which sub-samples covered during raster stage.
 Can modify also sub-sample coverage for custom coverage mask
DX 11.0 Compute Tiled based deferred shading/lighting MSAA is simpler
 Loop through MSAA tagged sub-samples
Advances in Real-Time Rendering course, Siggraph 2013 4
DEFERRED MSAAHEADS UP !
Simple theory, troublesome practice
 At least with complex deferred renderers
Non-MSAA friendly code accumulates fast.
 Breaks regularly, as new techniques added without MSAA consideration
 Even if still works.. Very often you’ll need to pinpoint and fix non-msaa friendly techniques, as these introduce visual
artifacts.
 E.g. white/dark outlines, or no AA at all
Do it upfront. Retrofitting a renderer to support Deferred MSAA is some work
 And it is very finiky
Advances in Real-Time Rendering course, Siggraph 2013 5
DEFERRED MSAACUSTOM RESOLVE & PER-SAMPLE MASK
Post G-Buffer, perform a custom msaa resolve
 Pre-resolves sample 0, for pixel frequency passes such as lighting/other MSAA dependent passes
 In same pass create sub-sample mask (compare samples similarity, mark if mismatching)
 Avoid default SV_COVERAGE, since it results in redundant processing on regions not requiring MSAA
SV_Coverage Custom Per-Sample Mask
Advances in Real-Time Rendering course, Siggraph 2013 6
DEFERRED MSAASTENCIL BATCHING [SOUSA13]
Batching per-sample stencil mask with regular stencil buffer usage
 Reserve 1 bit from stencil buffer
 Update with sub-sample mask
 Tag entire pixel-quad instead of just single pixel -> improves stencil culling efficiency
 Make usage of stencil read/write bitmask to avoid per-sample bit override
 StencilWriteMask = 0x7F
 Restore whenever a stencil clear occurs
Not possible due to extreme stencil usage?
 Use clip/discard
 Extra overhead also from additional texture read for per-sample mask
Advances in Real-Time Rendering course, Siggraph 2013 7
DEFERRED MSAAPIXEL AND SAMPLE FREQUENCY PASSES
Pixel Frequency Passes
 Set stencil read mask to reserved bits for per-pixel regions (~0x80)
 Bind pre-resolved (non-multisampled) targets SRVs
 Render pass as usual
Sample Frequency Passes
 Set stencil read mask to reserved bit for per-sample regions (0x80)
 Bind multisampled targets SRVs
 Index current sub-sample via SV_SAMPLEINDEX
 Render pass as usual
Advances in Real-Time Rendering course, Siggraph 2013 8
DEFERRED MSAAALPHA TEST SSAA
Alpha testing requires ad hoc solution
 Default SV_Coverage only applies to triangle edges
Create your own sub-sample coverage mask
 E.g. check if current sub-sample uses AT or not and set bit
static const float2 vMSAAOffsets[2] = {float2(0.25, 0.25),float2(-0.25,-0.25)};
const float2 vDDX = ddx(vTexCoord.xy);
const float2 vDDY = ddy(vTexCoord.xy);
[unroll] for(int s = 0; s < nSampleCount; ++s)
{
float2 vTexOffset = vMSAAOffsets[s].x * vDDX + (vMSAAOffsets[s].y * vDDY);
float fAlpha = tex2D(DiffuseSmp, vTexCoord + vTexOffset).w;
uCoverageMask |= ((fAlpha-fAlphaRef) >= 0)? (uint(0x1)<<i) : 0;
}
Alpha Test SSAA Disabled
Alpha Test SSAA Enabled
Advances in Real-Time Rendering course, Siggraph 2013 9
DEFERRED MSAAPERFORMANCE SHORTCUTS
Deferred cascades sun shadow maps
 Render shadows as usual at pixel frequency
 Bilateral upscale during deferred shading composite pass
Advances in Real-Time Rendering course, Siggraph 2013 10
DEFERRED MSAAPERFORMANCE SHORTCUTS (2)
Non-opaque techniques accessing depth (e.g. Soft-Particles)
 Recommendation to tackle via per-sample frequency is fairly slow on real world scenarios
 Using Max Depth works ok for most cases and N-times faster
Advances in Real-Time Rendering course, Siggraph 2013 11
MSAAPERFORMANCE SHORTCUTS (3)
Many games, also doing:
 Skipping Alpha Test Super Sampling
 Use alpha to coverage instead, or even no alpha test AA (let morphological AA tackle that)
 Render only opaque with MSAA
 Then render transparents withouth MSAA
 Assuming HDR rendering: note that tone mapping is implicitly done post-resolve resulting is loss of detail on high
contrast regions
Advances in Real-Time Rendering course, Siggraph 2013 12
DEFERRED MSAAMSAA FRIENDLINESS
Look out for these:
 No MSAA noticeably working, or noticeable bright/dark silhouettes.
Incorrect Incorrect
Advances in Real-Time Rendering course, Siggraph 2013 13
DEFERRED MSAAMSAA FRIENDLINESS
Look out for these:
 No MSAA noticeably working, or noticeable bright/dark silhouettes.
Fixed Fixed
Advances in Real-Time Rendering course, Siggraph 2013 14
DEFERRED MSAARECAP
Accessing and/or rendering to Multisampled RTs?
 Then you need to care about accessing and outputting correct sub-sample
In general always strive to minimize BW
 Avoid vanilla deferred lighting
 Prefer fully deferred, hybrids, or just skip deferred altogether.
 If deferred, prefer thin g-buffers
 Each additional target on g-buffer incurs in export rate overhead [Thibieroz11]
 NV/AMD (GCN): Export Cost = Cost(RT0)+Cost(RT1)..., AMD (older hw): Export Cost = (Num RTs) * (Slowest RT)
 Fat formats are half rate sampling cost for bilinear filtering modes on GCN [Thibieroz13]
 For lighting/some hdr post processes: 32 bit R11G11B10F fmt suffices for most cases
Advances in Real-Time Rendering course, Siggraph 2013 15
ANTIALIASING + 4K RESOLUTIONSWILL WE NEED MSAA AT ALL?
Likely can start getting creative here
Source 4kHighly Compressed 4k
Advances in Real-Time Rendering course, Siggraph 2013 16
ANTIALIASING THE QUEST FOR BETTER (AND FAST) AA
2011: the boom year of alternative AA modes (and naming combos)
 FXAA, MLAA, SMAA, SRAA, DEAA, GBAA, DLAA, ETC AA
 “Filtering Approaches for Real-Time Anti-Aliasing” [Jimenez et all 11]
Shading Anti-aliasing
 “Mip mapping normal maps” [Toksvig04]
 “Spectacular Specular: LEAN and CLEAN Specular Highlights” [Baker11]
 “Rock-Solid Shading: Image Stability withouth Sacrificing Detail” [Hill12]
Advances in Real-Time Rendering course, Siggraph 2013 17
TEMPORAL SSAASMAA 2TX/4X REVIEW [JIMENEZ11][SOUSA11]
Morphological AA + MSAA + Temporal SSAA combo
 Balanced cost/quality tradeoff, techniques complement each other.
 Temporal component uses 2 sub-pixel buffers.
 Each frame adds a sub-pixel jitter for 2x SSAA.
 Reproject previous frame and blend between current and
previous frames, via velocity length weighting.
 Preserves image sharpness + reasonable temporal stability
Naive Blending Reprojection |V| weighting
Advances in Real-Time Rendering course, Siggraph 2013 18
TEMPORAL AACOMMON ROBUSTNESS FLAWS
Relying on opaque geometry information
 Can’t handle signal (color) changes nor transparency.
 For correct result, all opaque geometry must output velocity
Pathological cases
 Alpha blended surfaces (e.g. particles), lighting/shadow/reflections/uv animation/etc
 Any scatter and alike post processes, before the AA resolve
Can result in distracting errors
 E.g. “ghosting” on transparency, lighting, shadows and such
 Silhouettes might appear, from scatter and alike post processes (e.g. bloom)
Multi-GPU
 Simplest solution: force resource sync
 NVIDIA exposes driver hint to force sync resource , via NVAPI. This is solution used by NVIDIAs TXAA
Note to hw vendors: would be great if all vendors exposed such (even better if Multi-GPU API functionality generalized)
Incorrect Blending
Advances in Real-Time Rendering course, Siggraph 2013 19
SMAA 1TXA MORE ROBUST TEMPORAL AA
Concept: Only track signal changes, don’t rely on geometry information
 For higher temporal stability: accumulate multiple frames in an accumulation buffer, alike TXAA [Lottes12]
 Re-project accumulation buffer
 Weighting: Map acc. buffer colors into the range of curr. frame neighborhood color extents [Malan2012]; different weight
for hi/low frequency regions (for sharpness preservation).
Current Frame (t0)
M
Accumulation Buffer (tN)
TL TR
BL BR
M
Advances in Real-Time Rendering course, Siggraph 2013 20
SMAA 1TXA MORE ROBUST TEMPORAL AA (2)
Concept: Only track signal changes, don’t rely on geometry information
 For higher temporal stability: accumulate multiple frames in an accumulation buffer, alike TXAA [Lottes12]
 Re-project accumulation buffer
 Weighting: Map acc. buffer colors into the range of curr. frame neighborhood color extents [Malan2012]; different weight
for hi/low frequency regions (for sharpness preservation).
Smaa 1txRegular TAA
Advances in Real-Time Rendering course, Siggraph 2013 21
SMAA 1TXA MORE ROBUST TEMPORAL AA (3)
Sample code
float3 cM = tex2D(tex0, tc.xy);
float3 cAcc = tex2D(tex0, reproj_tc.xy);
float3 cTL = tex2D(tex0, tc0.xy);
float3 cTR = tex2D(tex0, tc0.zw);
float3 cBL = tex2D(tex0, tc1.xy);
float3 cBR = tex2D(tex0, tc1.zw);
float3 cMax = max(cTL, max(cTR, max(cBL, cBR)));
float3 cMin = min(cTL, min(cTR, min(cBL, cBR)));
float3 wk = abs((cTL+cTR+cBL+cBR)*0.25-cM);
return lerp(cM, clamp(cAcc, cMin, cMax), saturate(rcp(lerp(kl, kh, wk)));
Advances in Real-Time Rendering course, Siggraph 2013 22
DEPTH OF FIELD
Advances in Real-Time Rendering course, Siggraph 2013 23
DEPTH OF FIELDPLAUSIBLE DOF: PARAMETERIZATION
Artist friendly parameters is one reason why games DOF tends to look wrong
 Typical controls such as “focus range” + “blur amount” and others have not much physical meaning
 CoC depends mainly on f-stops, focal length and focal distance. These last 2 directly affect FOV.
 If you want more Bokeh, you need to max your focal length + widen aperture. This means also getting closer or further
from subject for proper framing.
Not the typical way a game artist/programmer thinks about DOF.
Wider fov, less bokeh Shallow fov, more bokeh
Advances in Real-Time Rendering course, Siggraph 2013 24
50 mm
DEPTH OF FIELDFOCAL LENGTH
200 mm
Advances in Real-Time Rendering course, Siggraph 2013 25
DEPTH OF FIELDF-STOPS
2 f-stops 22 f-stops8 f-stops
Advances in Real-Time Rendering course, Siggraph 2013 26
DEPTH OF FIELDF-STOPS (2)
2 f-stops 2.8 f-stops 4 f-stops 5.6 f-stops
Advances in Real-Time Rendering course, Siggraph 2013 27
DEPTH OF FIELDFOCAL DISTANCE
0.5 m
Advances in Real-Time Rendering course, Siggraph 2013 28
DEPTH OF FIELDFOCAL DISTANCE
0.75 m
Advances in Real-Time Rendering course, Siggraph 2013 29
DEPTH OF FIELDFOCAL DISTANCE
1.0 m
Advances in Real-Time Rendering course, Siggraph 2013 30
DEPTH OF FIELDPLAUSIBLE DOF: BOKEH
Out of focus region is commonly referred in photography as “Bokeh” (Japanese word for blur)
Bokeh shape has direct relation to camera aperture size (aka f-stops) and diaphragm blades count
 Bigger aperture = more “circular” bokeh, smaller aperture = more polygonal bokeh
Polygonal bokeh look depends on diaphragm blades count
Blades count varies on lens characteristics
 Bigger aperture = more light enters, smaller aperture = less light
On night shots, you might notice often more circular bokeh and more motion blur
Bokeh kernel is flat
 Almost same amount of light enters camera iris from all directions
Edges might be in shadow, this is commonly known as Vignetting
Poor lenses manufacturing may introduce a vast array of optical aberrations [Wiki01]
 This is main reason why gaussian blur, diffusion dof, and derivative techniques look wrong/visually unpleasant
Advances in Real-Time Rendering course, Siggraph 2013 31
DEPTH OF FIELDSTATE OF THE ART OVERVIEW
Scatter based techniques [Cyril05][Sawada07][3DMark11][Mittring11][Sousa11]
 Render 1 quad or tri per-pixel, scale based on CoC
Simple implementation and nice results. Downside: performance, particularly on shallow DOF
Variable/inconsistent fillrate hit, depending on near/far layers resolution and aperture size might reach >5 ms
Quad generation phase has fixed cost attached.
Advances in Real-Time Rendering course, Siggraph 2013 32
DEPTH OF FIELDSTATE OF THE ART OVERVIEW (2)
Gather based: separable (inflexible kernel) vs. kernel flexibility
[Macintosh12]
[Gotanda09]
[White11] [Andreev 12]
[Kawase09]
Advances in Real-Time Rendering course, Siggraph 2013 33
DEPTH OF FIELDA PLAUSIBLE AND EFFICIENT DOF RECONSTRUCTION FILTER
Separable flexible filter: low bandwidth requirement + different bokeh shape possible
 1st pass N^2 taps (e.g: 7x7).
 2nd pass N^2 taps (e.g: 3x3) for flood filling shape
 R11G11B10F: downscaled HDR scene; R8G8: CoC
 Done at half resolution
 Far/Near fields processed in same pass
 Limit offset range to minimize undersampling
 Higher specs hw can have higher tap count
Diaphragm and optical aberrations sim
Physically based CoC
Advances in Real-Time Rendering course, Siggraph 2013 34
DEPTH OF FIELDLENS REVIEW
Pinhole “Lens”
 A camera withouth lens
 Light has to pass through single small aperture before hitting image plane
 Tipical realtime rendering
Thin lens
 Camera lenses have finite dimension
 Light refracts through lens until hitting image plane.
 F = Focal lenght
 P = Plane in focus
 I = Image distance
Circle of
Confusion
Image Plane
P I
F
Advances in Real-Time Rendering course, Siggraph 2013 35
DEPTH OF FIELDLENS REVIEW (2)
The thin lens equation gives relation between:
 F = Focal length (where light starts getting in focus)
 P = Plane in focus (camera focal distance)
 I = Image distance (where image is projected in focus)
Circle of Confusion [Potmesil81]
 f = f-stops (aka as the f-number or focal ratio)
 D = Object distance
 A = Aperture diameter
Simplifies to:
 Note: f and F are known variables from camera setup
 Folds down into a single mad in shader
Camera FOV:
 Typical film formats (or sensor), 35mm/70mm
 Can alternatively derive focal length from FOV
Advances in Real-Time Rendering course, Siggraph 2013 36
DEPTH OF FIELDSAMPLING
Concentric Mapping [Shirley97] used for uniform sample distribution
 Maps unit square to unit circle
 Square mapped to (a,b) [-1,1]2 and divided into 4 regions by lines a=b, a=-b
 First region given by:
Diaphragm simulation by morphing samples to n-gons
 Via a modified equation for the regular polygon.
Advances in Real-Time Rendering course, Siggraph 2013 37
DEPTH OF FIELDSAMPLING: 2ND ITERATION
To floodfill final shape, composite via boolean union, similarly to [McIntosh12]
+
U
+...=
U...=
Advances in Real-Time Rendering course, Siggraph 2013 38
DEPTH OF FIELDSEPARABLE FILTER PASSES
1st iteration: 49 taps (0.426ms) 2nd iteration: 9 taps (0.094 ms; 441 taps accumulated)
Advances in Real-Time Rendering course, Siggraph 2013 39
DEPTH OF FIELDREFERENCE VS SEPARABLE FILTER
144 taps (1.31ms) 58 taps (0.52ms)
Advances in Real-Time Rendering course, Siggraph 2013 40
DEPTH OF FIELDDIAPHRAGM SIMULATION IN ACTION
4f-stops2f-stops
Advances in Real-Time Rendering course, Siggraph 2013 41
Tile Min/Max CoC
 Downscale CoC target k times (k = tile count)
 Take min fragment for far field, max fragment for near field
 R8G8 storage
Used to process near/far fields in same pass
 Dynamic branching using Tile Min/Max CoC for both fields
 Balances cost between far/near
 Also used for scatter as gather approximation for near field
Can fold cost with other post-processes
 Initial downscale cost folded with HDR scene downscale for bloom,
also pack near/far fields HDR input into R11G11B10F - all in 1 pass
DEPTH OF FIELDTILE MIN/MAX COC
Tile min-CoC (far field)
Tile max-CoC (near field)
Advances in Real-Time Rendering course, Siggraph 2013 42
DEPTH OF FIELDFAR + NEAR FIELD PROCESSING
Both fields use half resolution input
 Careful: downscale is source of error due to bilinear filtering
 Use custom bilinear (bilateral) filter for downscaling
Far Field
 Scale kernel size and weight samples with far CoC [Scheumerman05]
 Pre-multiply layer with far CoC [Gotanda09]
Prevents bleeding artifacts from bilinear/separable filter
No weighting CoC weighting
CoC weighting +
CoC pre-multiply
Advances in Real-Time Rendering course, Siggraph 2013 43
DEPTH OF FIELDFAR + NEAR FIELD PROCESSING
Both fields use half resolution input
 Careful: downscale is source of error due to bilinear filtering
 Use custom bilinear (bilateral) filter for downscaling
Far Field
 Scale kernel size and weight samples with far CoC [Scheumerman05]
 Pre-multiply layer with far CoC [Gotanda09]
Prevents bleeding artifacts from bilinear/separable filter
Near Field
 Scatter as gather aproximation
 Scale kernel size + weight fragments with Tile Max CoC against
near CoC
 Pre-multiply with near CoC
Only want to blur near field fragments (cheap partial occlusion approximation) Far Field Near Field
Advances in Real-Time Rendering course, Siggraph 2013 44
DEPTH OF FIELDFINAL COMPOSITE
Far field: upscale via bilateral filter
 Take 4 taps from half res CoC, compare against full res CoC
 Weighted using bicubic filtering for quality [Sigg05]
 Far field CoC used for blending
Near field: upscale carelessly
 Half resolution near field CoC used for blending
 Can bleed as much as possible
 Also using bicubic filtering
Carefull with blending
 Linear blending doesn’t look good (signal frequency soup)
Can be seen in many games, including all Crysis series (puts hat of shame)
 Simple to address: use non-linear blend factor instead.
Linear blend
Non-linear blend (better)
Advances in Real-Time Rendering course, Siggraph 2013 45
MOTION BLUR
Advances in Real-Time Rendering course, Siggraph 2013 46
MOTION BLURSHUTTER SPEED AND F-STOPS REVIEW
Amount of motion blur is relative to camera shutter speed and f-stops usage
 The longer the exposure (slower shutter), the more light received (and the bigger amount of motion blur), and vice-versa
 The lower f-stops the faster the exposure can be (and have less motion blur), and vice versa
2 f-stops, shutter 1/20 sec 4 f-stops, shutter 1/6 sec
Advances in Real-Time Rendering course, Siggraph 2013 47
MOTION BLURSTATE OF THE ART OVERVIEW
Scatter via geometry expansion [Green03][Sawada07]
 Require additional geometry pass + gs shader usage *
Advances in Real-Time Rendering course, Siggraph 2013 48
MOTION BLURSTATE OF THE ART OVERVIEW (2)
Scatter as gather [Sousa08][Gotanda09][Sousa11][Maguire12]
 E.g.velocity dilation, velocity blur, tile max velocity; single vs. multiple pass composite; depth/v/obj ID masking; single pass DOF+MB
Advances in Real-Time Rendering course, Siggraph 2013 49
MOTION BLURRECONSTRUCTION FILTER FOR PLAUSIBLE MB [MCGUIRE12]
Tile Max Velocity and Tile Neighbor Max Velocity
 Downscale Velocity buffer by k times (k is tile count)
 Take max length velocity at each step
Motion Blur Pass
 Tile Neighbor Max for early out
 Tile Max Velocity as center velocity tap
 At each iteration step weight against full
resolution ||V|| and Depth
Advances in Real-Time Rendering course, Siggraph 2013 50
MOTION BLURAN IMPROVED RECONSTRUCTION FILTER
Performant Quality
 Simplify and vectorize inner loop weight computation (ends up in couple mads)
 Fat buffers sampling are half rate on GCN hw with bilinear (point filtering is fullrate, but doesn’t look good due to aliasing)
 Inputs: R11G11B10F for scene , bake ||V|| and 8 bit depth into a R8G8 target
 Make it separable, 2 passes [Sousa08]
1 iteration: 6 taps (0.236 ms) 2nd iter: 6 taps (+0.236 ms; 36 taps acc.)
Advances in Real-Time Rendering course, Siggraph 2013 51
MOTION BLURAN IMPROVED RECONSTRUCTION FILTER (2)
Inner loop sample
const float2 tc = min_tc + blur_step * s;
const float lensq_xy = abs(min_len_xy + len_xy_step * s);
const float2 vy = tex2Dlod(tex1, float4(tc.xy, 0, 0)); // x =||v||, y=depth
const float2 cmp_z = DepthCmp(float2(vx.y, vy.y), float2(vy.y, vx.y), 1);
const float4 cmp_v = VelCmp(lensq_xy, float2(vy.x, lensq_vx));
const float w = (dot(cmp_z.xy, cmp_v.xy) + (cmp_v.z * cmp_v.w) * 2);
acc.rgb += tex2Dlod(tex0, float4(tc.xy, 0, 0)) * w;
wacc += w;
float2 DepthCmp(float2 z0, float2 z1, float2 fSoftZ) {
return saturate( (1.0f + z0* fSoftZ) - z1* fSoftZ );
}
float4 VelCmp(float lensq_xy, float2 vxy) {
return saturate((1.0f - lensq_xy.xxxx *rcp(vxy.xyxy)) + float4(0.0f, 0.0f, 0.95f, 0.95f));
}
Advances in Real-Time Rendering course, Siggraph 2013 52
Output object velocity in G-Buffer (only when required)
 Avoids separate geometry passes.
 Rigid geometry: object distance < distance threshold
 Deformable geometry: if amount of movement > movement threshold
 Moving geometry rendered last
 R8G8 fmt
Composite with camera velocity
 Velocity encoded in gamma 2.0 space
 Precision still insufficient, but not much noticeable in practice
MOTION BLURAN IMPROVED RECONSTRUCTION FILTER (3)
Encode
Decode
Object velocity
Composite with camera velocity
Advances in Real-Time Rendering course, Siggraph 2013 53
MOTION BLURMB OR DOF FIRST?
In real world MB/DOF occur simultaneously
 A dream implementation: big N^2 kernel + batched DOF/MB
 Or sprite based with MB quad stretching
 Full resolution! 1 Billion taps! FP16! Multiple layers! 
But... performance still matters (consoles):
 DOF before MB introduces less error when MB happening on focus
This is due MB is a scatter as gather op relying on geometry data.
Any other similar op after will introduce error. And vice-versa.
Error from MB after DOF is less noticeable.
 Order swap makes DOF harder to fold with other posts
Additional overhead
MB before DOF MB after DOF
Advances in Real-Time Rendering course, Siggraph 2013 54
FINAL REMARKS
Practical MSAA details
 Do’s and Dont’s
SMAA 1TX: A More Robust Temporal AA
 For just 4 extra texture ops and couple alu
A Plausible and Performant DOF Reconstruction Filter
 Separable flexible filter, any bokeh kernel shape doable
 1st pass: 0.426ms, 2nd pass: 0.094ms. Sum: 0.52ms for reconstruction filter *
An Improved Reconstruction Filter for Plausible Motion Blur
 Separable, 1st pass: 0.236 ms, 2nd pass: 0.236ms. Sum: 0.472ms for reconstruction filter *
* 1080p + AMD 7970
Advances in Real-Time Rendering course, Siggraph 2013 55
SPECIAL THANKS
Natalya Tartachuck
Michael Kopietz, Nicolas Schulz, Christopher Evans, Carsten Wenzel, Christopher Raine,
Nick Kasyan, Magnus Larbrant, Pierre Donzallaz
Advances in Real-Time Rendering course, Siggraph 2013 56
WE ARE HIRING !
Advances in Real-Time Rendering course, Siggraph 2013 57
QUESTIONS ?
Tiago@Crytek.Com / Twitter: Crytek_Tiago
Advances in Real-Time Rendering course, Siggraph 2013 58
REFERENCES
Potmesil M., Chakravarty I. “Synthetic Image Generation with a Lens and Aperture Camera Model”, 1981
Shirley P., Chiu K., “A Low Distortion Map Between Disk and Square”, 1997
Green S. “Stupid OpenGL Tricks”, 2003
Toksvig M., “Mip Mapping Normal Maps”, 2004
Scheuermann T., Tatarchuk N. “Improved Depth of Field Rendering”, Shader X3, 2005
Sigg C., Hadwiger M., “Fast Third-Order Texture Filtgering”, 2005
Cyril, P et al. “Photographic Depth of Field Rendering”, 2005
Gritz L., Eon E. “The Importance of Being Linear”, 2007
Sawada Y., Talk at Game Developers Conference, http://www.beyond3d.com/content/news/499 , 2007
Sousa T. “Crysis Next Gen Effects”, 2008
Gotanda Y. ”Star Ocean 4: Flexible Shader Management and Post Processing”, 2009
Yang L. et al, “Amortized SuperSampling”, 2009
Kawase M., “Anti-Downsized Buffer Artifacts”, 2011
Sousa T., “CryENGINE 3 Rendering Techniques”, 2011
Binks D., “Dynamic Resolution Rendering”, 2011
Sousa T., Schulz N. , Kasyan N., “Secrets of CryENGINE 3 Graphics Technology”, 2011
Sousa T., “Anti-Aliasing Methods in CryENGINE 3”, 2011
Jimenez J. et.al, “Filtering Approaches for Real-Time Anti-Aliasing”, 2011
Thibieroz N., “Deferred Shading Optimizations”, 2011
Advances in Real-Time Rendering course, Siggraph 2013 59
REFERENCES
Cloward B., Otstott A., “Cinematic Character Lighting in Star Wars: The Old Republic”, 2011
Mittring M., Dudash B., “The Technology Behind the DirectX11 Unreal Engine “Samaritan” Demo“, 2011
Futuremark, 3DMark11 “White Paper”, 2011
Baker D., “Spectacular Specular: LEAN and CLEAN Specular Highlights”, 2011
White J., Brisebois C., “More Performance! Five Rendering Ideas from Batlefield 3 and Need for Speed: The Run”, 2011
Lottes T., “Nvidia TXAA”, http://www.nvidia.in/object/txaa-anti-aliasing-technology-in.html#gameContent=1 , 2012
McGuire M. et al, “A Reconstruction Filter for Plausible Motion Blur”, 2012
McIntosh L. Et al, “Efficiently Simulating Bokeh”, 2012
Malan H., “Real-Time Global Illumination and Reflections in Dust 514”, 2012
Hill S., Baker D., “Rock-Solid Shading: Image Stability withouth Sacrificing Detail”, 2012
Sousa T., Wenzel C., Raine C., “Rendering Technologies of Crysis 3”, 2013
Thibieroz N., Gruen H., “DirectX11 Performance Reloaded”, 2013
Andreev D., “Rendering Tricks in Dead Space 3”, 2013
http://en.wikipedia.org/wiki/Lens_%28optics%29
http://en.wikipedia.org/wiki/Pinhole_camera
http://en.wikipedia.org/wiki/F-number
http://en.wikipedia.org/wiki/Angle_of_view

Más contenido relacionado

La actualidad más candente

Five Rendering Ideas from Battlefield 3 & Need For Speed: The Run
Five Rendering Ideas from Battlefield 3 & Need For Speed: The RunFive Rendering Ideas from Battlefield 3 & Need For Speed: The Run
Five Rendering Ideas from Battlefield 3 & Need For Speed: The RunElectronic Arts / DICE
 
Hable John Uncharted2 Hdr Lighting
Hable John Uncharted2 Hdr LightingHable John Uncharted2 Hdr Lighting
Hable John Uncharted2 Hdr Lightingozlael ozlael
 
Rendering Technologies from Crysis 3 (GDC 2013)
Rendering Technologies from Crysis 3 (GDC 2013)Rendering Technologies from Crysis 3 (GDC 2013)
Rendering Technologies from Crysis 3 (GDC 2013)Tiago Sousa
 
Star Ocean 4 - Flexible Shader Managment and Post-processing
Star Ocean 4 - Flexible Shader Managment and Post-processingStar Ocean 4 - Flexible Shader Managment and Post-processing
Star Ocean 4 - Flexible Shader Managment and Post-processingumsl snfrzb
 
Moving Frostbite to Physically Based Rendering
Moving Frostbite to Physically Based RenderingMoving Frostbite to Physically Based Rendering
Moving Frostbite to Physically Based RenderingElectronic Arts / DICE
 
Bindless Deferred Decals in The Surge 2
Bindless Deferred Decals in The Surge 2Bindless Deferred Decals in The Surge 2
Bindless Deferred Decals in The Surge 2Philip Hammer
 
Physically Based and Unified Volumetric Rendering in Frostbite
Physically Based and Unified Volumetric Rendering in FrostbitePhysically Based and Unified Volumetric Rendering in Frostbite
Physically Based and Unified Volumetric Rendering in FrostbiteElectronic Arts / DICE
 
Dissecting the Rendering of The Surge
Dissecting the Rendering of The SurgeDissecting the Rendering of The Surge
Dissecting the Rendering of The SurgePhilip Hammer
 
Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...
Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...
Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...Johan Andersson
 
Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14
Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14
Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14AMD Developer Central
 
CryENGINE 3 Rendering Techniques
CryENGINE 3 Rendering TechniquesCryENGINE 3 Rendering Techniques
CryENGINE 3 Rendering TechniquesTiago Sousa
 
Rendering AAA-Quality Characters of Project A1
Rendering AAA-Quality Characters of Project A1Rendering AAA-Quality Characters of Project A1
Rendering AAA-Quality Characters of Project A1Ki Hyunwoo
 
Past, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in GamesPast, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in GamesColin Barré-Brisebois
 
Siggraph2016 - The Devil is in the Details: idTech 666
Siggraph2016 - The Devil is in the Details: idTech 666Siggraph2016 - The Devil is in the Details: idTech 666
Siggraph2016 - The Devil is in the Details: idTech 666Tiago Sousa
 
Parallel Futures of a Game Engine (v2.0)
Parallel Futures of a Game Engine (v2.0)Parallel Futures of a Game Engine (v2.0)
Parallel Futures of a Game Engine (v2.0)Johan Andersson
 
4K Checkerboard in Battlefield 1 and Mass Effect Andromeda
4K Checkerboard in Battlefield 1 and Mass Effect Andromeda4K Checkerboard in Battlefield 1 and Mass Effect Andromeda
4K Checkerboard in Battlefield 1 and Mass Effect AndromedaElectronic Arts / DICE
 
An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...
An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...
An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...Codemotion
 

La actualidad más candente (20)

Five Rendering Ideas from Battlefield 3 & Need For Speed: The Run
Five Rendering Ideas from Battlefield 3 & Need For Speed: The RunFive Rendering Ideas from Battlefield 3 & Need For Speed: The Run
Five Rendering Ideas from Battlefield 3 & Need For Speed: The Run
 
Hable John Uncharted2 Hdr Lighting
Hable John Uncharted2 Hdr LightingHable John Uncharted2 Hdr Lighting
Hable John Uncharted2 Hdr Lighting
 
Rendering Technologies from Crysis 3 (GDC 2013)
Rendering Technologies from Crysis 3 (GDC 2013)Rendering Technologies from Crysis 3 (GDC 2013)
Rendering Technologies from Crysis 3 (GDC 2013)
 
Star Ocean 4 - Flexible Shader Managment and Post-processing
Star Ocean 4 - Flexible Shader Managment and Post-processingStar Ocean 4 - Flexible Shader Managment and Post-processing
Star Ocean 4 - Flexible Shader Managment and Post-processing
 
Stochastic Screen-Space Reflections
Stochastic Screen-Space ReflectionsStochastic Screen-Space Reflections
Stochastic Screen-Space Reflections
 
Moving Frostbite to Physically Based Rendering
Moving Frostbite to Physically Based RenderingMoving Frostbite to Physically Based Rendering
Moving Frostbite to Physically Based Rendering
 
Bindless Deferred Decals in The Surge 2
Bindless Deferred Decals in The Surge 2Bindless Deferred Decals in The Surge 2
Bindless Deferred Decals in The Surge 2
 
Light prepass
Light prepassLight prepass
Light prepass
 
Physically Based and Unified Volumetric Rendering in Frostbite
Physically Based and Unified Volumetric Rendering in FrostbitePhysically Based and Unified Volumetric Rendering in Frostbite
Physically Based and Unified Volumetric Rendering in Frostbite
 
Dissecting the Rendering of The Surge
Dissecting the Rendering of The SurgeDissecting the Rendering of The Surge
Dissecting the Rendering of The Surge
 
Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...
Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...
Frostbite Rendering Architecture and Real-time Procedural Shading & Texturing...
 
Lighting the City of Glass
Lighting the City of GlassLighting the City of Glass
Lighting the City of Glass
 
Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14
Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14
Vertex Shader Tricks by Bill Bilodeau - AMD at GDC14
 
CryENGINE 3 Rendering Techniques
CryENGINE 3 Rendering TechniquesCryENGINE 3 Rendering Techniques
CryENGINE 3 Rendering Techniques
 
Rendering AAA-Quality Characters of Project A1
Rendering AAA-Quality Characters of Project A1Rendering AAA-Quality Characters of Project A1
Rendering AAA-Quality Characters of Project A1
 
Past, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in GamesPast, Present and Future Challenges of Global Illumination in Games
Past, Present and Future Challenges of Global Illumination in Games
 
Siggraph2016 - The Devil is in the Details: idTech 666
Siggraph2016 - The Devil is in the Details: idTech 666Siggraph2016 - The Devil is in the Details: idTech 666
Siggraph2016 - The Devil is in the Details: idTech 666
 
Parallel Futures of a Game Engine (v2.0)
Parallel Futures of a Game Engine (v2.0)Parallel Futures of a Game Engine (v2.0)
Parallel Futures of a Game Engine (v2.0)
 
4K Checkerboard in Battlefield 1 and Mass Effect Andromeda
4K Checkerboard in Battlefield 1 and Mass Effect Andromeda4K Checkerboard in Battlefield 1 and Mass Effect Andromeda
4K Checkerboard in Battlefield 1 and Mass Effect Andromeda
 
An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...
An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...
An introduction to Realistic Ocean Rendering through FFT - Fabio Suriano - Co...
 

Similar a Graphics Gems from CryENGINE 3 (Siggraph 2013)

LIAO TSEN YUNG Cover Letter
LIAO TSEN YUNG Cover LetterLIAO TSEN YUNG Cover Letter
LIAO TSEN YUNG Cover LetterTsen Yung Liao
 
D3 D10 Unleashed New Features And Effects
D3 D10 Unleashed   New Features And EffectsD3 D10 Unleashed   New Features And Effects
D3 D10 Unleashed New Features And EffectsThomas Goddard
 
The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...
The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...
The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...Johan Andersson
 
Anti-Aliasing Methods in CryENGINE 3
Anti-Aliasing Methods in CryENGINE 3Anti-Aliasing Methods in CryENGINE 3
Anti-Aliasing Methods in CryENGINE 3Tiago Sousa
 
FlameWorks GTC 2014
FlameWorks GTC 2014FlameWorks GTC 2014
FlameWorks GTC 2014Simon Green
 
Crysis 2-key-rendering-features
Crysis 2-key-rendering-featuresCrysis 2-key-rendering-features
Crysis 2-key-rendering-featuresRaimundo Renato
 
High Performance Pedestrian Detection On TEGRA X1
High Performance Pedestrian Detection On TEGRA X1High Performance Pedestrian Detection On TEGRA X1
High Performance Pedestrian Detection On TEGRA X1NVIDIA
 
NVIDIA Graphics, Cg, and Transparency
NVIDIA Graphics, Cg, and TransparencyNVIDIA Graphics, Cg, and Transparency
NVIDIA Graphics, Cg, and TransparencyMark Kilgard
 
INNOVA - SPIE Remote Sensing 2019
INNOVA - SPIE Remote Sensing 2019 INNOVA - SPIE Remote Sensing 2019
INNOVA - SPIE Remote Sensing 2019 Andrea Di Pasquale
 
Mapping Water features from SAR Imagery
Mapping Water features from SAR ImageryMapping Water features from SAR Imagery
Mapping Water features from SAR Imagerybramantiyo marjuki
 
Advanced Game Development with the Mobile 3D Graphics API
Advanced Game Development with the Mobile 3D Graphics APIAdvanced Game Development with the Mobile 3D Graphics API
Advanced Game Development with the Mobile 3D Graphics APITomi Aarnio
 
Look Ma, No Jutter! Optimizing Performance Across Oculus Mobile
Look Ma, No Jutter! Optimizing Performance Across Oculus MobileLook Ma, No Jutter! Optimizing Performance Across Oculus Mobile
Look Ma, No Jutter! Optimizing Performance Across Oculus MobileUnity Technologies
 
Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012
Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012
Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012Joshua Mora
 
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time Raytracing
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time RaytracingSIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time Raytracing
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time RaytracingElectronic Arts / DICE
 
Masked Software Occlusion Culling
Masked Software Occlusion CullingMasked Software Occlusion Culling
Masked Software Occlusion CullingIntel® Software
 
PG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard Hoffnung
PG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard HoffnungPG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard Hoffnung
PG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard HoffnungAMD Developer Central
 
Borderless Per Face Texture Mapping
Borderless Per Face Texture MappingBorderless Per Face Texture Mapping
Borderless Per Face Texture Mappingbasisspace
 

Similar a Graphics Gems from CryENGINE 3 (Siggraph 2013) (20)

LIAO TSEN YUNG Cover Letter
LIAO TSEN YUNG Cover LetterLIAO TSEN YUNG Cover Letter
LIAO TSEN YUNG Cover Letter
 
Bending the Graphics Pipeline
Bending the Graphics PipelineBending the Graphics Pipeline
Bending the Graphics Pipeline
 
D3 D10 Unleashed New Features And Effects
D3 D10 Unleashed   New Features And EffectsD3 D10 Unleashed   New Features And Effects
D3 D10 Unleashed New Features And Effects
 
The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...
The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...
The Intersection of Game Engines & GPUs: Current & Future (Graphics Hardware ...
 
Anti-Aliasing Methods in CryENGINE 3
Anti-Aliasing Methods in CryENGINE 3Anti-Aliasing Methods in CryENGINE 3
Anti-Aliasing Methods in CryENGINE 3
 
FlameWorks GTC 2014
FlameWorks GTC 2014FlameWorks GTC 2014
FlameWorks GTC 2014
 
Crysis 2-key-rendering-features
Crysis 2-key-rendering-featuresCrysis 2-key-rendering-features
Crysis 2-key-rendering-features
 
High Performance Pedestrian Detection On TEGRA X1
High Performance Pedestrian Detection On TEGRA X1High Performance Pedestrian Detection On TEGRA X1
High Performance Pedestrian Detection On TEGRA X1
 
NVIDIA Graphics, Cg, and Transparency
NVIDIA Graphics, Cg, and TransparencyNVIDIA Graphics, Cg, and Transparency
NVIDIA Graphics, Cg, and Transparency
 
INNOVA - SPIE Remote Sensing 2019
INNOVA - SPIE Remote Sensing 2019 INNOVA - SPIE Remote Sensing 2019
INNOVA - SPIE Remote Sensing 2019
 
SPIE Remote Sensing 2019
SPIE Remote Sensing 2019SPIE Remote Sensing 2019
SPIE Remote Sensing 2019
 
Mapping Water features from SAR Imagery
Mapping Water features from SAR ImageryMapping Water features from SAR Imagery
Mapping Water features from SAR Imagery
 
Advanced Game Development with the Mobile 3D Graphics API
Advanced Game Development with the Mobile 3D Graphics APIAdvanced Game Development with the Mobile 3D Graphics API
Advanced Game Development with the Mobile 3D Graphics API
 
Look Ma, No Jutter! Optimizing Performance Across Oculus Mobile
Look Ma, No Jutter! Optimizing Performance Across Oculus MobileLook Ma, No Jutter! Optimizing Performance Across Oculus Mobile
Look Ma, No Jutter! Optimizing Performance Across Oculus Mobile
 
Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012
Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012
Do Theoretical Flo Ps Matter For Real Application’S Performance Kaust 2012
 
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time Raytracing
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time RaytracingSIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time Raytracing
SIGGRAPH 2018 - Full Rays Ahead! From Raster to Real-Time Raytracing
 
Masked Software Occlusion Culling
Masked Software Occlusion CullingMasked Software Occlusion Culling
Masked Software Occlusion Culling
 
Reyes
ReyesReyes
Reyes
 
PG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard Hoffnung
PG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard HoffnungPG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard Hoffnung
PG-4037, Fast modal analysis with NX Nastran and GPUs, by Leonard Hoffnung
 
Borderless Per Face Texture Mapping
Borderless Per Face Texture MappingBorderless Per Face Texture Mapping
Borderless Per Face Texture Mapping
 

Último

The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024Rafal Los
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking MenDelhi Call girls
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slidespraypatel2
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfEnterprise Knowledge
 
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...gurkirankumar98700
 
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | DelhiFULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhisoniya singh
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Allon Mureinik
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersThousandEyes
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptxHampshireHUG
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure servicePooja Nehwal
 
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...HostedbyConfluent
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking MenDelhi Call girls
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreternaman860154
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Igalia
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024Results
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonetsnaman860154
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsMaria Levchenko
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking MenDelhi Call girls
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdfhans926745
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesSinan KOZAK
 

Último (20)

The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024The 7 Things I Know About Cyber Security After 25 Years | April 2024
The 7 Things I Know About Cyber Security After 25 Years | April 2024
 
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
08448380779 Call Girls In Diplomatic Enclave Women Seeking Men
 
Slack Application Development 101 Slides
Slack Application Development 101 SlidesSlack Application Development 101 Slides
Slack Application Development 101 Slides
 
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdfThe Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
The Role of Taxonomy and Ontology in Semantic Layers - Heather Hedden.pdf
 
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
Kalyanpur ) Call Girls in Lucknow Finest Escorts Service 🍸 8923113531 🎰 Avail...
 
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | DelhiFULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
FULL ENJOY 🔝 8264348440 🔝 Call Girls in Diplomatic Enclave | Delhi
 
Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)Injustice - Developers Among Us (SciFiDevCon 2024)
Injustice - Developers Among Us (SciFiDevCon 2024)
 
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for PartnersEnhancing Worker Digital Experience: A Hands-on Workshop for Partners
Enhancing Worker Digital Experience: A Hands-on Workshop for Partners
 
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
04-2024-HHUG-Sales-and-Marketing-Alignment.pptx
 
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure serviceWhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
WhatsApp 9892124323 ✓Call Girls In Kalyan ( Mumbai ) secure service
 
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
Transforming Data Streams with Kafka Connect: An Introduction to Single Messa...
 
08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men08448380779 Call Girls In Civil Lines Women Seeking Men
08448380779 Call Girls In Civil Lines Women Seeking Men
 
Presentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreterPresentation on how to chat with PDF using ChatGPT code interpreter
Presentation on how to chat with PDF using ChatGPT code interpreter
 
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
Raspberry Pi 5: Challenges and Solutions in Bringing up an OpenGL/Vulkan Driv...
 
A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024A Call to Action for Generative AI in 2024
A Call to Action for Generative AI in 2024
 
How to convert PDF to text with Nanonets
How to convert PDF to text with NanonetsHow to convert PDF to text with Nanonets
How to convert PDF to text with Nanonets
 
Handwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed textsHandwritten Text Recognition for manuscripts and early printed texts
Handwritten Text Recognition for manuscripts and early printed texts
 
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men08448380779 Call Girls In Greater Kailash - I Women Seeking Men
08448380779 Call Girls In Greater Kailash - I Women Seeking Men
 
[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf[2024]Digital Global Overview Report 2024 Meltwater.pdf
[2024]Digital Global Overview Report 2024 Meltwater.pdf
 
Unblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen FramesUnblocking The Main Thread Solving ANRs and Frozen Frames
Unblocking The Main Thread Solving ANRs and Frozen Frames
 

Graphics Gems from CryENGINE 3 (Siggraph 2013)

  • 1. Tiago Sousa, R&D Principal Graphics Engineer GRAPHICS GEMS
  • 2. Advances in Real-Time Rendering course, Siggraph 2013 2 AGENDA Anti-aliasing  Practical Deferred MSAA  Temporal Antialiasing: SMAA 1TX Camera Post-Processing  Depth of Field  Motion Blur Sharing results from ongoing research  Results not used in a shipped game yet 
  • 3. Advances in Real-Time Rendering course, Siggraph 2013 3 ANTIALIASINGDEFERRED MSAA REVIEW The problem: Multiple passes + r/w from Multisampled RTs  DX 10.1 introduced SV_SampleIndex / SV_Coverage system value semantics.  Allows to solve via multipass for pixel/sample frequency passes [Thibieroz11] SV_SampleIndex  Forces pixel shader execution for each sub-sample and provides index of the sub-sample currently executed  Index can be used to fetch sub-sample from a Multisampled RT. E.g. FooMS.Load( UnnormScreenCoord, nSampleIndex) SV_Coverage  Indicates to pixel shader which sub-samples covered during raster stage.  Can modify also sub-sample coverage for custom coverage mask DX 11.0 Compute Tiled based deferred shading/lighting MSAA is simpler  Loop through MSAA tagged sub-samples
  • 4. Advances in Real-Time Rendering course, Siggraph 2013 4 DEFERRED MSAAHEADS UP ! Simple theory, troublesome practice  At least with complex deferred renderers Non-MSAA friendly code accumulates fast.  Breaks regularly, as new techniques added without MSAA consideration  Even if still works.. Very often you’ll need to pinpoint and fix non-msaa friendly techniques, as these introduce visual artifacts.  E.g. white/dark outlines, or no AA at all Do it upfront. Retrofitting a renderer to support Deferred MSAA is some work  And it is very finiky
  • 5. Advances in Real-Time Rendering course, Siggraph 2013 5 DEFERRED MSAACUSTOM RESOLVE & PER-SAMPLE MASK Post G-Buffer, perform a custom msaa resolve  Pre-resolves sample 0, for pixel frequency passes such as lighting/other MSAA dependent passes  In same pass create sub-sample mask (compare samples similarity, mark if mismatching)  Avoid default SV_COVERAGE, since it results in redundant processing on regions not requiring MSAA SV_Coverage Custom Per-Sample Mask
  • 6. Advances in Real-Time Rendering course, Siggraph 2013 6 DEFERRED MSAASTENCIL BATCHING [SOUSA13] Batching per-sample stencil mask with regular stencil buffer usage  Reserve 1 bit from stencil buffer  Update with sub-sample mask  Tag entire pixel-quad instead of just single pixel -> improves stencil culling efficiency  Make usage of stencil read/write bitmask to avoid per-sample bit override  StencilWriteMask = 0x7F  Restore whenever a stencil clear occurs Not possible due to extreme stencil usage?  Use clip/discard  Extra overhead also from additional texture read for per-sample mask
  • 7. Advances in Real-Time Rendering course, Siggraph 2013 7 DEFERRED MSAAPIXEL AND SAMPLE FREQUENCY PASSES Pixel Frequency Passes  Set stencil read mask to reserved bits for per-pixel regions (~0x80)  Bind pre-resolved (non-multisampled) targets SRVs  Render pass as usual Sample Frequency Passes  Set stencil read mask to reserved bit for per-sample regions (0x80)  Bind multisampled targets SRVs  Index current sub-sample via SV_SAMPLEINDEX  Render pass as usual
  • 8. Advances in Real-Time Rendering course, Siggraph 2013 8 DEFERRED MSAAALPHA TEST SSAA Alpha testing requires ad hoc solution  Default SV_Coverage only applies to triangle edges Create your own sub-sample coverage mask  E.g. check if current sub-sample uses AT or not and set bit static const float2 vMSAAOffsets[2] = {float2(0.25, 0.25),float2(-0.25,-0.25)}; const float2 vDDX = ddx(vTexCoord.xy); const float2 vDDY = ddy(vTexCoord.xy); [unroll] for(int s = 0; s < nSampleCount; ++s) { float2 vTexOffset = vMSAAOffsets[s].x * vDDX + (vMSAAOffsets[s].y * vDDY); float fAlpha = tex2D(DiffuseSmp, vTexCoord + vTexOffset).w; uCoverageMask |= ((fAlpha-fAlphaRef) >= 0)? (uint(0x1)<<i) : 0; } Alpha Test SSAA Disabled Alpha Test SSAA Enabled
  • 9. Advances in Real-Time Rendering course, Siggraph 2013 9 DEFERRED MSAAPERFORMANCE SHORTCUTS Deferred cascades sun shadow maps  Render shadows as usual at pixel frequency  Bilateral upscale during deferred shading composite pass
  • 10. Advances in Real-Time Rendering course, Siggraph 2013 10 DEFERRED MSAAPERFORMANCE SHORTCUTS (2) Non-opaque techniques accessing depth (e.g. Soft-Particles)  Recommendation to tackle via per-sample frequency is fairly slow on real world scenarios  Using Max Depth works ok for most cases and N-times faster
  • 11. Advances in Real-Time Rendering course, Siggraph 2013 11 MSAAPERFORMANCE SHORTCUTS (3) Many games, also doing:  Skipping Alpha Test Super Sampling  Use alpha to coverage instead, or even no alpha test AA (let morphological AA tackle that)  Render only opaque with MSAA  Then render transparents withouth MSAA  Assuming HDR rendering: note that tone mapping is implicitly done post-resolve resulting is loss of detail on high contrast regions
  • 12. Advances in Real-Time Rendering course, Siggraph 2013 12 DEFERRED MSAAMSAA FRIENDLINESS Look out for these:  No MSAA noticeably working, or noticeable bright/dark silhouettes. Incorrect Incorrect
  • 13. Advances in Real-Time Rendering course, Siggraph 2013 13 DEFERRED MSAAMSAA FRIENDLINESS Look out for these:  No MSAA noticeably working, or noticeable bright/dark silhouettes. Fixed Fixed
  • 14. Advances in Real-Time Rendering course, Siggraph 2013 14 DEFERRED MSAARECAP Accessing and/or rendering to Multisampled RTs?  Then you need to care about accessing and outputting correct sub-sample In general always strive to minimize BW  Avoid vanilla deferred lighting  Prefer fully deferred, hybrids, or just skip deferred altogether.  If deferred, prefer thin g-buffers  Each additional target on g-buffer incurs in export rate overhead [Thibieroz11]  NV/AMD (GCN): Export Cost = Cost(RT0)+Cost(RT1)..., AMD (older hw): Export Cost = (Num RTs) * (Slowest RT)  Fat formats are half rate sampling cost for bilinear filtering modes on GCN [Thibieroz13]  For lighting/some hdr post processes: 32 bit R11G11B10F fmt suffices for most cases
  • 15. Advances in Real-Time Rendering course, Siggraph 2013 15 ANTIALIASING + 4K RESOLUTIONSWILL WE NEED MSAA AT ALL? Likely can start getting creative here Source 4kHighly Compressed 4k
  • 16. Advances in Real-Time Rendering course, Siggraph 2013 16 ANTIALIASING THE QUEST FOR BETTER (AND FAST) AA 2011: the boom year of alternative AA modes (and naming combos)  FXAA, MLAA, SMAA, SRAA, DEAA, GBAA, DLAA, ETC AA  “Filtering Approaches for Real-Time Anti-Aliasing” [Jimenez et all 11] Shading Anti-aliasing  “Mip mapping normal maps” [Toksvig04]  “Spectacular Specular: LEAN and CLEAN Specular Highlights” [Baker11]  “Rock-Solid Shading: Image Stability withouth Sacrificing Detail” [Hill12]
  • 17. Advances in Real-Time Rendering course, Siggraph 2013 17 TEMPORAL SSAASMAA 2TX/4X REVIEW [JIMENEZ11][SOUSA11] Morphological AA + MSAA + Temporal SSAA combo  Balanced cost/quality tradeoff, techniques complement each other.  Temporal component uses 2 sub-pixel buffers.  Each frame adds a sub-pixel jitter for 2x SSAA.  Reproject previous frame and blend between current and previous frames, via velocity length weighting.  Preserves image sharpness + reasonable temporal stability Naive Blending Reprojection |V| weighting
  • 18. Advances in Real-Time Rendering course, Siggraph 2013 18 TEMPORAL AACOMMON ROBUSTNESS FLAWS Relying on opaque geometry information  Can’t handle signal (color) changes nor transparency.  For correct result, all opaque geometry must output velocity Pathological cases  Alpha blended surfaces (e.g. particles), lighting/shadow/reflections/uv animation/etc  Any scatter and alike post processes, before the AA resolve Can result in distracting errors  E.g. “ghosting” on transparency, lighting, shadows and such  Silhouettes might appear, from scatter and alike post processes (e.g. bloom) Multi-GPU  Simplest solution: force resource sync  NVIDIA exposes driver hint to force sync resource , via NVAPI. This is solution used by NVIDIAs TXAA Note to hw vendors: would be great if all vendors exposed such (even better if Multi-GPU API functionality generalized) Incorrect Blending
  • 19. Advances in Real-Time Rendering course, Siggraph 2013 19 SMAA 1TXA MORE ROBUST TEMPORAL AA Concept: Only track signal changes, don’t rely on geometry information  For higher temporal stability: accumulate multiple frames in an accumulation buffer, alike TXAA [Lottes12]  Re-project accumulation buffer  Weighting: Map acc. buffer colors into the range of curr. frame neighborhood color extents [Malan2012]; different weight for hi/low frequency regions (for sharpness preservation). Current Frame (t0) M Accumulation Buffer (tN) TL TR BL BR M
  • 20. Advances in Real-Time Rendering course, Siggraph 2013 20 SMAA 1TXA MORE ROBUST TEMPORAL AA (2) Concept: Only track signal changes, don’t rely on geometry information  For higher temporal stability: accumulate multiple frames in an accumulation buffer, alike TXAA [Lottes12]  Re-project accumulation buffer  Weighting: Map acc. buffer colors into the range of curr. frame neighborhood color extents [Malan2012]; different weight for hi/low frequency regions (for sharpness preservation). Smaa 1txRegular TAA
  • 21. Advances in Real-Time Rendering course, Siggraph 2013 21 SMAA 1TXA MORE ROBUST TEMPORAL AA (3) Sample code float3 cM = tex2D(tex0, tc.xy); float3 cAcc = tex2D(tex0, reproj_tc.xy); float3 cTL = tex2D(tex0, tc0.xy); float3 cTR = tex2D(tex0, tc0.zw); float3 cBL = tex2D(tex0, tc1.xy); float3 cBR = tex2D(tex0, tc1.zw); float3 cMax = max(cTL, max(cTR, max(cBL, cBR))); float3 cMin = min(cTL, min(cTR, min(cBL, cBR))); float3 wk = abs((cTL+cTR+cBL+cBR)*0.25-cM); return lerp(cM, clamp(cAcc, cMin, cMax), saturate(rcp(lerp(kl, kh, wk)));
  • 22. Advances in Real-Time Rendering course, Siggraph 2013 22 DEPTH OF FIELD
  • 23. Advances in Real-Time Rendering course, Siggraph 2013 23 DEPTH OF FIELDPLAUSIBLE DOF: PARAMETERIZATION Artist friendly parameters is one reason why games DOF tends to look wrong  Typical controls such as “focus range” + “blur amount” and others have not much physical meaning  CoC depends mainly on f-stops, focal length and focal distance. These last 2 directly affect FOV.  If you want more Bokeh, you need to max your focal length + widen aperture. This means also getting closer or further from subject for proper framing. Not the typical way a game artist/programmer thinks about DOF. Wider fov, less bokeh Shallow fov, more bokeh
  • 24. Advances in Real-Time Rendering course, Siggraph 2013 24 50 mm DEPTH OF FIELDFOCAL LENGTH 200 mm
  • 25. Advances in Real-Time Rendering course, Siggraph 2013 25 DEPTH OF FIELDF-STOPS 2 f-stops 22 f-stops8 f-stops
  • 26. Advances in Real-Time Rendering course, Siggraph 2013 26 DEPTH OF FIELDF-STOPS (2) 2 f-stops 2.8 f-stops 4 f-stops 5.6 f-stops
  • 27. Advances in Real-Time Rendering course, Siggraph 2013 27 DEPTH OF FIELDFOCAL DISTANCE 0.5 m
  • 28. Advances in Real-Time Rendering course, Siggraph 2013 28 DEPTH OF FIELDFOCAL DISTANCE 0.75 m
  • 29. Advances in Real-Time Rendering course, Siggraph 2013 29 DEPTH OF FIELDFOCAL DISTANCE 1.0 m
  • 30. Advances in Real-Time Rendering course, Siggraph 2013 30 DEPTH OF FIELDPLAUSIBLE DOF: BOKEH Out of focus region is commonly referred in photography as “Bokeh” (Japanese word for blur) Bokeh shape has direct relation to camera aperture size (aka f-stops) and diaphragm blades count  Bigger aperture = more “circular” bokeh, smaller aperture = more polygonal bokeh Polygonal bokeh look depends on diaphragm blades count Blades count varies on lens characteristics  Bigger aperture = more light enters, smaller aperture = less light On night shots, you might notice often more circular bokeh and more motion blur Bokeh kernel is flat  Almost same amount of light enters camera iris from all directions Edges might be in shadow, this is commonly known as Vignetting Poor lenses manufacturing may introduce a vast array of optical aberrations [Wiki01]  This is main reason why gaussian blur, diffusion dof, and derivative techniques look wrong/visually unpleasant
  • 31. Advances in Real-Time Rendering course, Siggraph 2013 31 DEPTH OF FIELDSTATE OF THE ART OVERVIEW Scatter based techniques [Cyril05][Sawada07][3DMark11][Mittring11][Sousa11]  Render 1 quad or tri per-pixel, scale based on CoC Simple implementation and nice results. Downside: performance, particularly on shallow DOF Variable/inconsistent fillrate hit, depending on near/far layers resolution and aperture size might reach >5 ms Quad generation phase has fixed cost attached.
  • 32. Advances in Real-Time Rendering course, Siggraph 2013 32 DEPTH OF FIELDSTATE OF THE ART OVERVIEW (2) Gather based: separable (inflexible kernel) vs. kernel flexibility [Macintosh12] [Gotanda09] [White11] [Andreev 12] [Kawase09]
  • 33. Advances in Real-Time Rendering course, Siggraph 2013 33 DEPTH OF FIELDA PLAUSIBLE AND EFFICIENT DOF RECONSTRUCTION FILTER Separable flexible filter: low bandwidth requirement + different bokeh shape possible  1st pass N^2 taps (e.g: 7x7).  2nd pass N^2 taps (e.g: 3x3) for flood filling shape  R11G11B10F: downscaled HDR scene; R8G8: CoC  Done at half resolution  Far/Near fields processed in same pass  Limit offset range to minimize undersampling  Higher specs hw can have higher tap count Diaphragm and optical aberrations sim Physically based CoC
  • 34. Advances in Real-Time Rendering course, Siggraph 2013 34 DEPTH OF FIELDLENS REVIEW Pinhole “Lens”  A camera withouth lens  Light has to pass through single small aperture before hitting image plane  Tipical realtime rendering Thin lens  Camera lenses have finite dimension  Light refracts through lens until hitting image plane.  F = Focal lenght  P = Plane in focus  I = Image distance Circle of Confusion Image Plane P I F
  • 35. Advances in Real-Time Rendering course, Siggraph 2013 35 DEPTH OF FIELDLENS REVIEW (2) The thin lens equation gives relation between:  F = Focal length (where light starts getting in focus)  P = Plane in focus (camera focal distance)  I = Image distance (where image is projected in focus) Circle of Confusion [Potmesil81]  f = f-stops (aka as the f-number or focal ratio)  D = Object distance  A = Aperture diameter Simplifies to:  Note: f and F are known variables from camera setup  Folds down into a single mad in shader Camera FOV:  Typical film formats (or sensor), 35mm/70mm  Can alternatively derive focal length from FOV
  • 36. Advances in Real-Time Rendering course, Siggraph 2013 36 DEPTH OF FIELDSAMPLING Concentric Mapping [Shirley97] used for uniform sample distribution  Maps unit square to unit circle  Square mapped to (a,b) [-1,1]2 and divided into 4 regions by lines a=b, a=-b  First region given by: Diaphragm simulation by morphing samples to n-gons  Via a modified equation for the regular polygon.
  • 37. Advances in Real-Time Rendering course, Siggraph 2013 37 DEPTH OF FIELDSAMPLING: 2ND ITERATION To floodfill final shape, composite via boolean union, similarly to [McIntosh12] + U +...= U...=
  • 38. Advances in Real-Time Rendering course, Siggraph 2013 38 DEPTH OF FIELDSEPARABLE FILTER PASSES 1st iteration: 49 taps (0.426ms) 2nd iteration: 9 taps (0.094 ms; 441 taps accumulated)
  • 39. Advances in Real-Time Rendering course, Siggraph 2013 39 DEPTH OF FIELDREFERENCE VS SEPARABLE FILTER 144 taps (1.31ms) 58 taps (0.52ms)
  • 40. Advances in Real-Time Rendering course, Siggraph 2013 40 DEPTH OF FIELDDIAPHRAGM SIMULATION IN ACTION 4f-stops2f-stops
  • 41. Advances in Real-Time Rendering course, Siggraph 2013 41 Tile Min/Max CoC  Downscale CoC target k times (k = tile count)  Take min fragment for far field, max fragment for near field  R8G8 storage Used to process near/far fields in same pass  Dynamic branching using Tile Min/Max CoC for both fields  Balances cost between far/near  Also used for scatter as gather approximation for near field Can fold cost with other post-processes  Initial downscale cost folded with HDR scene downscale for bloom, also pack near/far fields HDR input into R11G11B10F - all in 1 pass DEPTH OF FIELDTILE MIN/MAX COC Tile min-CoC (far field) Tile max-CoC (near field)
  • 42. Advances in Real-Time Rendering course, Siggraph 2013 42 DEPTH OF FIELDFAR + NEAR FIELD PROCESSING Both fields use half resolution input  Careful: downscale is source of error due to bilinear filtering  Use custom bilinear (bilateral) filter for downscaling Far Field  Scale kernel size and weight samples with far CoC [Scheumerman05]  Pre-multiply layer with far CoC [Gotanda09] Prevents bleeding artifacts from bilinear/separable filter No weighting CoC weighting CoC weighting + CoC pre-multiply
  • 43. Advances in Real-Time Rendering course, Siggraph 2013 43 DEPTH OF FIELDFAR + NEAR FIELD PROCESSING Both fields use half resolution input  Careful: downscale is source of error due to bilinear filtering  Use custom bilinear (bilateral) filter for downscaling Far Field  Scale kernel size and weight samples with far CoC [Scheumerman05]  Pre-multiply layer with far CoC [Gotanda09] Prevents bleeding artifacts from bilinear/separable filter Near Field  Scatter as gather aproximation  Scale kernel size + weight fragments with Tile Max CoC against near CoC  Pre-multiply with near CoC Only want to blur near field fragments (cheap partial occlusion approximation) Far Field Near Field
  • 44. Advances in Real-Time Rendering course, Siggraph 2013 44 DEPTH OF FIELDFINAL COMPOSITE Far field: upscale via bilateral filter  Take 4 taps from half res CoC, compare against full res CoC  Weighted using bicubic filtering for quality [Sigg05]  Far field CoC used for blending Near field: upscale carelessly  Half resolution near field CoC used for blending  Can bleed as much as possible  Also using bicubic filtering Carefull with blending  Linear blending doesn’t look good (signal frequency soup) Can be seen in many games, including all Crysis series (puts hat of shame)  Simple to address: use non-linear blend factor instead. Linear blend Non-linear blend (better)
  • 45. Advances in Real-Time Rendering course, Siggraph 2013 45 MOTION BLUR
  • 46. Advances in Real-Time Rendering course, Siggraph 2013 46 MOTION BLURSHUTTER SPEED AND F-STOPS REVIEW Amount of motion blur is relative to camera shutter speed and f-stops usage  The longer the exposure (slower shutter), the more light received (and the bigger amount of motion blur), and vice-versa  The lower f-stops the faster the exposure can be (and have less motion blur), and vice versa 2 f-stops, shutter 1/20 sec 4 f-stops, shutter 1/6 sec
  • 47. Advances in Real-Time Rendering course, Siggraph 2013 47 MOTION BLURSTATE OF THE ART OVERVIEW Scatter via geometry expansion [Green03][Sawada07]  Require additional geometry pass + gs shader usage *
  • 48. Advances in Real-Time Rendering course, Siggraph 2013 48 MOTION BLURSTATE OF THE ART OVERVIEW (2) Scatter as gather [Sousa08][Gotanda09][Sousa11][Maguire12]  E.g.velocity dilation, velocity blur, tile max velocity; single vs. multiple pass composite; depth/v/obj ID masking; single pass DOF+MB
  • 49. Advances in Real-Time Rendering course, Siggraph 2013 49 MOTION BLURRECONSTRUCTION FILTER FOR PLAUSIBLE MB [MCGUIRE12] Tile Max Velocity and Tile Neighbor Max Velocity  Downscale Velocity buffer by k times (k is tile count)  Take max length velocity at each step Motion Blur Pass  Tile Neighbor Max for early out  Tile Max Velocity as center velocity tap  At each iteration step weight against full resolution ||V|| and Depth
  • 50. Advances in Real-Time Rendering course, Siggraph 2013 50 MOTION BLURAN IMPROVED RECONSTRUCTION FILTER Performant Quality  Simplify and vectorize inner loop weight computation (ends up in couple mads)  Fat buffers sampling are half rate on GCN hw with bilinear (point filtering is fullrate, but doesn’t look good due to aliasing)  Inputs: R11G11B10F for scene , bake ||V|| and 8 bit depth into a R8G8 target  Make it separable, 2 passes [Sousa08] 1 iteration: 6 taps (0.236 ms) 2nd iter: 6 taps (+0.236 ms; 36 taps acc.)
  • 51. Advances in Real-Time Rendering course, Siggraph 2013 51 MOTION BLURAN IMPROVED RECONSTRUCTION FILTER (2) Inner loop sample const float2 tc = min_tc + blur_step * s; const float lensq_xy = abs(min_len_xy + len_xy_step * s); const float2 vy = tex2Dlod(tex1, float4(tc.xy, 0, 0)); // x =||v||, y=depth const float2 cmp_z = DepthCmp(float2(vx.y, vy.y), float2(vy.y, vx.y), 1); const float4 cmp_v = VelCmp(lensq_xy, float2(vy.x, lensq_vx)); const float w = (dot(cmp_z.xy, cmp_v.xy) + (cmp_v.z * cmp_v.w) * 2); acc.rgb += tex2Dlod(tex0, float4(tc.xy, 0, 0)) * w; wacc += w; float2 DepthCmp(float2 z0, float2 z1, float2 fSoftZ) { return saturate( (1.0f + z0* fSoftZ) - z1* fSoftZ ); } float4 VelCmp(float lensq_xy, float2 vxy) { return saturate((1.0f - lensq_xy.xxxx *rcp(vxy.xyxy)) + float4(0.0f, 0.0f, 0.95f, 0.95f)); }
  • 52. Advances in Real-Time Rendering course, Siggraph 2013 52 Output object velocity in G-Buffer (only when required)  Avoids separate geometry passes.  Rigid geometry: object distance < distance threshold  Deformable geometry: if amount of movement > movement threshold  Moving geometry rendered last  R8G8 fmt Composite with camera velocity  Velocity encoded in gamma 2.0 space  Precision still insufficient, but not much noticeable in practice MOTION BLURAN IMPROVED RECONSTRUCTION FILTER (3) Encode Decode Object velocity Composite with camera velocity
  • 53. Advances in Real-Time Rendering course, Siggraph 2013 53 MOTION BLURMB OR DOF FIRST? In real world MB/DOF occur simultaneously  A dream implementation: big N^2 kernel + batched DOF/MB  Or sprite based with MB quad stretching  Full resolution! 1 Billion taps! FP16! Multiple layers!  But... performance still matters (consoles):  DOF before MB introduces less error when MB happening on focus This is due MB is a scatter as gather op relying on geometry data. Any other similar op after will introduce error. And vice-versa. Error from MB after DOF is less noticeable.  Order swap makes DOF harder to fold with other posts Additional overhead MB before DOF MB after DOF
  • 54. Advances in Real-Time Rendering course, Siggraph 2013 54 FINAL REMARKS Practical MSAA details  Do’s and Dont’s SMAA 1TX: A More Robust Temporal AA  For just 4 extra texture ops and couple alu A Plausible and Performant DOF Reconstruction Filter  Separable flexible filter, any bokeh kernel shape doable  1st pass: 0.426ms, 2nd pass: 0.094ms. Sum: 0.52ms for reconstruction filter * An Improved Reconstruction Filter for Plausible Motion Blur  Separable, 1st pass: 0.236 ms, 2nd pass: 0.236ms. Sum: 0.472ms for reconstruction filter * * 1080p + AMD 7970
  • 55. Advances in Real-Time Rendering course, Siggraph 2013 55 SPECIAL THANKS Natalya Tartachuck Michael Kopietz, Nicolas Schulz, Christopher Evans, Carsten Wenzel, Christopher Raine, Nick Kasyan, Magnus Larbrant, Pierre Donzallaz
  • 56. Advances in Real-Time Rendering course, Siggraph 2013 56 WE ARE HIRING !
  • 57. Advances in Real-Time Rendering course, Siggraph 2013 57 QUESTIONS ? Tiago@Crytek.Com / Twitter: Crytek_Tiago
  • 58. Advances in Real-Time Rendering course, Siggraph 2013 58 REFERENCES Potmesil M., Chakravarty I. “Synthetic Image Generation with a Lens and Aperture Camera Model”, 1981 Shirley P., Chiu K., “A Low Distortion Map Between Disk and Square”, 1997 Green S. “Stupid OpenGL Tricks”, 2003 Toksvig M., “Mip Mapping Normal Maps”, 2004 Scheuermann T., Tatarchuk N. “Improved Depth of Field Rendering”, Shader X3, 2005 Sigg C., Hadwiger M., “Fast Third-Order Texture Filtgering”, 2005 Cyril, P et al. “Photographic Depth of Field Rendering”, 2005 Gritz L., Eon E. “The Importance of Being Linear”, 2007 Sawada Y., Talk at Game Developers Conference, http://www.beyond3d.com/content/news/499 , 2007 Sousa T. “Crysis Next Gen Effects”, 2008 Gotanda Y. ”Star Ocean 4: Flexible Shader Management and Post Processing”, 2009 Yang L. et al, “Amortized SuperSampling”, 2009 Kawase M., “Anti-Downsized Buffer Artifacts”, 2011 Sousa T., “CryENGINE 3 Rendering Techniques”, 2011 Binks D., “Dynamic Resolution Rendering”, 2011 Sousa T., Schulz N. , Kasyan N., “Secrets of CryENGINE 3 Graphics Technology”, 2011 Sousa T., “Anti-Aliasing Methods in CryENGINE 3”, 2011 Jimenez J. et.al, “Filtering Approaches for Real-Time Anti-Aliasing”, 2011 Thibieroz N., “Deferred Shading Optimizations”, 2011
  • 59. Advances in Real-Time Rendering course, Siggraph 2013 59 REFERENCES Cloward B., Otstott A., “Cinematic Character Lighting in Star Wars: The Old Republic”, 2011 Mittring M., Dudash B., “The Technology Behind the DirectX11 Unreal Engine “Samaritan” Demo“, 2011 Futuremark, 3DMark11 “White Paper”, 2011 Baker D., “Spectacular Specular: LEAN and CLEAN Specular Highlights”, 2011 White J., Brisebois C., “More Performance! Five Rendering Ideas from Batlefield 3 and Need for Speed: The Run”, 2011 Lottes T., “Nvidia TXAA”, http://www.nvidia.in/object/txaa-anti-aliasing-technology-in.html#gameContent=1 , 2012 McGuire M. et al, “A Reconstruction Filter for Plausible Motion Blur”, 2012 McIntosh L. Et al, “Efficiently Simulating Bokeh”, 2012 Malan H., “Real-Time Global Illumination and Reflections in Dust 514”, 2012 Hill S., Baker D., “Rock-Solid Shading: Image Stability withouth Sacrificing Detail”, 2012 Sousa T., Wenzel C., Raine C., “Rendering Technologies of Crysis 3”, 2013 Thibieroz N., Gruen H., “DirectX11 Performance Reloaded”, 2013 Andreev D., “Rendering Tricks in Dead Space 3”, 2013 http://en.wikipedia.org/wiki/Lens_%28optics%29 http://en.wikipedia.org/wiki/Pinhole_camera http://en.wikipedia.org/wiki/F-number http://en.wikipedia.org/wiki/Angle_of_view