D3DX 8.1 D3DX 8.1

Anuj GosaliaAnuj GosaliaDevelopment LeadDevelopment LeadDirectXDirectX®® GraphicsGraphics

MicrosoftMicrosoft®® Corporation Corporation

D3DX ReleasesD3DX Releases

Shipped D3DX 8.0 with DirectX 8.0 SDKShipped D3DX 8.0 with DirectX 8.0 SDK D3DX 8.0b released to WebD3DX 8.0b released to Web

Bug fixes to D3DX 8.0, no new featuresBug fixes to D3DX 8.0, no new features

D3DX 8.1D3DX 8.1 Includes new featuresIncludes new features Now in BetaNow in Beta

Overview Of D3DX 8.1Overview Of D3DX 8.1

Mesh UtilitiesMesh Utilities Effect FrameworkEffect Framework

Shader assemblersShader assemblers

Texture UtilitiesTexture Utilities Math UtilitiesMath Utilities Miscellaneous UtilitiesMiscellaneous Utilities Authoring tool supportAuthoring tool support

Mesh LibraryMesh Library

Progressive meshesProgressive meshes N-Patch tessellationN-Patch tessellation Mesh optimizationMesh optimization Skinned meshesSkinned meshes Other mesh utilitiesOther mesh utilities

Bounding volume generation (sphere, box)Bounding volume generation (sphere, box) Ray intersections (mesh, sphere, box)Ray intersections (mesh, sphere, box) Mesh cleanupMesh cleanup More…More…

Mesh BasicsMesh Basics

Vertex Buffer, Index Buffer, Vertex Buffer, Index Buffer, and Attributesand Attributes

Indexed Triangle listsIndexed Triangle lists 16/32-bit indices supported16/32-bit indices supported Supports file I/O (via .X)Supports file I/O (via .X) Can be used independently of .X filesCan be used independently of .X files DrawSubset is only for convenienceDrawSubset is only for convenience

Not the only way to draw a meshNot the only way to draw a mesh Manipulates adjacency if requestedManipulates adjacency if requested

Attributes: Buffer? Table?Attributes: Buffer? Table?

Mesh has 1 DWORD per triangle (face)Mesh has 1 DWORD per triangle (face) Stored in mesh object as Attribute BufferStored in mesh object as Attribute Buffer

Semantics of values is up to the appSemantics of values is up to the app Need not be sequentialNeed not be sequential

Attribute TableAttribute Table A compact representation A compact representation

of the attribute bufferof the attribute buffer Generated by Attribute Sorting a meshGenerated by Attribute Sorting a mesh GetAttributeTable, no SetAttributeTableGetAttributeTable, no SetAttributeTable

Mesh RenderingMesh Rendering

DrawSubset() draws all triangles DrawSubset() draws all triangles of a given attributeof a given attribute

Needs Attribute TableNeeds Attribute Table Else it does linear search per faceElse it does linear search per face

Efficient if attributes are sequential, Efficient if attributes are sequential, starting from 0starting from 0 Else it does search of attribute tableElse it does search of attribute table

Uses Fixed Function FVF shaderUses Fixed Function FVF shader Avoid unless all above conditions metAvoid unless all above conditions met

Mesh Adjacency In D3DXMesh Adjacency In D3DX

Many mesh operations Many mesh operations require adjacencyrequire adjacency

Array of 3 DWORDs per faceArray of 3 DWORDs per face Each DWORD is a face indexEach DWORD is a face index 0xffffffff means no adjacent face0xffffffff means no adjacent face

All mesh operations that change All mesh operations that change adjacency will optionally return adjacency will optionally return updated adjacencyupdated adjacency

Load from .X returns adjacencyLoad from .X returns adjacency

Point RepresentativesPoint Representatives

Alternate way of encoding adjacency infoAlternate way of encoding adjacency info Keeps track of vertices which have the same Keeps track of vertices which have the same

position but replicated due to differing position but replicated due to differing attributes (like normals, tex coords, etc.)attributes (like normals, tex coords, etc.)

One DWORD per vertexOne DWORD per vertex All vertices in a set of replicated vertices point All vertices in a set of replicated vertices point

to any one of them as a “representative”to any one of them as a “representative” Non-replicated vertices point to themselvesNon-replicated vertices point to themselves

Meshes And AdjacencyMeshes And Adjacency

Can convert from PRep to adjacency Can convert from PRep to adjacency and backand back

Generating adjacency from scratchGenerating adjacency from scratch Can use identity Prep, ignoring duplicatesCan use identity Prep, ignoring duplicates

Works in some casesWorks in some cases

GenerateAdjacecncy() will identify vertices GenerateAdjacecncy() will identify vertices with same position (i.e., infer PRep)with same position (i.e., infer PRep) Slower than aboveSlower than above Will get correct adjacency if epsilon Will get correct adjacency if epsilon

is appropriateis appropriate

Remap ArraysRemap Arrays

Describes how mesh was rearrangedDescribes how mesh was rearranged 1 DWORD for each destination face / 1 DWORD for each destination face /

vertexvertex Indicates which face / vertex of source it Indicates which face / vertex of source it

came fromcame from Many-to-one mapping possibleMany-to-one mapping possible Allows mesh related data outside mesh Allows mesh related data outside mesh

object to be updated with the meshobject to be updated with the mesh

Mesh OptimizationMesh Optimization

StripifyStripify Rearrange vertices of a mesh in strip orderRearrange vertices of a mesh in strip order

Vertex cache optimizeVertex cache optimize Based on Hugues Hoppe’s Based on Hugues Hoppe’s

Siggraph ‘99 paperSiggraph ‘99 paper Hardware specific optimizationHardware specific optimization

Both need adjacency informationBoth need adjacency information ConvertPointRepsToAdjacency with NULL ConvertPointRepsToAdjacency with NULL

(identity) PRep array will suffice(identity) PRep array will suffice

Mesh OptimizationMesh Optimization

Attribute sortAttribute sort Sorts faces and vertices on the Sorts faces and vertices on the

attribute idsattribute ids Splits shared vertices if necessarySplits shared vertices if necessary Generates Attribute TableGenerates Attribute Table

Compact MeshCompact Mesh Eliminates vertices not referred to Eliminates vertices not referred to

by the index arrayby the index array

Sharing Vertex BuffersSharing Vertex Buffers

Typically Optimize re-arranges vertices Typically Optimize re-arranges vertices and indicesand indices

If vertices already ordered by attribute, If vertices already ordered by attribute, src & dest mesh can share VBssrc & dest mesh can share VBs

D3DXMESHOPT_SHAREVBD3DXMESHOPT_SHAREVB Useful for clones and optimizingUseful for clones and optimizing

Offline cache optimizationOffline cache optimization

Best done at load timeBest done at load time Algorithm is fastAlgorithm is fast

““Default” is Geforce 1,2Default” is Geforce 1,2 Works well on all cardsWorks well on all cards

Optimize on above or card with no Optimize on above or card with no hardware T&Lhardware T&L

Meshes and Tri-StripsMeshes and Tri-Strips

D3DXConvertMeshSubsetToStripsD3DXConvertMeshSubsetToStrips D3DXConvertMeshSubsetToSingleStripD3DXConvertMeshSubsetToSingleStrip Returns new Index Buffer separate from Returns new Index Buffer separate from

the mesh objectthe mesh object Works on any meshWorks on any mesh

Helps to optimize it for vertex cache or Helps to optimize it for vertex cache or stripifystripify

May be a performance win in some May be a performance win in some specific casesspecific cases

Use OptimizeMesh sample to see what Use OptimizeMesh sample to see what works bestworks best

Progressive MeshesProgressive MeshesOverviewOverview

Generate an ID3DXPMesh object from Generate an ID3DXPMesh object from high poly-count mesh using high poly-count mesh using ID3DXSPMesh objectID3DXSPMesh object Done either offline or load timeDone either offline or load time

Render the ID3DXPMesh object at any Render the ID3DXPMesh object at any LOD at runtimeLOD at runtime

Generate a bunch of ID3DXMesh Generate a bunch of ID3DXMesh objects from ID3DXPMesh objectobjects from ID3DXPMesh object

Progressive MeshesProgressive MeshesMesh SimplificationMesh Simplification

Based on Garland-Heckbert Based on Garland-Heckbert quadric error metricquadric error metric

Incorporates refinements by Hugues Incorporates refinements by Hugues Hoppe to accommodate normal and Hoppe to accommodate normal and attribute space metricsattribute space metrics

Needs accurate adjacency informationNeeds accurate adjacency information

Progressive MeshesProgressive MeshesMesh Simplification(2)Mesh Simplification(2)

API for simplification via ID3DXSPMesh objectAPI for simplification via ID3DXSPMesh object No more batch files No more batch files Allows you to incorporate automated LOD Allows you to incorporate automated LOD

generation in your internal toolsgeneration in your internal tools

User controls to influence simplification User controls to influence simplification processprocess Assigning weights to verticesAssigning weights to vertices Weighing the importance of various vertex attributesWeighing the importance of various vertex attributes

Progressive MeshesProgressive MeshesHalf-edge collapsesHalf-edge collapses

Chooses one of the two original vertices Chooses one of the two original vertices during each edge collapseduring each edge collapse

No significant quality degradationNo significant quality degradation Mesh vertices never change with LODMesh vertices never change with LOD

Enables mixing PM and mesh deformation Enables mixing PM and mesh deformation algorithms like morphing and skinningalgorithms like morphing and skinning

Reduces the amount of information Reduces the amount of information stored in a vertex split recordstored in a vertex split record LOD changes are fasterLOD changes are faster

Progressive MeshesProgressive MeshesDynamic LOD changesDynamic LOD changes

ID3DXPMesh object allows dynamic ID3DXPMesh object allows dynamic LOD changes to arbitrary LOD changes to arbitrary face/vertex countsface/vertex counts

LOD changes are fast enough LOD changes are fast enough to do at runtimeto do at runtime

Modifies the index buffer and Modifies the index buffer and the adjacencythe adjacency

Progressive MeshesProgressive MeshesCloningCloning

Support sharing the vertex data Support sharing the vertex data across clonesacross clones

Can “clone” multiple ID3DXMesh Can “clone” multiple ID3DXMesh objects from a progressive mesh, all objects from a progressive mesh, all of which share the same VBof which share the same VB Can even optimize the resultant mesh while Can even optimize the resultant mesh while

sharing the original VBsharing the original VB

Progressive MeshesProgressive MeshesPersistencyPersistency

Persist to IStreamPersist to IStream Can embed PMs in any custom file formatCan embed PMs in any custom file format

ID3DXPMesh::SaveID3DXPMesh::Save D3DXCreatePMeshFromStreamD3DXCreatePMeshFromStream

Progressive MeshesProgressive MeshesOptimizationOptimization

PMesh face ordering may not be PMesh face ordering may not be cache optimalcache optimal

Can at least make base mesh optimizedCan at least make base mesh optimized ID3DXPMesh::OptimizeBaseLODID3DXPMesh::OptimizeBaseLOD

Use multiple clones of PMesh with Use multiple clones of PMesh with increasing base LODsincreasing base LODs ID3DXPMesh::TrimByVerticesID3DXPMesh::TrimByVertices ID3DXPMesh::TrimByFacesID3DXPMesh::TrimByFaces Can share VB across clonesCan share VB across clones

Switch to PMesh with highest base LODSwitch to PMesh with highest base LOD

N-Patch TessellationN-Patch Tessellation

D3DX provides software N-Patch D3DX provides software N-Patch tessellationtessellation

Uses adjacency to share vertices Uses adjacency to share vertices in tessellated meshin tessellated mesh

Assumes mesh is smoothAssumes mesh is smooth Any sharp edges due to normal Any sharp edges due to normal

discontinuity will cause cracksdiscontinuity will cause cracks Use D3DXWeldVertices to merge normals Use D3DXWeldVertices to merge normals

within epsilonwithin epsilon Improved in D3DX 8.1 to make Improved in D3DX 8.1 to make

welding normals lot easierwelding normals lot easier

Other Mesh UtilitiesOther Mesh Utilities

Compute bounding box and sphereCompute bounding box and sphere Compute normalsCompute normals Ray mesh intersectionRay mesh intersection

Returns triangle index and barycentric Returns triangle index and barycentric coordinates of point of intersection if hitcoordinates of point of intersection if hit

Ray box and sphere intersectionRay box and sphere intersection Clean-up topology for simplificationClean-up topology for simplification Cloning for VB and IB format Cloning for VB and IB format

conversionconversion

Mesh Library ImprovementsMesh Library Improvements

D3DXSplitMeshD3DXSplitMesh Use to split large 32-bit meshes into Use to split large 32-bit meshes into

multiple 16-bit meshesmultiple 16-bit meshes Splits shared verticesSplits shared vertices

Minimized if mesh is vertex cache optimizedMinimized if mesh is vertex cache optimized

D3DXWeldVerticesD3DXWeldVertices Takes per component epsilonsTakes per component epsilons Does partial weldsDoes partial welds

Mesh IntersectionMesh Intersection

Intersect ray with tri, mesh or mesh Intersect ray with tri, mesh or mesh subsetsubset

Returns face and barycentric Returns face and barycentric coordinates of intersectioncoordinates of intersection

Optionally returns list of all Optionally returns list of all intersectionsintersections

Needs no precomputationNeeds no precomputation Efficient algorithm for hit testing, etcEfficient algorithm for hit testing, etc Not efficient for too many intersections for Not efficient for too many intersections for

the same meshthe same mesh

Compute Tangent SpaceCompute Tangent Space

Create a per vertex coordinate systemCreate a per vertex coordinate system Normal define one axisNormal define one axis Texture coordinate (u,v) gradients Texture coordinate (u,v) gradients

used to orient tangentsused to orient tangents Use u to define one tangent & compute Use u to define one tangent & compute

binorm by cross productbinorm by cross product Or use u & v to define both tangentsOr use u & v to define both tangents

Compute Tangent Space (2)Compute Tangent Space (2)

Mesh texture parameterizations can Mesh texture parameterizations can have orientation flipshave orientation flips

Cross product binormal can be reverse Cross product binormal can be reverse from v space gradient in some partsfrom v space gradient in some parts

Solution: Encode binormal sign per Solution: Encode binormal sign per vertexvertex Use 4D vector for encoding per-vertex Use 4D vector for encoding per-vertex

tangenttangent Put sign in 4Put sign in 4thth component component Invert computed binormal in vertex shaderInvert computed binormal in vertex shader

Skinned MeshesSkinned Meshes

Plug-ins for authoring tools to export Plug-ins for authoring tools to export skinning dataskinning data 3D Studio Max and Character Studio3D Studio Max and Character Studio Maya (work in progress)Maya (work in progress)

.X files extended to handle .X files extended to handle skinning dataskinning data D3DX functions to load skinned meshesD3DX functions to load skinned meshes

ID3DXSkinMesh independent of .X filesID3DXSkinMesh independent of .X files

Skinned Mesh ObjectSkinned Mesh Object

Contains a mesh object plus Contains a mesh object plus skinning dataskinning data

Skinning data supplied as a bone Skinning data supplied as a bone and a list of vertices it affectsand a list of vertices it affects And a weight corresponding to each vertexAnd a weight corresponding to each vertex

Though not hardware friendly, this input Though not hardware friendly, this input method is simple and generalmethod is simple and general

Can convert to optimized formsCan convert to optimized forms

Skinning Technique #1Skinning Technique #1

Direct3DDirect3D®® 7.0 style 7.0 style Per vertex weightsPer vertex weights Up to 4 bones (matrices) per triangleUp to 4 bones (matrices) per triangle

Or patch if using R/T-PatchesOr patch if using R/T-Patches

ConvertToBlendedMesh generates ConvertToBlendedMesh generates a mesh with per vertex weightsa mesh with per vertex weights

Can cause mesh to have Can cause mesh to have many “subsets”many “subsets”

Works with well N-Patch tessellationWorks with well N-Patch tessellation

Skinning Technique #2Skinning Technique #2

Introduced in Direct3D 8.0Introduced in Direct3D 8.0 Per vertex indices refer to matrices from Per vertex indices refer to matrices from

a palette that affect it a palette that affect it Up to 4 indices per vertex, 12 per faceUp to 4 indices per vertex, 12 per face Up to 256 matrices in a paletteUp to 256 matrices in a palette

Reduces API calls and matrix changesReduces API calls and matrix changes ConvertToIndexedBlendedMesh ConvertToIndexedBlendedMesh

generates mesh with per vertex weights generates mesh with per vertex weights and matrix indicesand matrix indices

Skinning Technique #3Skinning Technique #3

Software skinning in D3DXSoftware skinning in D3DX Arbitrary number of influences Arbitrary number of influences

per vertexper vertex Useful for skinning curved surface Useful for skinning curved surface

control meshcontrol mesh Useful for accessing post skinned Useful for accessing post skinned

mesh datamesh data Hit testing skinned meshesHit testing skinned meshes

GenerateSkinnedMesh() / GenerateSkinnedMesh() / UpdateSkinnedMesh() does thisUpdateSkinnedMesh() does this

ConvertToBlendedMeshConvertToBlendedMesh

Truncates bone influences when Truncates bone influences when >4 per triangle exists>4 per triangle exists Keeps the 4 most important weightsKeeps the 4 most important weights Uses adjacency info to avoid cracksUses adjacency info to avoid cracks

Orders bone combinations by Orders bone combinations by increasing # of influencesincreasing # of influences Enables using GeForce’s restricted Enables using GeForce’s restricted

skinned support by rendering a prefix skinned support by rendering a prefix of the mesh in hardwareof the mesh in hardware

Use software for the restUse software for the rest

ConvertToIndexedBlended…ConvertToIndexedBlended…

Will truncate if >4 influences per vertexWill truncate if >4 influences per vertex Handles palette sizes < num bonesHandles palette sizes < num bones

But must be > maxFaceInflBut must be > maxFaceInfl

Partitions mesh into subsets that Partitions mesh into subsets that fit in a palettefit in a palette

Output can be used with vertex shadersOutput can be used with vertex shaders Output mesh has only necessary # Output mesh has only necessary #

of weightsof weights Use Clone to pad extra weights if shader Use Clone to pad extra weights if shader

expects fixed #expects fixed #

Skinning PerformanceSkinning Performance

Minimize # of bone combinations?Minimize # of bone combinations? Can merge subset combinationsCan merge subset combinations Increases # of blendsIncreases # of blends

Improve matrix coherence across Improve matrix coherence across combinations?combinations? Can’t prevent extra DrawPrim callsCan’t prevent extra DrawPrim calls Can’t prevent matrix concatenationCan’t prevent matrix concatenation Does not seem worthwhileDoes not seem worthwhile

Skinning Performance (2)Skinning Performance (2)

Non-hardware T&L devicesNon-hardware T&L devices Indexed palette skinning using FF pipeline Indexed palette skinning using FF pipeline

or vertex shaders is bestor vertex shaders is best

On GeForce 1,2 and Radeon, On GeForce 1,2 and Radeon, non-indexed skinning is fastestnon-indexed skinning is fastest

On Geforce 3 indexed skinning using On Geforce 3 indexed skinning using vertex shader is fastest ?vertex shader is fastest ?

Disclaimer: Your mileage may vary…Disclaimer: Your mileage may vary…

SW Skinning PerformaceSW Skinning Performace

Skin on CPU instead of GPUSkin on CPU instead of GPU CPU/GPU load balancingCPU/GPU load balancing Multipass renderingMultipass rendering

33% faster skinning in D3DX 8.133% faster skinning in D3DX 8.1 Consider using multiple streamsConsider using multiple streams

Minimize data processed by CPUMinimize data processed by CPU

Skinning PMeshesSkinning PMeshes

Skinning causes mesh to be split into Skinning causes mesh to be split into subsets, adversely affecting subsets, adversely affecting simplification qualitysimplification quality

Using Indexed skinning reduces Using Indexed skinning reduces subsets (1 if palette size >= num bones)subsets (1 if palette size >= num bones)

Call ConvertTo* and use result Call ConvertTo* and use result to create PMeshto create PMesh

Simplification And SkinningSimplification And Skinning

Simplification ignores geometry Simplification ignores geometry changes due to skinningchanges due to skinning

Default pose of mesh (figure mode?) Default pose of mesh (figure mode?) may not be best to simplifymay not be best to simplify Many joint (elbows, knees, etc.) are straightMany joint (elbows, knees, etc.) are straight Geometric error when simplifying across Geometric error when simplifying across

joints lower than would be when joint joints lower than would be when joint is bentis bent

Choose some different pose for Choose some different pose for simplification (How?)simplification (How?)

Skinning And NPatchesSkinning And NPatches

Tessellating indices is messyTessellating indices is messy Use software skinning of control point Use software skinning of control point

meshmesh Use only if hardware is doing full Use only if hardware is doing full

tessellationtessellation

Use non-indexed skinning Use non-indexed skinning of tessellated meshof tessellated mesh ConvertToBlendedMesh firstConvertToBlendedMesh first Tessellate the resultTessellate the result Update bone combination table with new Update bone combination table with new

attribute tableattribute table

Call To ActionCall To Action

Try out new features in DirectX 8.1Try out new features in DirectX 8.1 Give us feedbackGive us feedback Tell us about bugs and performance Tell us about bugs and performance

issuesissues What else would you like to see?What else would you like to see? Hang around for the next talk…Hang around for the next talk…

AcknowledgementsAcknowledgements

Thanks to Origin Systems for Thanks to Origin Systems for permission to use Unicorn modelpermission to use Unicorn model

Thanks to NewTek for permission to use Thanks to NewTek for permission to use the monster modelthe monster model

Questions ?Questions ?

D3DX 8.1 D3DX 8.1

Anuj GosaliaAnuj GosaliaDevelopment LeadDevelopment LeadDirectXDirectX®® Graphics Graphics

Microsoft CorporationMicrosoft Corporation

Overview Of D3DX 8.1Overview Of D3DX 8.1

Mesh UtilitiesMesh Utilities Effect FrameworkEffect Framework

Shader assemblersShader assemblers

Texture UtilitiesTexture Utilities Math UtilitiesMath Utilities Miscellaneous UtilitiesMiscellaneous Utilities Authoring tool supportAuthoring tool support

Effect FrameworkEffect Framework

Encapsulation of device stateEncapsulation of device state Enables scalable rendering techniquesEnables scalable rendering techniques Allows controlled fallbackAllows controlled fallback Can’t just switch to multi-passCan’t just switch to multi-pass

Older hardware can’t do more passes Older hardware can’t do more passes since alpha blending fill rate is lesssince alpha blending fill rate is less

Helps rapid prototypingHelps rapid prototyping Runtime interpretation of text-based Runtime interpretation of text-based

effect definitioneffect definition

Effect FrameworkEffect FrameworkFallback TechniquesFallback Techniques

Uses controlled effect fallbacksUses controlled effect fallbacks

EffectEffect

TechniqueTechnique

PassPassImplementationImplementation

Simple text file (.fx) to define effectsSimple text file (.fx) to define effects

Effect FrameworkEffect FrameworkFallback TechniquesFallback Techniques

Techniques are grouped by their quality Techniques are grouped by their quality or “LOD”or “LOD”

Techniques can be chosen based on Techniques can be chosen based on what hardware creates successfullywhat hardware creates successfully

Can test performance in back bufferCan test performance in back buffer User responsible for drawing geometryUser responsible for drawing geometry

Effect FrameworkEffect FrameworkCreating EffectsCreating Effects

D3DXCompileEffectFromFileD3DXCompileEffectFromFile Parses text fileParses text file

D3DXCreateEffectD3DXCreateEffect Use compiled effect to create Use compiled effect to create

an effect objectan effect object

State for each pass is encoded State for each pass is encoded as state blocksas state blocks

Effect Data typesEffect Data types

DWORD, FLOATDWORD, FLOAT VECTOR, MATRIXVECTOR, MATRIX TEXTURETEXTURE VERTEXSHADER, PIXELSHADERVERTEXSHADER, PIXELSHADER STRINGSTRING

Enables user-data associated with effectsEnables user-data associated with effects Not used to program device stateNot used to program device state

Parameterized EffectsParameterized Effects

Effects can have parameters Effects can have parameters of various typesof various types

Parameters augment static Parameters augment static state description in the .fx filesstate description in the .fx files

How (and which) parameters get How (and which) parameters get used defined by the effectused defined by the effect

Effect ImprovemetsEffect Improvemets

Support for longer namesSupport for longer names No longer limited to FourCCNo longer limited to FourCC

Enable ordinal or string based Enable ordinal or string based parameter resolutionparameter resolution

Block comment /* */ supportBlock comment /* */ support Merge ID3DXEffect and ID3DXTechniqueMerge ID3DXEffect and ID3DXTechnique

Need to carry around only 1 pointerNeed to carry around only 1 pointer

OnLost() and OnReset() methodsOnLost() and OnReset() methods

Effect FrameworkEffect FrameworkShader AssembliesShader Assemblies

In-line or load from fileIn-line or load from file VertexVertex

D3DXAssembleVertexShader()D3DXAssembleVertexShader() D3DXAssembleVertexShaderFromFile()D3DXAssembleVertexShaderFromFile()

PixelPixel D3DXAssemblePixelShader()D3DXAssemblePixelShader() D3DXAssemblePixelShaderFromFile()D3DXAssemblePixelShaderFromFile()

Shape LibraryShape Library

Regular polygonRegular polygon BoxBox Cylinder/ConeCylinder/Cone SphereSphere TorusTorus And, of course, the teapotAnd, of course, the teapot Optional adjacency info availableOptional adjacency info available

2D Text2D Text

Draw text to surface using GDIDraw text to surface using GDI Render to off screen DCRender to off screen DC Blit to an internal textureBlit to an internal texture Render using quadRender using quad

Cache output by rendering to a textureCache output by rendering to a texture Supports all GDI features: italics, Supports all GDI features: italics,

kerning, international fonts, etc.kerning, international fonts, etc. ID3DXFont::DrawTextID3DXFont::DrawText

Dynamic 2D textDynamic 2D text

Using GDI every time can be slowUsing GDI every time can be slow Render alphabet to a textureRender alphabet to a texture Render a quad per characterRender a quad per character Texture coordinates into the texture Texture coordinates into the texture

depend on the characterdepend on the character Works well with simple fontsWorks well with simple fonts

Not for international fonts, kerning, etc.Not for international fonts, kerning, etc.

CD3DFont in sample framework CD3DFont in sample framework does thisdoes this

3D Text3D Text

D3DXCreateTextD3DXCreateText Extrudes a string rendered using Extrudes a string rendered using

a TrueType fonta TrueType font Returns a mesh objectReturns a mesh object Does not handleDoes not handle

Kerning, etc.Kerning, etc. International font spacingInternational font spacing

Sprites (not point sprites!)Sprites (not point sprites!)

Draws image in a texture to screenDraws image in a texture to screen Using a textured quadUsing a textured quad

Alpha blendingAlpha blending Rotation, scalesRotation, scales Arbitrary transforms & warpsArbitrary transforms & warps For performanceFor performance

Draw multiple sprites between Begin/EndDraw multiple sprites between Begin/End Draw mutiple sprites from same textureDraw mutiple sprites from same texture

Rendering to TexturesRendering to Textures

ID3DXRenderToSurface abstractionID3DXRenderToSurface abstraction BeginBegin

Setup render targets, viewportsSetup render targets, viewports Use intermediate surface if necessaryUse intermediate surface if necessary call BeginScenecall BeginScene

EndEnd CleanupCleanup Call EndSceneCall EndScene Blit to dest if necessaryBlit to dest if necessary

Texture UtilitiesTexture Utilities

Image file loadersImage file loaders JPG, PNG, TGA, BMP, PPM, DDSJPG, PNG, TGA, BMP, PPM, DDS Supports files in memorySupports files in memory

Format conversionFormat conversion Image re-samplingImage re-sampling

Better filtering optionsBetter filtering options Supports wrap modesSupports wrap modes

Mip-map generationMip-map generation Color-key to alpha conversionColor-key to alpha conversion

DXTn Encode QualityDXTn Encode Quality

New high quality compression New high quality compression algorithmalgorithm

Fast enough for load-time compressionFast enough for load-time compression 75-95% of earlier algorithm75-95% of earlier algorithm

Dithers while encodingDithers while encoding Avoids blocking of smooth gradientsAvoids blocking of smooth gradients

Improved encoding for alpha imagesImproved encoding for alpha images

DXTn Encoding examplesDXTn Encoding examples

Texture utilities updateTexture utilities update

D3DXGetImageInfoFrom*()D3DXGetImageInfoFrom*() Info about image before loading itInfo about image before loading it Include file format infoInclude file format info Enables calling appropriate load functionEnables calling appropriate load function

D3DXLoadSurfaceFromSurface D3DXLoadSurfaceFromSurface performanceperformance Will use hardware if possibleWill use hardware if possible

Support for dynamic texturesSupport for dynamic textures

Image SaveImage Save

D3DXSaveSurfaceToFileD3DXSaveSurfaceToFile BMPsBMPs

8-bit paletted8-bit paletted 24-bit RGB24-bit RGB

DDSDDS All formatsAll formats Mip-maps, cube-maps, volumesMip-maps, cube-maps, volumes

New scratch poolNew scratch pool

D3DPOOL_SCRATCHD3DPOOL_SCRATCH Allows creation of resources that are Allows creation of resources that are

not limited by device capabilitiesnot limited by device capabilities Create-Destroy, Lock-UnlockCreate-Destroy, Lock-Unlock

Can set to device, use in renderingCan set to device, use in rendering

Use with D3DX to convert to something Use with D3DX to convert to something useableuseable

e.g., Load high-prec height field and e.g., Load high-prec height field and convert to device prec normal mapconvert to device prec normal map

Texture FillTexture Fill

Texture fill functionsTexture fill functions D3DXFillTextureD3DXFillTexture D3DXFillCubeTextureD3DXFillCubeTexture D3DXFillVolumeTextureD3DXFillVolumeTexture

Handles mip-mapsHandles mip-maps Callback function gets a 2D/3D Callback function gets a 2D/3D

location and size of texellocation and size of texel Encode functions as look-up tables for Encode functions as look-up tables for

pixel shaderspixel shaders

Bump MappingBump Mapping

D3DXComputeNormalMapD3DXComputeNormalMap Converts a height field to a normal mapConverts a height field to a normal map Looks at 8 neighbors to calculate slopeLooks at 8 neighbors to calculate slope Calculates occlusion term in alphaCalculates occlusion term in alpha

Rough estimate of what fraction of the Rough estimate of what fraction of the hemisphere at that location in the height field is hemisphere at that location in the height field is “sky”“sky”

Smooth gradients can have aliasingSmooth gradients can have aliasing Use high-precision height fieldUse high-precision height field D3DX now supports 16-bit formatsD3DX now supports 16-bit formats

Math Library ImprovementsMath Library Improvements

D3DXQuaternionSqaudSetupD3DXQuaternionSqaudSetup Use with D3DXQuaternionSqaudUse with D3DXQuaternionSqaud

D3DXMatrixMultiplyTransposeD3DXMatrixMultiplyTranspose For matrices in vertex shadersFor matrices in vertex shaders

D3DXFresnelTermD3DXFresnelTerm Useful along with texture fill functionsUseful along with texture fill functions

Math library optimizationMath library optimization

CPU specific optimizations CPU specific optimizations for most important functionsfor most important functions 3DNow, SSE and SSE23DNow, SSE and SSE2 Vector, matrix, quaternion, interpolation, …Vector, matrix, quaternion, interpolation, …

Auto-detect CPU typeAuto-detect CPU type First call to an optimized math function First call to an optimized math function

detects CPUdetects CPU Patches jump table so no additional Patches jump table so no additional

overhead for subsequent callsoverhead for subsequent calls

Aligned MatricesAligned Matrices

Support for 16-byte aligned matricesSupport for 16-byte aligned matrices D3DXMATRIXA16D3DXMATRIXA16 Uses declspec(align:16) on new compilersUses declspec(align:16) on new compilers

Visual C++Visual C++®® 6 + processor pack6 + processor pack Visual C++ 7 (future product)Visual C++ 7 (future product) Not in Visual C++ 6 service packsNot in Visual C++ 6 service packs

Aligns on stack, members, globalsAligns on stack, members, globals Overloaded new / delete for aligned heap Overloaded new / delete for aligned heap

allocationsallocations Use with care when embedding in structsUse with care when embedding in structs

Authoring Tool SupportAuthoring Tool Support

Feature adoption gated by art pipelineFeature adoption gated by art pipeline Longer lead times for content creationLonger lead times for content creation Tool evolution rateTool evolution rate Duplicated effort for custom toolsDuplicated effort for custom tools

Every shop writes own export plug-insEvery shop writes own export plug-ins

We will provide source and samplesWe will provide source and samples To help reduce learning curveTo help reduce learning curve

Look under “extras” directory in SDKLook under “extras” directory in SDK

Authoring Tool Authoring Tool Plug-InsPlug-Ins

MeshesMeshes PatchesPatches Transform hierarchyTransform hierarchy Materials and TexturesMaterials and Textures SkinningSkinning AnimationAnimation

3D Studio Max 3.x, 4.03D Studio Max 3.x, 4.0 Support Character Studio 2.x, 3.0Support Character Studio 2.x, 3.0

Biped animations stored as sampled Biped animations stored as sampled matricesmatrices

Physique skinning exportedPhysique skinning exported Might need to re-apply physique to old data Might need to re-apply physique to old data

files like babyenv.maxfiles like babyenv.max

COM Skin support plannedCOM Skin support planned Patch exportPatch export

Work in progressWork in progress Cannot export patches created by surface Cannot export patches created by surface

modifiermodifier

A|W Maya 2.x, 3.xA|W Maya 2.x, 3.x

Support Rigid and Smooth skinningSupport Rigid and Smooth skinning Maya 2.x only has NURBSMaya 2.x only has NURBS

Can convert to patches using a scriptCan convert to patches using a script Use this before calling the exporterUse this before calling the exporter Supports export of skinned patchesSupports export of skinned patches

Maya 3.x has patchesMaya 3.x has patches Do not have export option yetDo not have export option yet

Call To ActionCall To Action

Try out new features in DirectX 8.1Try out new features in DirectX 8.1 Give us feedbackGive us feedback Tell us about bugs and performance Tell us about bugs and performance

issuesissues What else would you like to see?What else would you like to see? Hang around for the next talk…Hang around for the next talk…

AcknowledgementsAcknowledgements

Thanks to Origin Systems for Thanks to Origin Systems for permission to use Unicorn modelpermission to use Unicorn model

Thanks to NewTek for permission to use Thanks to NewTek for permission to use the monster modelthe monster model

Questions ?Questions ?