This section of the archives stores flipcode's complete Developer Toolbox collection, featuring a variety of mini-articles and source code contributions from our readers.

 

  Textured Lines In D3D
  Submitted by



Here's a way to solve a recurrent problem : emulating glLineWidth in Direct3D. Better : you can use this code to draw textured lines, which are often needed to create a lot of special effects (lightning bolts, toon-like silhouettes, etc). The code computes a world quad from a segment in world space. Simply render the quad instead of the segment! If you have a line-strip, compute one quad for each segment, then render everything as a single non-indexed tristrip of 4 * NbSegments vertices. You should be able to use it with any flexible vertex format (FVF). All matrices are 4x4 D3D-compliant matrices. mRenderWidth is the width of the render window (ex: 640) I hope it will be useful to some of you. I know I would have loved to have this right from the start (read: included in D3D...). By the way, it's not supposed to be optimized so don't bother telling me it's not :)

Pierre

Download Associated File: d3dtexturedlines.txt (4,586 bytes)

/*

Here's a way to solve a recurrent problem : emulating glLineWidth in Direct3D. Better : you can use this code to draw textured lines, which are often needed to create a lot of special effects (lightning bolts, toon-like silhouettes, etc). The code computes a world quad from a segment in world space. Simply render the quad instead of the segment! If you have a line-strip, compute one quad for each segment, then render everything as a single non-indexed tristrip of 4 * NbSegments vertices. You should be able to use it with any flexible vertex format (FVF). Here's the main routine :

*/
////////////////////////////////////////////////////////////////////////////////////////////////////////// /** * Computes a screen quad aligned with a world segment. * \param inverseview [in] inverse view matrix * \param view [in] view matrix * \param proj [in] projection matrix * \param verts [out] 4 quad vertices in world space (forming a tri-strip) * \param uvs [out] 4 quad uvs * \param stride [in] size of vertex (FVF stride) * \param p0 [in] segment's first point in world space * \param p1 [in] segment's second point in world space * \param size [in] size of segment/quad * \param constantsize [in] true to keep the quad's screen size constant (e.g. needed to emulate glLineWidth) */ ////////////////////////////////////////////////////////////////////////////////////////////////////////// void Renderer::ComputeScreenQuad(const ViewMatrix& inverseview, const ViewMatrix& view, const ProjMatrix& proj, ubyte* verts, ubyte* uvs, udword stride, const Point& p0, const Point& p1, float size, bool constantsize) { // Compute delta in camera space Point Delta; TransformPoint3x3(Delta, p1-p0, view); // Compute size factors float SizeP0 = size; float SizeP1 = size; if(constantsize) { // Compute scales so that screen-size is constant SizeP0 *= ComputeConstantScale(p0, view, proj); SizeP1 *= ComputeConstantScale(p1, view, proj); } // Compute quad vertices float Theta0 = atan2f(-Delta.x, -Delta.y); float c0 = SizeP0 * cosf(Theta0); float s0 = SizeP0 * sinf(Theta0); ComputePoint(*((Point*)verts), c0, -s0, inverseview, p0); verts+=stride; ComputePoint(*((Point*)verts), -c0, s0, inverseview, p0); verts+=stride; float Theta1 = atan2f(Delta.x, Delta.y); float c1 = SizeP1 * cosf(Theta1); float s1 = SizeP1 * sinf(Theta1); ComputePoint(*((Point*)verts), -c1, s1, inverseview, p1); verts+=stride; ComputePoint(*((Point*)verts), c1, -s1, inverseview, p1); verts+=stride; // Output uvs if needed if(uvs) { *((float*)uvs) = 0.0f; *((float*)(uvs+4)) = 1.0f; uvs+=stride; *((float*)uvs) = 0.0f; *((float*)(uvs+4)) = 0.0f; uvs+=stride; *((float*)uvs) = 1.0f; *((float*)(uvs+4)) = 1.0f; uvs+=stride; *((float*)uvs) = 1.0f; *((float*)(uvs+4)) = 0.0f; uvs+=stride; } }

inline void ComputePoint(Point& dest, float x, float y, const Matrix4x4& rot, const Point& trans) { dest.x = trans.x + x * rot.m[0][0] + y * rot.m[1][0]; dest.y = trans.y + x * rot.m[0][1] + y * rot.m[1][1]; dest.z = trans.z + x * rot.m[0][2] + y * rot.m[1][2]; }

And extra used functions: // Quickly rotates a vector, using the 3x3 part of a 4x4 matrix inline void TransformPoint3x3(Point& dest, const Point& source, const Matrix4x4& rot) { dest.x = source.x * rot.m[0][0] + source.y * rot.m[1][0] + source.z * rot.m[2][0]; dest.y = source.x * rot.m[0][1] + source.y * rot.m[1][1] + source.z * rot.m[2][1]; dest.z = source.x * rot.m[0][2] + source.y * rot.m[1][2] + source.z * rot.m[2][2]; }

float Renderer::ComputeConstantScale(const Point& pos, const ViewMatrix& view, const ProjMatrix& proj) { Point ppcam0 = pos * view; Point ppcam1 = ppcam0; ppcam1.x += 1.0f; float l1 = 1.0f/(ppcam0.x*proj.m[0][3] + ppcam0.y*proj.m[1][3] + ppcam0.z*proj.m[2][3] + proj.m[3][3]); float c1 = (ppcam0.x*proj.m[0][0] + ppcam0.y*proj.m[1][0] + ppcam0.z*proj.m[2][0] + proj.m[3][0])*l1; float l2 = 1.0f/(ppcam1.x*proj.m[0][3] + ppcam1.y*proj.m[1][3] + ppcam1.z*proj.m[2][3] + proj.m[3][3]); float c2 = (ppcam1.x*proj.m[0][0] + ppcam1.y*proj.m[1][0] + ppcam1.z*proj.m[2][0] + proj.m[3][0])*l2; float CorrectScale = 1.0f/(c2 - c1); return CorrectScale / float(mRenderWidth); } All matrices are 4x4 D3D-compliant matrices. mRenderWidth is the width of the render window (ex: 640) I hope it will be useful to some of you. I know I would have loved to have this right from the start (read: included in D3D...). By the way, it's not supposed to be optimized so don't bother telling me it's not :) Pierre

The zip file viewer built into the Developer Toolbox made use of the zlib library, as well as the zlibdll source additions.

 

Copyright 1999-2008 (C) FLIPCODE.COM and/or the original content author(s). All rights reserved.
Please read our Terms, Conditions, and Privacy information.