Major rework of tnl module

New array_cache module
Support 8 texture units in core mesa (now support 8 everywhere)
Rework core mesa statechange operations to avoid flushing on many
noop statechanges.

12 files changed, 821 insertions, 204 deletions

diff --git a/src/mesa/math/m_clip_tmp.h b/src/mesa/math/m_clip_tmp.h
index 321d3a9..d977698 100644
--- a/src/mesa/math/m_clip_tmp.h
+++ b/src/mesa/math/m_clip_tmp.h
@@ -1,4 +1,4 @@
-/* $Id: m_clip_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+/* $Id: m_clip_tmp.h,v 1.2 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -78,16 +78,16 @@ static GLvector4f * _XFORMAPI TAG(cliptest_points4)( GLvector4f *clip_vec,
 	 c++;
 	 tmpAndMask &= mask;
 	 tmpOrMask |= mask;
-	 vProj[i][0] = 0;	/* no longer required? */
+	 vProj[i][0] = 0;
 	 vProj[i][1] = 0;
 	 vProj[i][2] = 0;
 	 vProj[i][3] = 1;
       } else {
 	 GLfloat oow = 1.0F / cw;	 
-	 vProj[i][3] = oow;
 	 vProj[i][0] = cx * oow;
 	 vProj[i][1] = cy * oow;
 	 vProj[i][2] = cz * oow;      
+	 vProj[i][3] = oow;
       }	 
    }
 
@@ -95,7 +95,7 @@ static GLvector4f * _XFORMAPI TAG(cliptest_points4)( GLvector4f *clip_vec,
    *andMask = (GLubyte) (c < count ? 0 : tmpAndMask);
 
    proj_vec->flags |= VEC_SIZE_4;
-   proj_vec->size = 3;
+   proj_vec->size = 4;
    proj_vec->count = clip_vec->count;
    return proj_vec;
 }
@@ -127,8 +127,6 @@ static GLvector4f * _XFORMAPI TAG(cliptest_points3)( GLvector4f *clip_vec,
       tmpAndMask &= mask;
    }
 
-   gl_vector4f_clean_elem(proj_vec, count, 3);
-      
    *orMask = tmpOrMask;
    *andMask = tmpAndMask;
    return clip_vec;
@@ -159,8 +157,6 @@ static GLvector4f * _XFORMAPI TAG(cliptest_points2)( GLvector4f *clip_vec,
       tmpAndMask &= mask;
    }
 
-   gl_vector4f_clean_elem(proj_vec, count, 3);
-
    *orMask = tmpOrMask;
    *andMask = tmpAndMask;
    return clip_vec;
diff --git a/src/mesa/math/m_copy_tmp.h b/src/mesa/math/m_copy_tmp.h
index b328537..3119b95 100644
--- a/src/mesa/math/m_copy_tmp.h
+++ b/src/mesa/math/m_copy_tmp.h
@@ -1,4 +1,4 @@
-/* $Id: m_copy_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+/* $Id: m_copy_tmp.h,v 1.2 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -51,24 +51,6 @@ static void TAG2(copy, BITS)(GLvector4f *to, const GLvector4f *f,	\
       }									\
 }
 
-
-
-/* static void TAG2(clean, BITS)(GLvector4f *to ) */
-/* { */
-/*    GLfloat (*t)[4] = to->data; */
-/*    GLuint i; */
-
-/*    if (BITS) */
-/*       for (i = 0 ; i < VB_SIZE ; i++) { */
-/*          if (BITS&1) t[i][0] = 0; */
-/* 	 if (BITS&2) t[i][1] = 0; */
-/* 	 if (BITS&4) t[i][2] = 0; */
-/* 	 if (BITS&8) t[i][3] = 1; */
-/*       } */
-/*    to->flags &= ~BITS; */
-/* } */
-
-
 /* We got them all here:
  */
 COPY_FUNC( 0x0 )		/* noop */
@@ -106,21 +88,4 @@ static void TAG2(init_copy, 0 ) ( void )
    gl_copy_tab[IDX][0xd] = TAG2(copy, 0xd);
    gl_copy_tab[IDX][0xe] = TAG2(copy, 0xe);
    gl_copy_tab[IDX][0xf] = TAG2(copy, 0xf);
-
-/*    gl_clean_tab[IDX][0x0] = TAG2(clean, 0x0); */
-/*    gl_clean_tab[IDX][0x1] = TAG2(clean, 0x1); */
-/*    gl_clean_tab[IDX][0x2] = TAG2(clean, 0x2); */
-/*    gl_clean_tab[IDX][0x3] = TAG2(clean, 0x3); */
-/*    gl_clean_tab[IDX][0x4] = TAG2(clean, 0x4); */
-/*    gl_clean_tab[IDX][0x5] = TAG2(clean, 0x5); */
-/*    gl_clean_tab[IDX][0x6] = TAG2(clean, 0x6); */
-/*    gl_clean_tab[IDX][0x7] = TAG2(clean, 0x7); */
-/*    gl_clean_tab[IDX][0x8] = TAG2(clean, 0x8); */
-/*    gl_clean_tab[IDX][0x9] = TAG2(clean, 0x9); */
-/*    gl_clean_tab[IDX][0xa] = TAG2(clean, 0xa); */
-/*    gl_clean_tab[IDX][0xb] = TAG2(clean, 0xb); */
-/*    gl_clean_tab[IDX][0xc] = TAG2(clean, 0xc); */
-/*    gl_clean_tab[IDX][0xd] = TAG2(clean, 0xd); */
-/*    gl_clean_tab[IDX][0xe] = TAG2(clean, 0xe); */
-/*    gl_clean_tab[IDX][0xf] = TAG2(clean, 0xf); */
 }
diff --git a/src/mesa/math/m_dotprod_tmp.h b/src/mesa/math/m_dotprod_tmp.h
index 637e35f..4bd875c 100644
--- a/src/mesa/math/m_dotprod_tmp.h
+++ b/src/mesa/math/m_dotprod_tmp.h
@@ -1,4 +1,4 @@
-/* $Id: m_dotprod_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+/* $Id: m_dotprod_tmp.h,v 1.2 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -31,18 +31,16 @@
 
 /* Note - respects the stride of the output vector.
  */
-static void TAG(dotprod_vec2)( GLvector4f *out_vec, 
-			     GLuint elt,
-			     const GLvector4f *coord_vec, 
-			     const GLfloat plane[4], 
-			     const GLubyte mask[])
+static void TAG(dotprod_vec2)( GLfloat *out, 
+			       GLuint outstride,
+			       const GLvector4f *coord_vec, 
+			       const GLfloat plane[4], 
+			       const GLubyte mask[])
 {
    GLuint stride = coord_vec->stride;
    GLfloat *coord = coord_vec->start;
    GLuint count = coord_vec->count;
 
-   GLuint outstride = out_vec->stride;
-   GLfloat *out = out_vec->start + elt;
    GLuint i;
    
    const GLfloat plane0 = plane[0], plane1 = plane[1], plane3 = plane[3];
@@ -56,21 +54,18 @@ static void TAG(dotprod_vec2)( GLvector4f *out_vec,
 		            plane3);
       }
    }   
-   out_vec->count = coord_vec->count;
 }
 
-static void TAG(dotprod_vec3)( GLvector4f *out_vec, 
-			     GLuint elt,
-			     const GLvector4f *coord_vec, 
-			     const GLfloat plane[4], 
-			     const GLubyte mask[])
+static void TAG(dotprod_vec3)( GLfloat *out, 
+			       GLuint outstride,
+			       const GLvector4f *coord_vec, 
+			       const GLfloat plane[4], 
+			       const GLubyte mask[])
 {
    GLuint stride = coord_vec->stride;
    GLfloat *coord = coord_vec->start;
    GLuint count = coord_vec->count;
 
-   GLuint outstride = out_vec->stride;
-   GLfloat *out = out_vec->start + elt;
    GLuint i;
 
    const GLfloat plane0 = plane[0], plane1 = plane[1], plane2 = plane[2];
@@ -86,21 +81,17 @@ static void TAG(dotprod_vec3)( GLvector4f *out_vec,
 		            plane3);
       }
    }   
-   out_vec->count = coord_vec->count;
 }
 
-static void TAG(dotprod_vec4)( GLvector4f *out_vec, 
-			     GLuint elt,
-			     const GLvector4f *coord_vec, 
-			     const GLfloat plane[4], 
-			     const GLubyte mask[])
+static void TAG(dotprod_vec4)( GLfloat *out, 
+			       GLuint outstride,
+			       const GLvector4f *coord_vec, 
+			       const GLfloat plane[4], 
+			       const GLubyte mask[])
 {
    GLuint stride = coord_vec->stride;
    GLfloat *coord = coord_vec->start;
    GLuint count = coord_vec->count;
-
-   GLuint outstride = out_vec->stride;
-   GLfloat *out = out_vec->start + elt;
    GLuint i;
 
    const GLfloat plane0 = plane[0], plane1 = plane[1], plane2 = plane[2];
@@ -116,7 +107,6 @@ static void TAG(dotprod_vec4)( GLvector4f *out_vec,
 		 coord[3] * plane3);
       }
    }   
-   out_vec->count = coord_vec->count;
 }
 
 
diff --git a/src/mesa/math/m_eval.c b/src/mesa/math/m_eval.c
new file mode 100644
index 0000000..a4ae039
--- /dev/null
+++ b/src/mesa/math/m_eval.c
@@ -0,0 +1,501 @@
+/* $Id: m_eval.c,v 1.1 2000/12/26 05:09:31 keithw Exp $ */
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  3.5
+ *
+ * Copyright (C) 1999-2000  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included
+ * in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/*
+ * eval.c was written by
+ * Bernd Barsuhn (bdbarsuh@cip.informatik.uni-erlangen.de) and
+ * Volker Weiss (vrweiss@cip.informatik.uni-erlangen.de).
+ *
+ * My original implementation of evaluators was simplistic and didn't
+ * compute surface normal vectors properly.  Bernd and Volker applied
+ * used more sophisticated methods to get better results.
+ *
+ * Thanks guys!
+ */
+
+
+#include "glheader.h"
+#include "config.h"
+#include "m_eval.h"
+
+static GLfloat inv_tab[MAX_EVAL_ORDER];
+
+
+
+/*
+ * Horner scheme for Bezier curves
+ *
+ * Bezier curves can be computed via a Horner scheme.
+ * Horner is numerically less stable than the de Casteljau
+ * algorithm, but it is faster. For curves of degree n
+ * the complexity of Horner is O(n) and de Casteljau is O(n^2).
+ * Since stability is not important for displaying curve
+ * points I decided to use the Horner scheme.
+ *
+ * A cubic Bezier curve with control points b0, b1, b2, b3 can be
+ * written as
+ *
+ *        (([3]        [3]     )     [3]       )     [3]
+ * c(t) = (([0]*s*b0 + [1]*t*b1)*s + [2]*t^2*b2)*s + [3]*t^2*b3
+ *
+ *                                           [n]
+ * where s=1-t and the binomial coefficients [i]. These can
+ * be computed iteratively using the identity:
+ *
+ * [n]               [n  ]             [n]
+ * [i] = (n-i+1)/i * [i-1]     and     [0] = 1
+ */
+
+
+void
+_math_horner_bezier_curve(const GLfloat *cp, GLfloat *out, GLfloat t,
+			  GLuint dim, GLuint order)
+{
+   GLfloat s, powert;
+   GLuint i, k, bincoeff;
+
+   if(order >= 2)
+   {
+      bincoeff = order-1;
+      s = 1.0-t;
+
+      for(k=0; k<dim; k++)
+	 out[k] = s*cp[k] + bincoeff*t*cp[dim+k];
+
+      for(i=2, cp+=2*dim, powert=t*t; i<order; i++, powert*=t, cp +=dim)
+      {
+	 bincoeff *= order-i;
+	 bincoeff *= inv_tab[i];
+
+	 for(k=0; k<dim; k++)
+	    out[k] = s*out[k] + bincoeff*powert*cp[k];
+      }
+   }
+   else /* order=1 -> constant curve */
+   {
+      for(k=0; k<dim; k++)
+	 out[k] = cp[k];
+   }
+}
+
+/*
+ * Tensor product Bezier surfaces
+ *
+ * Again the Horner scheme is used to compute a point on a
+ * TP Bezier surface. First a control polygon for a curve
+ * on the surface in one parameter direction is computed,
+ * then the point on the curve for the other parameter
+ * direction is evaluated.
+ *
+ * To store the curve control polygon additional storage
+ * for max(uorder,vorder) points is needed in the
+ * control net cn.
+ */
+
+void
+_math_horner_bezier_surf(GLfloat *cn, GLfloat *out, GLfloat u, GLfloat v,
+			 GLuint dim, GLuint uorder, GLuint vorder)
+{
+   GLfloat *cp = cn + uorder*vorder*dim;
+   GLuint i, uinc = vorder*dim;
+
+   if(vorder > uorder)
+   {
+      if(uorder >= 2)
+      {
+	 GLfloat s, poweru;
+	 GLuint j, k, bincoeff;
+
+	 /* Compute the control polygon for the surface-curve in u-direction */
+	 for(j=0; j<vorder; j++)
+	 {
+	    GLfloat *ucp = &cn[j*dim];
+
+	    /* Each control point is the point for parameter u on a */
+	    /* curve defined by the control polygons in u-direction */
+	    bincoeff = uorder-1;
+	    s = 1.0-u;
+
+	    for(k=0; k<dim; k++)
+	       cp[j*dim+k] = s*ucp[k] + bincoeff*u*ucp[uinc+k];
+
+	    for(i=2, ucp+=2*uinc, poweru=u*u; i<uorder;
+		i++, poweru*=u, ucp +=uinc)
+	    {
+	       bincoeff *= uorder-i;
+	       bincoeff *= inv_tab[i];
+
+	       for(k=0; k<dim; k++)
+		  cp[j*dim+k] = s*cp[j*dim+k] + bincoeff*poweru*ucp[k];
+	    }
+	 }
+
+	 /* Evaluate curve point in v */
+	 _math_horner_bezier_curve(cp, out, v, dim, vorder);
+      }
+      else /* uorder=1 -> cn defines a curve in v */
+	 _math_horner_bezier_curve(cn, out, v, dim, vorder);
+   }
+   else /* vorder <= uorder */
+   {
+      if(vorder > 1)
+      {
+	 GLuint i;
+
+	 /* Compute the control polygon for the surface-curve in u-direction */
+	 for(i=0; i<uorder; i++, cn += uinc)
+	 {
+	    /* For constant i all cn[i][j] (j=0..vorder) are located */
+	    /* on consecutive memory locations, so we can use        */
+	    /* horner_bezier_curve to compute the control points     */
+
+	    _math_horner_bezier_curve(cn, &cp[i*dim], v, dim, vorder);
+	 }
+
+	 /* Evaluate curve point in u */
+	 _math_horner_bezier_curve(cp, out, u, dim, uorder);
+      }
+      else  /* vorder=1 -> cn defines a curve in u */
+	 _math_horner_bezier_curve(cn, out, u, dim, uorder);
+   }
+}
+
+/*
+ * The direct de Casteljau algorithm is used when a point on the
+ * surface and the tangent directions spanning the tangent plane
+ * should be computed (this is needed to compute normals to the
+ * surface). In this case the de Casteljau algorithm approach is
+ * nicer because a point and the partial derivatives can be computed
+ * at the same time. To get the correct tangent length du and dv
+ * must be multiplied with the (u2-u1)/uorder-1 and (v2-v1)/vorder-1.
+ * Since only the directions are needed, this scaling step is omitted.
+ *
+ * De Casteljau needs additional storage for uorder*vorder
+ * values in the control net cn.
+ */
+
+void
+_math_de_casteljau_surf(GLfloat *cn, GLfloat *out, GLfloat *du, GLfloat *dv,
+			GLfloat u, GLfloat v, GLuint dim,
+			GLuint uorder, GLuint vorder)
+{
+   GLfloat *dcn = cn + uorder*vorder*dim;
+   GLfloat us = 1.0-u, vs = 1.0-v;
+   GLuint h, i, j, k;
+   GLuint minorder = uorder < vorder ? uorder : vorder;
+   GLuint uinc = vorder*dim;
+   GLuint dcuinc = vorder;
+
+   /* Each component is evaluated separately to save buffer space  */
+   /* This does not drasticaly decrease the performance of the     */
+   /* algorithm. If additional storage for (uorder-1)*(vorder-1)   */
+   /* points would be available, the components could be accessed  */
+   /* in the innermost loop which could lead to less cache misses. */
+
+#define CN(I,J,K) cn[(I)*uinc+(J)*dim+(K)]
+#define DCN(I, J) dcn[(I)*dcuinc+(J)]
+   if(minorder < 3)
+   {
+      if(uorder==vorder)
+      {
+	 for(k=0; k<dim; k++)
+	 {
+	    /* Derivative direction in u */
+	    du[k] = vs*(CN(1,0,k) - CN(0,0,k)) +
+	       v*(CN(1,1,k) - CN(0,1,k));
+
+	    /* Derivative direction in v */
+	    dv[k] = us*(CN(0,1,k) - CN(0,0,k)) +
+	       u*(CN(1,1,k) - CN(1,0,k));
+
+	    /* bilinear de Casteljau step */
+	    out[k] =  us*(vs*CN(0,0,k) + v*CN(0,1,k)) +
+	       u*(vs*CN(1,0,k) + v*CN(1,1,k));
+	 }
+      }
+      else if(minorder == uorder)
+      {
+	 for(k=0; k<dim; k++)
+	 {
+	    /* bilinear de Casteljau step */
+	    DCN(1,0) =    CN(1,0,k) -   CN(0,0,k);
+	    DCN(0,0) = us*CN(0,0,k) + u*CN(1,0,k);
+
+	    for(j=0; j<vorder-1; j++)
+	    {
+	       /* for the derivative in u */
+	       DCN(1,j+1) =    CN(1,j+1,k) -   CN(0,j+1,k);
+	       DCN(1,j)   = vs*DCN(1,j)    + v*DCN(1,j+1);
+
+	       /* for the `point' */
+	       DCN(0,j+1) = us*CN(0,j+1,k) + u*CN(1,j+1,k);
+	       DCN(0,j)   = vs*DCN(0,j)    + v*DCN(0,j+1);
+	    }
+
+	    /* remaining linear de Casteljau steps until the second last step */
+	    for(h=minorder; h<vorder-1; h++)
+	       for(j=0; j<vorder-h; j++)
+	       {
+		  /* for the derivative in u */
+		  DCN(1,j) = vs*DCN(1,j) + v*DCN(1,j+1);
+
+		  /* for the `point' */
+		  DCN(0,j) = vs*DCN(0,j) + v*DCN(0,j+1);
+	       }
+
+	    /* derivative direction in v */
+	    dv[k] = DCN(0,1) - DCN(0,0);
+
+	    /* derivative direction in u */
+	    du[k] =   vs*DCN(1,0) + v*DCN(1,1);
+
+	    /* last linear de Casteljau step */
+	    out[k] =  vs*DCN(0,0) + v*DCN(0,1);
+	 }
+      }
+      else /* minorder == vorder */
+      {
+	 for(k=0; k<dim; k++)
+	 {
+	    /* bilinear de Casteljau step */
+	    DCN(0,1) =    CN(0,1,k) -   CN(0,0,k);
+	    DCN(0,0) = vs*CN(0,0,k) + v*CN(0,1,k);
+	    for(i=0; i<uorder-1; i++)
+	    {
+	       /* for the derivative in v */
+	       DCN(i+1,1) =    CN(i+1,1,k) -   CN(i+1,0,k);
+	       DCN(i,1)   = us*DCN(i,1)    + u*DCN(i+1,1);
+
+	       /* for the `point' */
+	       DCN(i+1,0) = vs*CN(i+1,0,k) + v*CN(i+1,1,k);
+	       DCN(i,0)   = us*DCN(i,0)    + u*DCN(i+1,0);
+	    }
+
+	    /* remaining linear de Casteljau steps until the second last step */
+	    for(h=minorder; h<uorder-1; h++)
+	       for(i=0; i<uorder-h; i++)
+	       {
+		  /* for the derivative in v */
+		  DCN(i,1) = us*DCN(i,1) + u*DCN(i+1,1);
+
+		  /* for the `point' */
+		  DCN(i,0) = us*DCN(i,0) + u*DCN(i+1,0);
+	       }
+
+	    /* derivative direction in u */
+	    du[k] = DCN(1,0) - DCN(0,0);
+
+	    /* derivative direction in v */
+	    dv[k] =   us*DCN(0,1) + u*DCN(1,1);
+
+	    /* last linear de Casteljau step */
+	    out[k] =  us*DCN(0,0) + u*DCN(1,0);
+	 }
+      }
+   }
+   else if(uorder == vorder)
+   {
+      for(k=0; k<dim; k++)
+      {
+	 /* first bilinear de Casteljau step */
+	 for(i=0; i<uorder-1; i++)
+	 {
+	    DCN(i,0) = us*CN(i,0,k) + u*CN(i+1,0,k);
+	    for(j=0; j<vorder-1; j++)
+	    {
+	       DCN(i,j+1) = us*CN(i,j+1,k) + u*CN(i+1,j+1,k);
+	       DCN(i,j)   = vs*DCN(i,j)    + v*DCN(i,j+1);
+	    }
+	 }
+
+	 /* remaining bilinear de Casteljau steps until the second last step */
+	 for(h=2; h<minorder-1; h++)
+	    for(i=0; i<uorder-h; i++)
+	    {
+	       DCN(i,0) = us*DCN(i,0) + u*DCN(i+1,0);
+	       for(j=0; j<vorder-h; j++)
+	       {
+		  DCN(i,j+1) = us*DCN(i,j+1) + u*DCN(i+1,j+1);
+		  DCN(i,j)   = vs*DCN(i,j)   + v*DCN(i,j+1);
+	       }
+	    }
+
+	 /* derivative direction in u */
+	 du[k] = vs*(DCN(1,0) - DCN(0,0)) +
+	    v*(DCN(1,1) - DCN(0,1));
+
+	 /* derivative direction in v */
+	 dv[k] = us*(DCN(0,1) - DCN(0,0)) +
+	    u*(DCN(1,1) - DCN(1,0));
+
+	 /* last bilinear de Casteljau step */
+	 out[k] =  us*(vs*DCN(0,0) + v*DCN(0,1)) +
+	    u*(vs*DCN(1,0) + v*DCN(1,1));
+      }
+   }
+   else if(minorder == uorder)
+   {
+      for(k=0; k<dim; k++)
+      {
+	 /* first bilinear de Casteljau step */
+	 for(i=0; i<uorder-1; i++)
+	 {
+	    DCN(i,0) = us*CN(i,0,k) + u*CN(i+1,0,k);
+	    for(j=0; j<vorder-1; j++)
+	    {
+	       DCN(i,j+1) = us*CN(i,j+1,k) + u*CN(i+1,j+1,k);
+	       DCN(i,j)   = vs*DCN(i,j)    + v*DCN(i,j+1);
+	    }
+	 }
+
+	 /* remaining bilinear de Casteljau steps until the second last step */
+	 for(h=2; h<minorder-1; h++)
+	    for(i=0; i<uorder-h; i++)
+	    {
+	       DCN(i,0) = us*DCN(i,0) + u*DCN(i+1,0);
+	       for(j=0; j<vorder-h; j++)
+	       {
+		  DCN(i,j+1) = us*DCN(i,j+1) + u*DCN(i+1,j+1);
+		  DCN(i,j)   = vs*DCN(i,j)   + v*DCN(i,j+1);
+	       }
+	    }
+
+	 /* last bilinear de Casteljau step */
+	 DCN(2,0) =    DCN(1,0) -   DCN(0,0);
+	 DCN(0,0) = us*DCN(0,0) + u*DCN(1,0);
+	 for(j=0; j<vorder-1; j++)
+	 {
+	    /* for the derivative in u */
+	    DCN(2,j+1) =    DCN(1,j+1) -    DCN(0,j+1);
+	    DCN(2,j)   = vs*DCN(2,j)    + v*DCN(2,j+1);
+	
+	    /* for the `point' */
+	    DCN(0,j+1) = us*DCN(0,j+1 ) + u*DCN(1,j+1);
+	    DCN(0,j)   = vs*DCN(0,j)    + v*DCN(0,j+1);
+	 }
+
+	 /* remaining linear de Casteljau steps until the second last step */
+	 for(h=minorder; h<vorder-1; h++)
+	    for(j=0; j<vorder-h; j++)
+	    {
+	       /* for the derivative in u */
+	       DCN(2,j) = vs*DCN(2,j) + v*DCN(2,j+1);
+	
+	       /* for the `point' */
+	       DCN(0,j) = vs*DCN(0,j) + v*DCN(0,j+1);
+	    }
+
+	 /* derivative direction in v */
+	 dv[k] = DCN(0,1) - DCN(0,0);
+
+	 /* derivative direction in u */
+	 du[k] =   vs*DCN(2,0) + v*DCN(2,1);
+
+	 /* last linear de Casteljau step */
+	 out[k] =  vs*DCN(0,0) + v*DCN(0,1);
+      }
+   }
+   else /* minorder == vorder */
+   {
+      for(k=0; k<dim; k++)
+      {
+	 /* first bilinear de Casteljau step */
+	 for(i=0; i<uorder-1; i++)
+	 {
+	    DCN(i,0) = us*CN(i,0,k) + u*CN(i+1,0,k);
+	    for(j=0; j<vorder-1; j++)
+	    {
+	       DCN(i,j+1) = us*CN(i,j+1,k) + u*CN(i+1,j+1,k);
+	       DCN(i,j)   = vs*DCN(i,j)    + v*DCN(i,j+1);
+	    }
+	 }
+
+	 /* remaining bilinear de Casteljau steps until the second last step */
+	 for(h=2; h<minorder-1; h++)
+	    for(i=0; i<uorder-h; i++)
+	    {
+	       DCN(i,0) = us*DCN(i,0) + u*DCN(i+1,0);
+	       for(j=0; j<vorder-h; j++)
+	       {
+		  DCN(i,j+1) = us*DCN(i,j+1) + u*DCN(i+1,j+1);
+		  DCN(i,j)   = vs*DCN(i,j)   + v*DCN(i,j+1);
+	       }
+	    }
+
+	 /* last bilinear de Casteljau step */
+	 DCN(0,2) =    DCN(0,1) -   DCN(0,0);
+	 DCN(0,0) = vs*DCN(0,0) + v*DCN(0,1);
+	 for(i=0; i<uorder-1; i++)
+	 {
+	    /* for the derivative in v */
+	    DCN(i+1,2) =    DCN(i+1,1)  -   DCN(i+1,0);
+	    DCN(i,2)   = us*DCN(i,2)    + u*DCN(i+1,2);
+	
+	    /* for the `point' */
+	    DCN(i+1,0) = vs*DCN(i+1,0)  + v*DCN(i+1,1);
+	    DCN(i,0)   = us*DCN(i,0)    + u*DCN(i+1,0);
+	 }
+
+	 /* remaining linear de Casteljau steps until the second last step */
+	 for(h=minorder; h<uorder-1; h++)
+	    for(i=0; i<uorder-h; i++)
+	    {
+	       /* for the derivative in v */
+	       DCN(i,2) = us*DCN(i,2) + u*DCN(i+1,2);
+	
+	       /* for the `point' */
+	       DCN(i,0) = us*DCN(i,0) + u*DCN(i+1,0);
+	    }
+
+	 /* derivative direction in u */
+	 du[k] = DCN(1,0) - DCN(0,0);
+
+	 /* derivative direction in v */
+	 dv[k] =   us*DCN(0,2) + u*DCN(1,2);
+
+	 /* last linear de Casteljau step */
+	 out[k] =  us*DCN(0,0) + u*DCN(1,0);
+      }
+   }
+#undef DCN
+#undef CN
+}
+
+
+/*
+ * Do one-time initialization for evaluators.
+ */
+void _math_init_eval( void )
+{
+   GLuint i;
+
+   /* KW: precompute 1/x for useful x.
+    */
+   for (i = 1 ; i < MAX_EVAL_ORDER ; i++)
+      inv_tab[i] = 1.0 / i;
+}
+
diff --git a/src/mesa/math/m_eval.h b/src/mesa/math/m_eval.h
new file mode 100644
index 0000000..b478b39
--- /dev/null
+++ b/src/mesa/math/m_eval.h
@@ -0,0 +1,79 @@
+
+#ifndef _M_EVAL_H
+#define _M_EVAL_H
+
+#include "glheader.h"
+
+void _math_init_eval( void );
+
+
+/*
+ * Horner scheme for Bezier curves
+ *
+ * Bezier curves can be computed via a Horner scheme.
+ * Horner is numerically less stable than the de Casteljau
+ * algorithm, but it is faster. For curves of degree n
+ * the complexity of Horner is O(n) and de Casteljau is O(n^2).
+ * Since stability is not important for displaying curve
+ * points I decided to use the Horner scheme.
+ *
+ * A cubic Bezier curve with control points b0, b1, b2, b3 can be
+ * written as
+ *
+ *        (([3]        [3]     )     [3]       )     [3]
+ * c(t) = (([0]*s*b0 + [1]*t*b1)*s + [2]*t^2*b2)*s + [3]*t^2*b3
+ *
+ *                                           [n]
+ * where s=1-t and the binomial coefficients [i]. These can
+ * be computed iteratively using the identity:
+ *
+ * [n]               [n  ]             [n]
+ * [i] = (n-i+1)/i * [i-1]     and     [0] = 1
+ */
+
+
+void
+_math_horner_bezier_curve(const GLfloat *cp, GLfloat *out, GLfloat t,
+			  GLuint dim, GLuint order);
+
+
+/*
+ * Tensor product Bezier surfaces
+ *
+ * Again the Horner scheme is used to compute a point on a
+ * TP Bezier surface. First a control polygon for a curve
+ * on the surface in one parameter direction is computed,
+ * then the point on the curve for the other parameter
+ * direction is evaluated.
+ *
+ * To store the curve control polygon additional storage
+ * for max(uorder,vorder) points is needed in the
+ * control net cn.
+ */
+
+void
+_math_horner_bezier_surf(GLfloat *cn, GLfloat *out, GLfloat u, GLfloat v,
+			 GLuint dim, GLuint uorder, GLuint vorder);
+
+
+/*
+ * The direct de Casteljau algorithm is used when a point on the
+ * surface and the tangent directions spanning the tangent plane
+ * should be computed (this is needed to compute normals to the
+ * surface). In this case the de Casteljau algorithm approach is
+ * nicer because a point and the partial derivatives can be computed
+ * at the same time. To get the correct tangent length du and dv
+ * must be multiplied with the (u2-u1)/uorder-1 and (v2-v1)/vorder-1.
+ * Since only the directions are needed, this scaling step is omitted.
+ *
+ * De Casteljau needs additional storage for uorder*vorder
+ * values in the control net cn.
+ */
+
+void
+_math_de_casteljau_surf(GLfloat *cn, GLfloat *out, GLfloat *du, GLfloat *dv,
+			GLfloat u, GLfloat v, GLuint dim,
+			GLuint uorder, GLuint vorder);
+
+
+#endif
diff --git a/src/mesa/math/m_translate.c b/src/mesa/math/m_translate.c
index 945e357..1a0804a 100644
--- a/src/mesa/math/m_translate.c
+++ b/src/mesa/math/m_translate.c
@@ -1,4 +1,4 @@
-/* $Id: m_translate.c,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+/* $Id: m_translate.c,v 1.2 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -36,22 +36,67 @@
 
 #include "m_translate.h"
 
+
+
+typedef void (*trans_1f_func)(GLfloat *to,
+			      CONST void *ptr,
+			      GLuint stride,
+			      GLuint start, 
+			      GLuint n );
+
+typedef void (*trans_1ui_func)(GLuint *to,
+			       CONST void *ptr,
+			       GLuint stride,
+			       GLuint start, 
+			       GLuint n );
+
+typedef void (*trans_1ub_func)(GLubyte *to,
+			       CONST void *ptr,
+			       GLuint stride,
+			       GLuint start,
+			       GLuint n );
+
+typedef void (*trans_4ub_func)(GLubyte (*to)[4],
+			       CONST void *ptr,
+			       GLuint stride,
+			       GLuint start,
+			       GLuint n );
+
+typedef void (*trans_4f_func)(GLfloat (*to)[4],
+			      CONST void *ptr,
+			      GLuint stride,
+			      GLuint start, 
+			      GLuint n );
+
+typedef void (*trans_3f_func)(GLfloat (*to)[3],
+			      CONST void *ptr,
+			      GLuint stride,
+			      GLuint start, 
+			      GLuint n );
+
+
+
+
+#define TYPE_IDX(t) ((t) & 0xf)
+#define MAX_TYPES TYPE_IDX(GL_DOUBLE)+1      /* 0xa + 1 */
+
+
 /* This macro is used on other systems, so undefine it for this module */
 
 #undef	CHECK
 
-trans_1f_func gl_trans_1f_tab[MAX_TYPES];
-trans_1ui_func gl_trans_1ui_tab[MAX_TYPES];
-trans_1ub_func gl_trans_1ub_tab[MAX_TYPES];
-trans_3f_func  gl_trans_3f_tab[MAX_TYPES];
-trans_4ub_func gl_trans_4ub_tab[5][MAX_TYPES];
-trans_4f_func  gl_trans_4f_tab[5][MAX_TYPES];
+static trans_1f_func  _math_trans_1f_tab[MAX_TYPES];
+static trans_1ui_func _math_trans_1ui_tab[MAX_TYPES];
+static trans_1ub_func _math_trans_1ub_tab[MAX_TYPES];
+static trans_3f_func  _math_trans_3f_tab[MAX_TYPES];
+static trans_4ub_func _math_trans_4ub_tab[5][MAX_TYPES];
+static trans_4f_func  _math_trans_4f_tab[5][MAX_TYPES];
 
 
 #define PTR_ELT(ptr, elt) (((SRC *)ptr)[elt])
 
 
-#define TAB(x) gl_trans##x##_tab
+#define TAB(x) _math_trans##x##_tab
 #define ARGS   GLuint start, GLuint n
 #define SRC_START  start
 #define DST_START  0
@@ -471,8 +516,73 @@ static void init_translate_raw(void)
 
 
 
-void 
-_math_init_translate( void )
+void _math_init_translate( void )
 {
    init_translate_raw();
 }
+
+
+
+void _math_trans_1f(GLfloat *to,
+		    CONST void *ptr,
+		    GLuint stride,
+		    GLenum type,
+		    GLuint start,
+		    GLuint n )
+{
+   _math_trans_1f_tab[TYPE_IDX(type)]( to, ptr, stride, start, n );
+}
+
+void _math_trans_1ui(GLuint *to,
+		     CONST void *ptr,
+		     GLuint stride,
+		     GLenum type,
+		     GLuint start,
+		     GLuint n )
+{
+   _math_trans_1ui_tab[TYPE_IDX(type)]( to, ptr, stride, start, n ); 
+}
+
+void _math_trans_1ub(GLubyte *to,
+		     CONST void *ptr,
+		     GLuint stride,
+		     GLenum type,
+		     GLuint start,
+		     GLuint n )
+{
+   _math_trans_1ub_tab[TYPE_IDX(type)]( to, ptr, stride, start, n ); 
+}
+
+
+void _math_trans_4ub(GLubyte (*to)[4],
+		     CONST void *ptr,
+		     GLuint stride,
+		     GLenum type,
+		     GLuint size,
+		     GLuint start,
+		     GLuint n )
+{
+   _math_trans_4ub_tab[size][TYPE_IDX(type)]( to, ptr, stride, start, n ); 
+}
+
+void _math_trans_4f(GLfloat (*to)[4],
+		    CONST void *ptr,
+		    GLuint stride,
+		    GLenum type,
+		    GLuint size,
+		    GLuint start,
+		    GLuint n )
+{
+   _math_trans_4f_tab[size][TYPE_IDX(type)]( to, ptr, stride, start, n ); 
+}
+
+void _math_trans_3f(GLfloat (*to)[3],
+		    CONST void *ptr,
+		    GLuint stride,
+		    GLenum type,
+		    GLuint start,
+		    GLuint n )
+{
+   _math_trans_3f_tab[TYPE_IDX(type)]( to, ptr, stride, start, n ); 
+}
+
diff --git a/src/mesa/math/m_translate.h b/src/mesa/math/m_translate.h
index a109dfa..b6db1f6 100644
--- a/src/mesa/math/m_translate.h
+++ b/src/mesa/math/m_translate.h
@@ -1,4 +1,4 @@
-/* $Id: m_translate.h,v 1.2 2000/11/17 21:01:49 brianp Exp $ */
+/* $Id: m_translate.h,v 1.3 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -31,66 +31,52 @@
 #include "config.h"
 
 
-typedef void (*trans_1f_func)(GLfloat *to,
-			      CONST void *ptr,
-			      GLuint stride,
-			      GLuint start, 
-			      GLuint n );
 
-typedef void (*trans_1ui_func)(GLuint *to,
-			       CONST void *ptr,
-			       GLuint stride,
-			       GLuint start, 
-			       GLuint n );
-
-typedef void (*trans_1ub_func)(GLubyte *to,
-			       CONST void *ptr,
-			       GLuint stride,
-			       GLuint start,
-			       GLuint n );
-
-typedef void (*trans_4ub_func)(GLubyte (*to)[4],
-			       CONST void *ptr,
-			       GLuint stride,
-			       GLuint start,
-			       GLuint n );
-
-typedef void (*trans_4f_func)(GLfloat (*to)[4],
-			      CONST void *ptr,
-			      GLuint stride,
-			      GLuint start, 
-			      GLuint n );
-
-typedef void (*trans_3f_func)(GLfloat (*to)[3],
-			      CONST void *ptr,
-			      GLuint stride,
-			      GLuint start, 
-			      GLuint n );
-
-
-
-
-/* Translate GL_UNSIGNED_BYTE, etc to the indexes used in the arrays
- * below.
- */
-#define TYPE_IDX(t) ((t) & 0xf)
-
-#define MAX_TYPES TYPE_IDX(GL_DOUBLE)+1      /* 0xa + 1 */
-
-/* Only useful combinations are defined, thus there is no function to
- * translate eg, ubyte->float or ubyte->ubyte, which are never used.  
- */
-extern trans_1f_func gl_trans_1f_tab[MAX_TYPES];
-extern trans_1ui_func gl_trans_1ui_tab[MAX_TYPES];
-extern trans_1ub_func gl_trans_1ub_tab[MAX_TYPES];
-extern trans_3f_func  gl_trans_3f_tab[MAX_TYPES];
-extern trans_4ub_func gl_trans_4ub_tab[5][MAX_TYPES];
-extern trans_4f_func  gl_trans_4f_tab[5][MAX_TYPES];
-
-
-
-extern void 
-_math_init_translate( void );
+extern void _math_trans_1f(GLfloat *to,
+			   CONST void *ptr,
+			   GLuint stride,
+			   GLenum type,
+			   GLuint start,
+			   GLuint n );
+
+extern void _math_trans_1ui(GLuint *to,
+			    CONST void *ptr,
+			    GLuint stride,
+			    GLenum type,
+			    GLuint start,
+			    GLuint n );
+
+extern void _math_trans_1ub(GLubyte *to,
+			    CONST void *ptr,
+			    GLuint stride,
+			    GLenum type,
+			    GLuint start,
+			    GLuint n );
+
+extern void _math_trans_4ub(GLubyte (*to)[4],
+			    CONST void *ptr,
+			    GLuint stride,
+			    GLenum type,
+			    GLuint size,
+			    GLuint start,
+			    GLuint n );
+
+extern void _math_trans_4f(GLfloat (*to)[4],
+			   CONST void *ptr,
+			   GLuint stride,
+			   GLenum type,
+			   GLuint size,
+			   GLuint start,
+			   GLuint n );
+
+extern void _math_trans_3f(GLfloat (*to)[3],
+			   CONST void *ptr,
+			   GLuint stride,
+			   GLenum type,
+			   GLuint start,
+			   GLuint n );
+
+extern void _math_init_translate( void );
 
 
 #endif
diff --git a/src/mesa/math/m_vector.c b/src/mesa/math/m_vector.c
index 4dbf68f..e582ccc 100644
--- a/src/mesa/math/m_vector.c
+++ b/src/mesa/math/m_vector.c
@@ -1,4 +1,4 @@
-/* $Id: m_vector.c,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+/* $Id: m_vector.c,v 1.2 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -85,7 +85,7 @@ void gl_vector4f_init( GLvector4f *v, GLuint flags, GLfloat (*storage)[4] )
    v->data = storage;
    v->start = (GLfloat *) storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags | VEC_GOOD_STRIDE;
+   v->flags = size_bits[4] | flags ;
 }
 
 void gl_vector3f_init( GLvector3f *v, GLuint flags, GLfloat (*storage)[3] )
@@ -94,7 +94,7 @@ void gl_vector3f_init( GLvector3f *v, GLuint flags, GLfloat (*storage)[3] )
    v->data = storage;
    v->start = (GLfloat *) storage;
    v->count = 0;
-   v->flags = flags | VEC_GOOD_STRIDE;
+   v->flags = flags ;
 }
 
 void gl_vector1f_init( GLvector1f *v, GLuint flags, GLfloat *storage )
@@ -103,7 +103,7 @@ void gl_vector1f_init( GLvector1f *v, GLuint flags, GLfloat *storage )
    v->data = storage;
    v->start = (GLfloat *)storage;
    v->count = 0;
-   v->flags = flags | VEC_GOOD_STRIDE;
+   v->flags = flags ;
 }
 
 void gl_vector4ub_init( GLvector4ub *v, GLuint flags, GLubyte (*storage)[4] )
@@ -112,7 +112,7 @@ void gl_vector4ub_init( GLvector4ub *v, GLuint flags, GLubyte (*storage)[4] )
    v->data = storage;
    v->start = (GLubyte *) storage;
    v->count = 0;
-   v->flags = flags | VEC_GOOD_STRIDE;
+   v->flags = flags ;
 }
 
 void gl_vector1ub_init( GLvector1ub *v, GLuint flags, GLubyte *storage )
@@ -121,7 +121,7 @@ void gl_vector1ub_init( GLvector1ub *v, GLuint flags, GLubyte *storage )
    v->data = storage;
    v->start = (GLubyte *) storage;
    v->count = 0;
-   v->flags = flags | VEC_GOOD_STRIDE;
+   v->flags = flags ;
 }
 
 void gl_vector1ui_init( GLvector1ui *v, GLuint flags, GLuint *storage )
@@ -130,7 +130,7 @@ void gl_vector1ui_init( GLvector1ui *v, GLuint flags, GLuint *storage )
    v->data = storage;
    v->start = (GLuint *) storage;
    v->count = 0;
-   v->flags = flags | VEC_GOOD_STRIDE;
+   v->flags = flags ;
 }
 
 
@@ -153,7 +153,7 @@ void gl_vector4f_alloc( GLvector4f *v, GLuint flags, GLuint count,
    v->start = (GLfloat *) v->storage;
    v->data = (GLfloat (*)[4]) v->storage;
    v->count = 0;
-   v->flags = size_bits[4] | flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+   v->flags = size_bits[4] | flags | VEC_MALLOC ;
 }
 
 void gl_vector3f_alloc( GLvector3f *v, GLuint flags, GLuint count,
@@ -164,7 +164,7 @@ void gl_vector3f_alloc( GLvector3f *v, GLuint flags, GLuint count,
    v->start = (GLfloat *) v->storage;
    v->data = (GLfloat (*)[3]) v->storage;
    v->count = 0;
-   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+   v->flags = flags | VEC_MALLOC ;
 }
 
 void gl_vector1f_alloc( GLvector1f *v, GLuint flags, GLuint count,
@@ -175,7 +175,7 @@ void gl_vector1f_alloc( GLvector1f *v, GLuint flags, GLuint count,
       ALIGN_MALLOC( count * sizeof(GLfloat), alignment );
    v->data = v->start;
    v->count = 0;
-   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+   v->flags = flags | VEC_MALLOC ;
 }
 
 void gl_vector4ub_alloc( GLvector4ub *v, GLuint flags, GLuint count,
@@ -186,7 +186,7 @@ void gl_vector4ub_alloc( GLvector4ub *v, GLuint flags, GLuint count,
    v->start = (GLubyte *) v->storage;
    v->data = (GLubyte (*)[4]) v->storage;
    v->count = 0;
-   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+   v->flags = flags | VEC_MALLOC ;
 }
 
 void gl_vector1ub_alloc( GLvector1ub *v, GLuint flags, GLuint count,
@@ -197,7 +197,7 @@ void gl_vector1ub_alloc( GLvector1ub *v, GLuint flags, GLuint count,
    v->start = (GLubyte *) v->storage;      
    v->data = (GLubyte *) v->storage;
    v->count = 0;
-   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+   v->flags = flags | VEC_MALLOC ;
 }
 
 void gl_vector1ui_alloc( GLvector1ui *v, GLuint flags, GLuint count,
@@ -208,7 +208,7 @@ void gl_vector1ui_alloc( GLvector1ui *v, GLuint flags, GLuint count,
    v->start = (GLuint *) v->storage;      
    v->data = (GLuint *) v->storage;
    v->count = 0;
-   v->flags = flags | VEC_MALLOC | VEC_GOOD_STRIDE;
+   v->flags = flags | VEC_MALLOC ;
 }
 
 
diff --git a/src/mesa/math/m_vector.h b/src/mesa/math/m_vector.h
index c4af1ea..24a392a 100644
--- a/src/mesa/math/m_vector.h
+++ b/src/mesa/math/m_vector.h
@@ -1,4 +1,4 @@
-/* $Id: m_vector.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+/* $Id: m_vector.h,v 1.2 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -35,18 +35,13 @@
 #include "glheader.h"
 
 
-#define VEC_DIRTY_0        0x1	/* dirty flags not really used any more */
+#define VEC_DIRTY_0        0x1	
 #define VEC_DIRTY_1        0x2
 #define VEC_DIRTY_2        0x4
 #define VEC_DIRTY_3        0x8
 #define VEC_MALLOC         0x10 /* storage field points to self-allocated mem*/
-#define VEC_WRITABLE       0x20	/* keep both + and - bits for easy testing */
-#define VEC_NOT_WRITABLE   0x40
-#define VEC_GOOD_STRIDE    0x80	
-#define VEC_BAD_STRIDE     0x100
-
-#define VEC_WRITABLE_FLAGS (VEC_WRITABLE|VEC_NOT_WRITABLE)
-#define VEC_STRIDE_FLAGS   (VEC_GOOD_STRIDE|VEC_BAD_STRIDE)
+#define VEC_NOT_WRITEABLE  0x40	/* writable elements to hold clipped data */
+#define VEC_BAD_STRIDE     0x100 /* matches tnl's prefered stride */
 
 
 #define VEC_SIZE_1   VEC_DIRTY_0
diff --git a/src/mesa/math/m_xform.c b/src/mesa/math/m_xform.c
index a6e34e9..72aec3a 100644
--- a/src/mesa/math/m_xform.c
+++ b/src/mesa/math/m_xform.c
@@ -1,4 +1,4 @@
-/* $Id: m_xform.c,v 1.4 2000/11/24 10:25:11 keithw Exp $ */
+/* $Id: m_xform.c,v 1.5 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -41,6 +41,7 @@
 #include "macros.h"
 #include "mmath.h"
 
+#include "m_eval.h"
 #include "m_matrix.h"
 #include "m_translate.h"
 #include "m_xform.h"
@@ -250,4 +251,5 @@ _math_init( void )
    _math_init_transformation();
    _math_init_translate();
    _math_init_vertices();
+   _math_init_eval();
 }
diff --git a/src/mesa/math/m_xform.h b/src/mesa/math/m_xform.h
index 7fc5be3..046fc3d 100644
--- a/src/mesa/math/m_xform.h
+++ b/src/mesa/math/m_xform.h
@@ -1,4 +1,4 @@
-/* $Id: m_xform.h,v 1.2 2000/11/17 21:01:49 brianp Exp $ */
+/* $Id: m_xform.h,v 1.3 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -75,8 +75,8 @@ do {								\
 
 
 extern void gl_transform_vector( GLfloat u[4],
-				 const GLfloat v[4],
-                                 const GLfloat m[16] );
+				 CONST GLfloat v[4],
+                                 CONST GLfloat m[16] );
 
 
 extern void 
@@ -132,37 +132,36 @@ _math_init_transformation( void );
 #define CLIP_NEAR_BIT    0x10
 #define CLIP_FAR_BIT     0x20
 #define CLIP_USER_BIT    0x40
-#define CLIP_CULLED_BIT  0x80	/* Vertex has been culled */
 #define CLIP_ALL_BITS    0x3f
 
 
 typedef GLvector4f * (_XFORMAPIP clip_func)( GLvector4f *vClip,
-				  GLvector4f *vProj, 
-				  GLubyte clipMask[],
-				  GLubyte *orMask, 
-				  GLubyte *andMask );
+					     GLvector4f *vProj, 
+					     GLubyte clipMask[],
+					     GLubyte *orMask, 
+					     GLubyte *andMask );
 
-typedef void (*dotprod_func)( GLvector4f *out_vec, 
-			      GLuint elt,
-			      const GLvector4f *coord_vec, 
-			      const GLfloat plane[4], 
-			      const GLubyte mask[]);
+typedef void (*dotprod_func)( GLfloat *out, 
+			      GLuint out_stride,
+			      CONST GLvector4f *coord_vec, 
+			      CONST GLfloat plane[4], 
+			      CONST GLubyte mask[]);
 
 typedef void (*vec_copy_func)( GLvector4f *to, 
-			       const GLvector4f *from, 
-			       const GLubyte mask[]);
+			       CONST GLvector4f *from, 
+			       CONST GLubyte mask[]);
 
 
 
 /*
  * Functions for transformation of normals in the VB.
  */
-typedef void (_NORMAPIP normal_func)( const GLmatrix *mat,
-			     GLfloat scale,
-			     const GLvector3f *in,
-			     const GLfloat lengths[],
-			     const GLubyte mask[],
-			     GLvector3f *dest );
+typedef void (_NORMAPIP normal_func)( CONST GLmatrix *mat,
+				      GLfloat scale,
+				      CONST GLvector3f *in,
+				      CONST GLfloat lengths[],
+				      CONST GLubyte mask[],
+				      GLvector3f *dest );
 
 
 /* Flags for selecting a normal transformation function. 
@@ -181,26 +180,27 @@ typedef void (_NORMAPIP normal_func)( const GLmatrix *mat,
  *     parameter, to allow a unified interface.  
  */
 typedef void (_XFORMAPIP transform_func)( GLvector4f *to_vec,
-					  const GLfloat m[16],
-				const GLvector4f *from_vec, 
-				const GLubyte *clipmask,
-				const GLubyte flag );
+					  CONST GLfloat m[16],
+					  CONST GLvector4f *from_vec, 
+					  CONST GLubyte *clipmask,
+					  CONST GLubyte flag );
 
 
 extern GLvector4f *gl_project_points( GLvector4f *to,
-			       const GLvector4f *from );
+				      CONST GLvector4f *from );
 
 extern void gl_transform_bounds3( GLubyte *orMask, GLubyte *andMask,
-				  const GLfloat m[16],
-			   CONST GLfloat src[][3] );
+				  CONST GLfloat m[16],
+				  CONST GLfloat src[][3] );
 
 extern void gl_transform_bounds2( GLubyte *orMask, GLubyte *andMask,
-				  const GLfloat m[16],
-			   CONST GLfloat src[][3] );
+				  CONST GLfloat m[16],
+				  CONST GLfloat src[][3] );
 
 
 extern dotprod_func  gl_dotprod_tab[2][5];
 extern vec_copy_func gl_copy_tab[2][0x10];
+extern vec_copy_func gl_copy_clean_tab[2][5];
 extern clip_func     gl_clip_tab[5];
 extern normal_func   gl_normal_tab[0xf][0x4];
 
@@ -210,17 +210,13 @@ extern normal_func   gl_normal_tab[0xf][0x4];
 extern transform_func **(gl_transform_tab[2]);
 
 
-extern void gl_transform_point_sz( GLfloat Q[4], const GLfloat M[16],
-				   const GLfloat P[4], GLuint sz );
+extern void gl_transform_point_sz( GLfloat Q[4], CONST GLfloat M[16],
+				   CONST GLfloat P[4], GLuint sz );
 
 
 #define TransformRaw( to, mat, from ) \
       ( (*gl_transform_tab[0][(from)->size][(mat)->type])( to, (mat)->m, from, 0, 0 ), \
         (to) )
 
-#define Transform( to, mat, from, mask, cull ) \
-      ( (*gl_transform_tab[cull!=0][(from)->size][(mat)->type])( to, (mat)->m, from, mask, cull ), \
-	(to) )
-
 
 #endif
diff --git a/src/mesa/math/m_xform_tmp.h b/src/mesa/math/m_xform_tmp.h
index 289255a..cb94f88 100644
--- a/src/mesa/math/m_xform_tmp.h
+++ b/src/mesa/math/m_xform_tmp.h
@@ -1,4 +1,4 @@
-/* $Id: m_xform_tmp.h,v 1.1 2000/11/16 21:05:41 keithw Exp $ */
+/* $Id: m_xform_tmp.h,v 1.2 2000/12/26 05:09:31 keithw Exp $ */
 
 /*
  * Mesa 3-D graphics library
@@ -457,10 +457,7 @@ TAG(transform_points2_3d_no_rot)( GLvector4f *to_vec,
    to_vec->count = from_vec->count;
 }
 
-/* This may not be called too often, but I wouldn't say it was dead
- * code.  It's also hard to remove any of these functions if you are
- * attached to the assertions that have appeared in them.
- */
+
 static void _XFORMAPI
 TAG(transform_points2_perspective)( GLvector4f *to_vec,
 				    const GLfloat m[16],