summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorBrian Paul <[email protected]>2004-02-15 16:21:07 +0000
committerBrian Paul <[email protected]>2004-02-15 16:21:07 +0000
commit7c4a61c57fea4c8f112fc22278f643bee2fbaa45 (patch)
tree5eaaf2f2f505ae74a03d30a6c2e4ae936070d252
parent0200947f5e35f1dc7baae776d2e3a8206860335a (diff)
Minor clean-ups of variable scopes, initializations, etc.
New comments discussing FIXED_FRAC_BITS, SUB_PIXEL_BITS, max viewport size and rasterization accuracy.
-rw-r--r--src/mesa/swrast/s_tritemp.h80
1 files changed, 51 insertions, 29 deletions
diff --git a/src/mesa/swrast/s_tritemp.h b/src/mesa/swrast/s_tritemp.h
index 1949e7cd40f..74dc13e2997 100644
--- a/src/mesa/swrast/s_tritemp.h
+++ b/src/mesa/swrast/s_tritemp.h
@@ -1,8 +1,8 @@
/*
* Mesa 3-D graphics library
- * Version: 5.1
+ * Version: 6.1
*
- * Copyright (C) 1999-2003 Brian Paul All Rights Reserved.
+ * Copyright (C) 1999-2004 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"),
@@ -63,6 +63,34 @@
* This code was designed for the origin to be in the lower-left corner.
*
* Inspired by triangle rasterizer code written by Allen Akin. Thanks Allen!
+ *
+ *
+ * Some notes on rasterization accuracy:
+ *
+ * This code uses fixed point arithmetic (the GLfixed type) to iterate
+ * over the triangle edges and interpolate ancillary data (such as Z,
+ * color, secondary color, etc). The number of fractional bits in
+ * GLfixed and the value of SUB_PIXEL_BITS has a direct bearing on the
+ * accuracy of rasterization.
+ *
+ * If SUB_PIXEL_BITS=4 then we'll snap the vertices to the nearest
+ * 1/16 of a pixel. If we're walking up a long, nearly vertical edge
+ * (dx=1/16, dy=1024) we'll need 4 + 10 = 14 fractional bits in
+ * GLfixed to walk the edge without error. If the maximum viewport
+ * height is 4K pixels, then we'll need 4 + 12 = 16 fractional bits.
+ *
+ * Historically, Mesa has used 11 fractional bits in GLfixed, snaps
+ * vertices to 1/16 pixel and allowed a maximum viewport height of 2K
+ * pixels. 11 fractional bits is actually insufficient for accurately
+ * rasterizing some triangles. More recently, the maximum viewport
+ * height was increased to 4K pixels. Thus, Mesa should be using 16
+ * fractional bits in GLfixed. Unfortunately, there may be some issues
+ * with setting FIXED_FRAC_BITS=16, such as multiplication overflow.
+ * This will have to be examined in some detail...
+ *
+ * For now, if you find rasterization errors, particularly with tall,
+ * sliver triangles, try increasing FIXED_FRAC_BITS and/or decreasing
+ * SUB_PIXEL_BITS.
*/
@@ -558,24 +586,18 @@ static void NAME(GLcontext *ctx, const SWvertex *v0,
*/
{
- int subTriangle;
- GLfixed fx;
+ GLint subTriangle;
GLfixed fxLeftEdge = 0, fxRightEdge = 0;
GLfixed fdxLeftEdge = 0, fdxRightEdge = 0;
- GLfixed fdxOuter;
- int idxOuter;
- float dxOuter;
GLfixed fError = 0, fdError = 0;
- float adjx, adjy;
- GLfixed fy;
#ifdef PIXEL_ADDRESS
PIXEL_TYPE *pRow = NULL;
- int dPRowOuter = 0, dPRowInner; /* offset in bytes */
+ GLint dPRowOuter = 0, dPRowInner; /* offset in bytes */
#endif
#ifdef INTERP_Z
# ifdef DEPTH_TYPE
DEPTH_TYPE *zRow = NULL;
- int dZRowOuter = 0, dZRowInner; /* offset in bytes */
+ GLint dZRowOuter = 0, dZRowInner; /* offset in bytes */
# endif
GLfixed zLeft = 0, fdzOuter = 0, fdzInner;
#endif
@@ -659,31 +681,31 @@ static void NAME(GLcontext *ctx, const SWvertex *v0,
}
if (setupLeft && eLeft->lines > 0) {
- const SWvertex *vLower;
- GLfixed fsx = eLeft->fsx;
- fx = FixedCeil(fsx);
+ const SWvertex *vLower = eLeft->v0;
+ const GLfixed fsx = eLeft->fsx; /* no fractional part */
+ const GLfixed fsy = eLeft->fsy;
+ const GLfixed fx = FixedCeil(fsx); /* no fractional part */
+ const GLfloat adjx = (GLfloat) (fx - eLeft->fx0); /* SCALED! */
+ const GLfloat adjy = eLeft->adjy; /* SCALED! */
+ GLfixed fdxOuter;
+ GLint idxOuter;
+ GLfloat dxOuter;
+
fError = fx - fsx - FIXED_ONE;
fxLeftEdge = fsx - FIXED_EPSILON;
fdxLeftEdge = eLeft->fdxdy;
fdxOuter = FixedFloor(fdxLeftEdge - FIXED_EPSILON);
fdError = fdxOuter - fdxLeftEdge + FIXED_ONE;
idxOuter = FixedToInt(fdxOuter);
- dxOuter = (float) idxOuter;
- (void) dxOuter;
+ dxOuter = (GLfloat) idxOuter;
- fy = eLeft->fsy;
- span.y = FixedToInt(fy);
+ span.y = FixedToInt(fsy);
- adjx = (float)(fx - eLeft->fx0); /* SCALED! */
- adjy = eLeft->adjy; /* SCALED! */
-#ifndef __IBMCPP__
- (void) adjx; /* silence compiler warnings */
- (void) adjy; /* silence compiler warnings */
-#endif
- vLower = eLeft->v0;
-#ifndef __IBMCPP__
- (void) vLower; /* silence compiler warnings */
-#endif
+ /* silence warnings on some compilers */
+ (void) dxOuter;
+ (void) adjx;
+ (void) adjy;
+ (void) vLower;
#ifdef PIXEL_ADDRESS
{
@@ -694,7 +716,7 @@ static void NAME(GLcontext *ctx, const SWvertex *v0,
#endif
/*
* Now we need the set of parameter (z, color, etc.) values at
- * the point (fx, fy). This gives us properly-sampled parameter
+ * the point (fx, fsy). This gives us properly-sampled parameter
* values that we can step from pixel to pixel. Furthermore,
* although we might have intermediate results that overflow
* the normal parameter range when we step temporarily outside