tweaks to the LIT instructions

diff --git a/src/mesa/shader/nvvertexec.c b/src/mesa/shader/nvvertexec.c
index 79929f1..269c4c9 100644 (file)
--- a/src/mesa/shader/nvvertexec.c
+++ b/src/mesa/shader/nvvertexec.c
@@ -412,17 +412,12 @@ _mesa_exec_vertex_program(GLcontext *ctx, const struct vertex_program *program)
             break;
          case VP_OPCODE_LIT:
             {
-               const GLfloat epsilon = 1.0e-5F; /* XXX fix? */
+               const GLfloat epsilon = 1.0F / 256.0F; /* per NV spec */
                GLfloat t[4], lit[4];
                fetch_vector4( &inst->SrcReg[0], state, t );
-               if (t[3] < -(128.0F - epsilon))
-                   t[3] = - (128.0F - epsilon);
-               else if (t[3] > 128.0F - epsilon)
-                  t[3] = 128.0F - epsilon;
-               if (t[0] < 0.0)
-                  t[0] = 0.0;
-               if (t[1] < 0.0)
-                  t[1] = 0.0;
+               t[0] = MAX2(t[0], 0.0F);
+               t[1] = MAX2(t[1], 0.0F);
+               t[3] = CLAMP(t[3], -(128.0F - epsilon), (128.0F - epsilon));
                lit[0] = 1.0;
                lit[1] = t[0];
                lit[2] = (t[0] > 0.0) ? (GLfloat) exp(t[3] * log(t[1])) : 0.0F;

diff --git a/src/mesa/swrast/s_nvfragprog.c b/src/mesa/swrast/s_nvfragprog.c
index 88a6593..d083f83 100644 (file)
--- a/src/mesa/swrast/s_nvfragprog.c
+++ b/src/mesa/swrast/s_nvfragprog.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.1
+ * Version:  6.3
  *
  * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
  *
@@ -805,14 +805,16 @@ execute_program( GLcontext *ctx,
             break;
          case FP_OPCODE_LIT:
             {
+               const GLfloat epsilon = 1.0F / 256.0F; /* from NV VP spec */
                GLfloat a[4], result[4];
                fetch_vector4( ctx, &inst->SrcReg[0], machine, program, a );
-               if (a[0] < 0.0F)
-                  a[0] = 0.0F;
-               if (a[1] < 0.0F)
-                  a[1] = 0.0F;
+               a[0] = MAX2(a[0], 0.0F);
+               a[1] = MAX2(a[1], 0.0F);
+               /* XXX ARB version clamps a[3], NV version doesn't */
+               a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
                result[0] = 1.0F;
                result[1] = a[0];
+               /* XXX we could probably just use pow() here */
                result[2] = (a[0] > 0.0F) ? (GLfloat) exp(a[3] * log(a[1])) : 0.0F;
                result[3] = 1.0F;
                store_vector4( inst, machine, result );