nv2a: implement FFP LIGHT_SPOT

This commit is contained in:
Wilhelm Kovatch 2021-05-02 18:57:28 +02:00 committed by mborgerson
parent 8604d7e41e
commit e2e3572301
1 changed files with 25 additions and 2 deletions

View File

@ -566,8 +566,31 @@ GLSL_DEFINE(materialEmissionColor, GLSL_LTCTXA(NV_IGRAPH_XF_LTCTXA_CM_COL) ".xyz
/* Everything done already */
break;
case LIGHT_SPOT:
assert(false);
/*FIXME: calculate falloff */
/* From: https://docs.microsoft.com/en-us/windows/win32/direct3d9/attenuation-and-spotlight-factor#spotlight-factor */
/* Note: normalizing the direction does not produce the same result as on hardware.
* Instead, dividing the direction by 1-w produces the same result as on hardware.
* Rationale: the w component is always close (but not equal to) the difference
* between 1 and the the direction vector length (e.g. 1.0-2.6=-1.6, with w=-1.5).
* So, dividing by 1-w is the same as normalizing the vector and scaling it a bit.
* In the example: 1-(-1.5) = 2.5, 2.6/2.5 = 1.04. This also scales the subsequent
* dot product by the same amount, effectively increasing the spotlight radius.
*/
qstring_append_fmt(body,
" vec3 dir = lightSpotDirection(%d).xyz / (1.0 - lightSpotDirection(%d).w);\n"
" float rho = dot(dir, VP);\n"
" float theta = lightSpotFalloff(%d).x;\n"
" float phi = lightSpotFalloff(%d).y;\n"
" float falloff = lightSpotFalloff(%d).z;\n"
" float cosHalfPhi = cos(phi * 0.5);\n"
" float cosHalfTheta = cos(theta * 0.5);\n"
" if (rho > cosHalfTheta) {\n" /* do nothing => spotlightFactor = 1 */
" } else if (rho <= cosHalfPhi) {\n"
" attenuation = 0.0;\n" /* spotlightFactor = 0 */
" } else {\n"
" float spotlightFactor = pow((rho - cosHalfPhi) / (cosHalfTheta - cosHalfPhi), falloff);\n"
" attenuation = attenuation * spotlightFactor;\n"
" }\n",
i, i, i, i, i);
break;
default:
assert(false);