MikktspaceTangentGenerator fails with NPE

zzuegg · December 28, 2024, 10:23am

Since i was scrolling through the code too, the MikktspaceGenerator contains quite a few of todo’s. one of the comments seems to indicate that the original implementation uses some pointer magic.

github.com

jMonkeyEngine/jmonkeyengine/blob/fc32a26c930c3635573ef5d335bffb3453e71a35/jme3-core/src/main/java/com/jme3/util/mikktspace/MikktspaceTangentGenerator.java#L294


      
                      }
                  }
          
                  return true;
              }
          
          ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////
              // it is IMPORTANT that this function is called to evaluate the hash since
              // inlining could potentially reorder instructions and generate different
              // results for the same effective input value fVal.
              //TODO nehon: Wuuttt? something is fishy about this. How the fuck inlining can reorder instructions? Is that a C thing?
              static int findGridCell(final float min, final float max, final float val) {
                  final float fIndex = CELLS * ((val - min) / (max - min));
                  final int iIndex = (int) fIndex;
                  return iIndex < CELLS ? (iIndex >= 0 ? iIndex : 0) : (CELLS - 1);
              }
          
              static void generateSharedVerticesIndexList(int piTriList_in_and_out[], final MikkTSpaceContext mikkTSpace, final int iNrTrianglesIn) {
          
                  // Generate bounding box
                  TmpVert[] pTmpVert;

So the questions standing is, is the jme mikktspace generator tested against other implementations? since the primary reason for it’s existence is to have an industry standard that should produce the same results in every engine/normal map backer