Do not understand MotionBlurRenderPass

Dear ForumReaders,

can anyone help me with the following code-snip from

MotionBlurRenderPass (line 294):

      if( !freeze ) {
         for( int i = 0; i < dynamicObjects.size(); i++ ) {
            DynamicObject dynamicObject = dynamicObjects.get( i );
            Matrix4f modelMatrix = dynamicObject.modelMatrix;
            Spatial spatial = dynamicObject.spatial;

            spatial.getWorldRotation().toRotationMatrix( tmpMatrix );
            modelMatrix.multLocal( tmpMatrix );
            modelMatrix.m00 *= spatial.getWorldScale().x;
            modelMatrix.m11 *= spatial.getWorldScale().y;
            modelMatrix.m22 *= spatial.getWorldScale().z;
            modelMatrix.setTranslation( spatial.getWorldTranslation() );

Seems to be a little confusing to me - the rotation of the dynObj stored into tmpMatrix, then again multiplicated with the modelMatrix, then scaled, translated and afterwards transposed. I know transposing is useful for the inversion of a rotation, but in this case? What mathematical trick is done here - and for what? I know the algo behind the shader-code used here (Stupid OpenGL Shader Tricks, NVIDIA), but my brain is freezed with this snip - can anyone explain?

Thnx in advance for any help,
Best Regards Woidl

Ahhhh, think I understand it now - the Identity+MultLocal stuff is cause .toRotationMatrix() does not touch the 4th col/row.

It's just the update of the next transformation of the spatial. Think transposeLocal() is necessary for the shader - correct?