Question about structure memory allocation

Accelerated Computing CUDA CUDA Programming and Performance
1

Here are some parts of CUDA code from the CUDA course task. I think pointerp should point to the address of the first element in Body, but then Body has six floats, Not quite sure why p and buf are same. I try to write something like cudaMallocManaged(&p, 6*bytes), but why it cannot help to improve performance?

typedef struct { float x, y, z, vx, vy, vz; } Body;

int main(const int argc, const char** argv)
  int nBodies = 2<<11;
  int bytes = nBodies * sizeof(Body);
  float *buf;
  cudaMallocManaged(&buf, bytes); 
  Body *p = (Body*)buf;
2

Does it need more information?