I was thinking more how branch prediction is handled in the pipeline. Agner Fog states that on Haswell having branches in the same 16 byte block affects perform

I was thinking more how branch prediction is handled in the pipeline. Agner Fog states that on Haswell having branches in the same 16 byte block affects performance. Probably not enough to matter though.