It depends. The main purpose of .s (and operations with the -containg symbol, for example, ldc.i4.1 ) is to simply reduce the size of the code, and the advantage of reducing the size of the method is to make it possible to embed the method when creating your own code from the CIL of the calling method (The limit for x86 jitter is 32 IL bytes). Therefore, in this case, short instructions can improve application performance if they are used in the built-in candidate method.
Otherwise, since this is not a CIL run, the machine code generated by both short and normal operation codes should be the same (and also optimized when possible) native code.