[Open64-devel] WHIRL question

[C. B. <cbe...@pa...>]

(forwarding this question)

This code
        j = --i + i--;

Is transformed into this (pseudo)WHIRL:

STID(i) - ADD - LDID(i)                         ; this is the side
              \ INTCONST(-1)                    ; effect of the --i expression

STID(preg_I4) - LDID(i)				; storage of the previous
						; i value

STID(i) - ADD - LDID(i)                         ; this is the side
              \ INTCONST(-1)                    ; effect of the i-- expression

STID(j) - ADD - LDID(preg_I4)                   ; the actual
              \ LDID(i)                         ; j = .... expression

what is the advantage of this WHIRL instead of

STID(i) - ADD - LDID(i)                         ; this is the side
              \ INTCONST(-1)                    ; effect of the --i expression

STID(j) - ADD - LDID(i)                         ; the actual
              \ LDID(i)                         ; j = .... expression

STID(i) - ADD - LDID(i)                         ; this is the side
              \ INTCONST(-1)                    ; effect of the --i expression


the first version just uses one more WHIRL node and one more variable.
why is it used? is it because it's easier to rewrite C -> WHIRL in that way?