Rereading that made me wonder: if someone retargeted an old compiler (pcc, say) to produce i386 code,
I thought SVR3's was PCC (maybe PCC2). But I thought I remember that is had a i386 code. Certainly by SVR4 time.
IIRC, the time frame of SVR3's front end would have been original ANSI (i.e. White Book V2).
how much faster would it run than a VAX?
In the time frame of the SVR3 (mid/late 80s), the Intel processors was faster than the 1MIP (780 circa 1977) in raw computes. The issue was always I/O. Most PC did not have the same amount of I/O HW that much earlier Vaxen.
I see that there is a pcc derivative at <http://pcc.ludd.ltu.se/>, but supposedly it has been heavily rewritten for C99 compliance and other things.
And my point is that by the time of C99, it was a different language than the early 1970s when Dennis created fit or the original PDP-11/20 he and Ken used for the first UNIX kernel and tools implementations.