[philiplu]

That sounds reasonable. I'd forgotten about the p-code version. I think that was dead by the time I started. I was also fortunate enough to avoid working on the 16-bit compiler backend, and just worked on the 32-bit, referred to internally as the n386 backend, which was pretty much a complete rewrite. The p-code backend would have been a variant of the 16-bit version.

[tragomaskhalos]

The 'own C compiler' is from Joel Spolsky from when he worked at Microsoft: https://www.joelonsoftware.com/2001/10/14/in-defense-of-not-.... And indeed this states that it was p-code. I'm old enough to remember when p-code was - briefly - a very hot idea.

[philiplu]

I know we shipped p-code for general usage, not just internal like Excel. But it might have been internal-only in the mid to late 80s. I found a link [1] talking about some p-code internals, from April 1992, by the guy who hired me at Microsoft about a year before that. The compiler team definitely viewed Excel as one of their most important customers, and were willing to do lots of work to satisfy them.

So p-code wasn't actually dead when I started, but instead shipped with the C/C++ 7.0 compiler, which predated Visual C++. I never worked on C7, since that was 16-bit, and p-code wasn't part of any 32-bit compiler, as I recall.

[1] https://techshelps.github.io/MSDN/BACKGRND/html/msdn_c7pcode...