MP/M ran fine on the Z-80. Only the mnemonics were different between Intel and Zilog. The byte-level machine code of the 8080/8085 would run just fine on the Z-80.
I was once blessed with a z80 c compiler, hosted on msdos but targeting cp/m (maybe it was ported for use with embedded targets). Its output was 8080 assembler with the odd db here and there when it wanted to emit a z80 instruction. It shipped with a perfectly good z80 assembler, mind; the compiler just emitted 8080 mnemonics because reasons.
I always thought that the Zilog mnemonics were better than Intel's. I learned Intel first, but immediately recognized that Zilog's were better once I became aware of them.
The only C compiler I had on CP/M was BDS C. It was fairly limited and had no floating point support whatsoever. Maybe that's the one you're thinking of?
This one was called Aztec and Manx C, depending on what part you looked at. It lacked bit field struct members and I think a couple of other features but don't remember now.
I learned Zilog first and agree with you. I had to keep an 8080 manual handy for a while to make any sense out of that compiler output.
I used Aztec C for the 68k on some embedded projects for work in the late 80's. The 68k version was a full-featured C implementation complete with IEEE-754 floating-point math and bit fields. I liked the fact that Aztec included the source code for parts of their libraries so one could write their own drivers for stdio devices. I still have that code somewhere...
yes - but as a Z80 programmer it was hard to use the 8080-oriented tooling on the MP/M machines. I mean there was a direct correspondence from 8080 to Z80 but man... 8080 was ugly.
