most awesome

I've already tested this and can confirm that it does indeed deliver the big bandwidth increase

@breakintoprogram please merge ASAP 😁

Thanks. That looks reasonable to me. I'll merge and test later.

Awesome. It's surprising (to me) at the first glance that this change makes such a notable difference !

Awesome. It's surprising (to me) at the first glance that this change makes such a notable difference !

the reason for this I believe is that essentially the previous version was calling readByte_b 4 times, each time of which runs thru a while loop waiting for a byte to become available. if the comms buffer were full (or rather just had those 4 bytes available, which is fairly likely), there would always be bytes available so that's 3x the amount of checking loops needed. additionally the readBytes call here would be a simple memory copy, so there's no shuffling of bytes around to form the 32-bit word.

I kinda get why it's a fairly big improvement - and conversely am also still surprised by how big an improvement it is. 😁

gives me some ideas for some further improvements that can be made - but those will have to wait a little while.

I was a bit surprised as well. But you have to remember this is a 32bit CPU. It's most happy to fetch 32bit around. It could basically as costly to fetch 32bit at once or just 8bit.
The emulator for example does run the old code at the same speed, too. It just that the esp32 was limiting here. And now I can show the ez80 does limit. If you remove the last waitstate for the flash memory currently in place you even get 800kbits.