richardman what am I really thinking ? Microchip killing off the serial processing market with a pile of shit hardware and software that doesn't work real time. What is that interrupt latency that isn't documented, I shouldn't have to measure it with a scope but I have to, because to measure is to know. Serial processing, thats old school, and Microchip screwing things up with abstracted software.
I don't think Microchip got a clue about processing, they can't design hardware peripherals that work let alone software.
Of course you have to measure this. How else are you going to know? Or could you tell me exactly what software timing requirements you have on a STM32F427 Cortex m4 chip with 8 USARTs firing off RXNE, TXE & IDLE interrupts, timer interrupts, DMA transfer complete requests, ethernet/USB ticks while some DMA request is stalling the complete peripheral bus because the APB1 DMA peripheral is trying to access the APB2 bus?
No - what's even more great is that the USART implementation on (I believe any) STM32 doesn't have a FIFO. That's great when it's serving 8 USARTs @ 115200 Baud. The only answer you will find is "DMA is it all".. but no not really. What about variable packet sizes, response latency when you use DMA timeouts, or DMA stream/request collisions with other peripherals? In order to find this all out, you are forced to look up the
1731 page user manual and hope you don't miss anything. Else you're basically screwed.
And don't get me started on the STM timers: could they have made that any more complex? And it still doesn't do what I wanted it to do - like generate DMA requests with ETR pin at that frequency - not at half of it.
Although the peripherals of PIC are usually very simple and scarce of configuration bits - I do like some of the flexibility. E.g. there is no messing about with DMA request tables; just enter an IRQn and it's all good.
Can't say much good about the 8-bit PIC core. I stopped using them for anything requiring more than 1 C source file. If anything is gained from this acquisition; they should at least merge the AVR core with some of the PICs pluses.
I don't agree that Atmel/Microchip should watch carefully are the hobbyists. The "market share" of hobbyists is absolutely nothing - yet we find ourselves very important in the market. Although many students & young engineers will probably start using a particular microcontroller they have used in their quality time, for anything that's really work related it's all about the features, familiarity with other engineers and development time/cost.