How large is ARM market share in automotive? From outsider perspective, it looks to be concentrated in entertainment subsystems, the real time control still seems to be dominated by 16/32 bit non-ARM microcontrollers (Renesas, Toshiba, NXP).
Source: the very few automotive parts I've repaired or reverse-engineered, mostly designed in early 2000s.
You are right, but ARM is getting stronger. In IVI you mostly have no other choice (only Intel), since the main contenders (SH4, top-end PowerPC) are now discontinued. Which I personally thought was a good thing, because in IVI you want to reuse lots of existing software from outside of automotive, which often has only support for Intel and ARM.
For other domains other architectures .e.g from Renesas) are still going strong. 16bit is however nearly dead, since most standardized software (Autosar) requires 32bit. However I also see ARM cores getting traction there. But even if they are coming in it will take some time until you see them in real cars. Keep in mind that development cycles are >= 3 years, which means even if someone decides to use an ARM core in a ECU which is designed now you won't see it before 2020 on the streets.
In both of those chips the CPUs themselfes are the smallest part but the on chip peripherals differ greatly by usecase. The automotive has a great plethora of different busses and i/o (digital, analog, complex PWM, CAN, LIN, FlexRay) while the typical ARM core has less but higher bandwith IOs.
The other big step is backwards compatibility. Going to a completely different architecture is a bigger step than going from one iteration to the next one.
So probably even the "new" designs for 2020 may use the successor of the current generation of uC or SoC chips.
Yes, peripherals differ. But that's not related to ARM vs sth. else but generally for automotive-specific vs other domain. Automotive specific ARM SoCs will contain the same peripherals than their RH850, PPC, etc. counterparts if required. Smartphone specific SoCs will of course favor other peripherals. But these won't be directly used in automotive anyway, because they are lacking the required certifications (temperature range, etc...)
I think ARM will dominate the next generation. The current offerings are either too limited (8051 anyone?) or too obscure to be worth keeping alive (powerpc).
Source: the very few automotive parts I've repaired or reverse-engineered, mostly designed in early 2000s.