| ??? 09/16/09 21:55 Read: times |
#168924 - The jury is still out ... Responding to: ???'s previous message |
Per Westermark said:
The thing is that some people build systems. Then these complete Linux++ boards may be possible platforms to start with.
But Linux isn't natively an RTOS. And when money count, you may want to save on flash and RAM. A lot of people want a simple microcontroller. The ARM7 or Cortex-M3 chips starts at around $1. That is significant if you need to produce 10k+/year of a product. If you save one dollar on the processor, and the Keil tools costs $4000, then the savings will cover the Keil license within 3-4 months. And the Keil tools may get your product off to market a month earlier than if you decode to use gcc with the same processor which may result in 1k more units sold. There are a huge number of products that are very price-pressed, but needs more computing power than the 8051 can mangage, but at the same time needs strict real-time performance and isn't so well suited for a Linux platform. Just because one has LINUX as a native development platform isn't a mandate to use LINUX in the target. Further, there are lots of 8-bit things that the ARM doesn't do as well as the 805x, particularly if speed is an issue. That doesn't apply across the board, but it does apply in some cases. That ARM9 board to which I referred certainly isn't top-of-the-line, but it can support LINUX on its own, requiring only the addition of mass storage. It does, after all, have an attached console, less keyboard, though there's a port for it. The last time I used an ARM9, it was still not capable, at 150 MHz, of outperforming a one-clocker 805x type at 33 MHz when doing byte-intensive processing. Of course, a 32-bit add was no problem. The reason ARM is gaining is that it can compete with the cheapest 8-bit chips. And it can compete with probably all 16-bit chips. And it can compete with most 32-bit chips that are intended for some form of embedded and/or low-power use. On the other hand, users of ARM chips should know that the peripherals can vary totally between different chips, and that when coding, the speed of the core and the availability of integer division or floating point can make a huge difference. It is easy to consider all ARM chips to be within the same family, but the diversity is huge, and knowledge of one ARM chip does not mean that you save so much time when moving to a different ARM chip, unless it is bought from the same manufacturer and they have tried to reuse peripheral functions. I'll not conclude that ARM is the be-all-and-end-all, but it is attractive. Of course, for FPGA implementation, there are very effective MIPS cores, among several others, that are supported with gcc, and which run LINUX just as well as ARM does. That, naturally, presents other attractive options. I do believe that off-the-shelf ARM-core CPU's, particularly if they have the peripherals that I want. What's really desirable about the ones that are shipped as part of a development platform supporting LINUX is that one can then rely on being able to get pretty decent and proven drivers for the desired peripherals on which one can base an embedded non-LINUX system. Moreover, one doesn't have to pay the $4k for KEIL software, even if it is spreadable over 10K units/year. RE |
