Difference between ARM and any other General Purpose Processor

Take your homework to your professor.

pete

"Sikandar" skrev i meddelandet news: snipped-for-privacy@e1g2000hsg.googlegroups.com...

An ARM processor uses an Instruction Set Architecture developed by ARM Ltd, and other processors don't :-)

"Ulf Samuelsson" a écrit dans le message de news:eu00b1$n2a$ snipped-for-privacy@aioe.org...

Why ARM procesor are used by many people ?

--

-Stan

It's a plot by the Powers That Be.

There is no ARM processor. There is a ARM processor core, that can be used to build an ARM based processor. And yes, there are a lot of ARM based processors, just like there are a lot of MIPS-based processors. I don't know which is used more often, but the processor you choose depends on the features you need on a processor, the processing power and the budget.

Kind regards, Johan Borkhuis

The ARM processor is a 5 digit processor, most others are 8, 16, 32 or 64 bit.

"Stan" skrev i meddelandet news:eu03rn$6u8$ snipped-for-privacy@s1.news.oleane.net...

ARM cores lived a pretty insignificant life until Apple decided to design the Newton PDA. The ARM chip exisited, and did the job so Apple Engineers selected it. Every ASIC vendor wanted to be in this business, and they were aware of the requirement of binary compatibility from their experience from the PC market, so a lot of the ASIC vendors licensed the ARM core. Intel and Motorola refused to license their cores for ASIC use. When Apple canned the Newton the ASIC vendors were sitting with a core, which started to become useful in different ASIC projects. Then some vendors started to promote ARM7 as general purpose microcontrollers, (Atmel was one of the first). DEC, then Intel started to promote the StrongARM followed by XScale. The ARM9 based AT91RM9200, introduced is very good for the embedded Linux market, and the AT91SAM926x family seems to even more successful. When NXP released their flash based ARM7, quickly followed by the Atmel SAM7 (rest of competition hardly counts) people began to see the possibility to use same toolset with many vendors. People believe that they will be able to live with ARM for a long time. At the same time, the makers of ASSPs and especially mobile phones and PDAs have converged on ARM and this has made it very attractive for third party vendors. When potential customers see the variety of support for ARM, then it is going to be a great motivator to select it. ARM cores are not really *excellent* cores, MIPS cores are generally faster, but good enough for many applications MIPS cores have other problems. IIRC You have to use a compiler since there are strange things in the compiler which would confuse an ordinary assembler programmer.

Historical accident mostly. Things really took off as ARM had the right product (the ARM7TDMI) in the right place (mobile phones) at the right time (circa 1995).

The ARM7TDMI gave the developers a full 32 bit processor which they could also use with the 16 bit THUMB instruction set.

While it might not seem important now the code size benefits of THUMB meant that the vendors could minimise the the amount of flash in their phone and when you are making millions of units being able to halve the amount of flash you need makes a very big difference to your bottom line.

-p

Just to make sure... that's only in Thumb mode, right?

...had to ask...

-f

You are probably referring to instructions being synthesized by the assembler (not the compiler), e.g. MIPS doesn't have an instruction to load a 32-bit constant, nevertheless the assembler understands e.g. "la " and generates an appropriate sequence of instructions. Also, the AT register ("assembler temporary") can be used as scratchpad register by the assembler at any time, so an assembly programmer should not use it. However, all these features can be disabled by assembly pragmas.

So, you *can* program MIPS in assembly if you really need to.

Rob

But don't forget they usually come in complementary pairs and used in parallel.

Many arms make light work

That is not specific to MIPS - any RISC processor that does not have variable length instructions will be unable to load full-width constants directly.

My guess is that he is referring to delay slots (a useful idea, which can greatly reduce the wasted cycles for pipeline flushes on branches). On the MIPS, branch and jump instructions do not act immediately - they work more as sort of "do the next two instructions, *then* branch" instructions. Thus in a function return, you'll get a "return" instruction followed by the last two instructions of the function. This can get very confusing if you are not used to it, especially if you want other conditional instructions during the delays. Additionally, the number of delay slots varies slightly in different circumstances, and between different MIPS cores. I believe the assembler can give you some help, but it's easier to let the compiler get it right!