Emotional stories about processors for first computers: part 9 (Acorn ARM)

[litwr]

The first ARM processors

The ARM-1 processor was an astonishing development, it continued the 6502 ideology (namely to make a processor that is easier, cheaper and better), and was released by Acorn in 1985. This was at the same time when Intel's technological miracle the 80386 processor appeared. ARM consisted of about ten times less transistors

Comments

[anonymous]

'Almost all operations, exceptions are jumps, multiplications and block copies, are performed in 1 clock cycle'
There is another exception : using the barrel shifter with a register adds another clock cycle (0 when it is a constant).

RSUB isn't the correct mnemonic for reverse substract, it is RSB
Also your examples should read :
ADD R1, R0, R0,LSL#6
RSB R1, R0, R0,LSL#6

You wrote :
'When calling subprograms, the stack is not used; instead, the return address is stored in the register allocated for it.'
To be more precise, it is R14, but this register is free to use as the other general-purpose registers, when not receiving the return address after a Branch with Link (BL) instruction was used.
The coder can also decide to store the content of R14 somewhere in memory (ex : STR R14,Stored_Return_Address) to free this register and work with it in the sub procedure ...
To come back to the callant routine, one can use MOV PC,R14, but LDR PC,Stored_Return_Address will work too ;-)

[litwr]

Thank you.

[anonymous]

On Stardot BigEd made a mistake when he wrote
'Good point, and indeed there's an assembler error which tells you that it can't be done. But if a function isn't as big as 2k, then it can reach a constant just beyond the final return instruction.
'
12 bits allow to address 4095 bytes, and furthermore the encoding of LDR(B) or STR(B) shows that there is a sign bit, outside the 12 bit field. It means from any place in your code, you can reach a value relative to the address of the Program Counter (R15) ranging from -4095 to +4095. So that's an 8kbyte area, not 2 ;-)

[litwr]

So we have 13 bit signed offset but it means 12 bits for any direction.

[litwr]

Well, there is not a contiguous series of 13 bits in the instruction, and the official documentation
doesn't use the 13 bit signed terminology.
You can check the LDR(B), STR(B) format in the VL86C010 RISC FAMILY DATA MANUAL, the pdf is easy to find on the net, see on page 28 :
- there is an up/down bit (bit #23) to specify whether the offset is to be added or substracted from base register
- the 12 bit offset is in bits 0 to 11
This 12 bits can hold an unsigned value, but they can also hold a register number and a shift applied to it
(bit #25 determines whether the offset is an immediate value or a possibly shifted register).

[anonymous]

On stardot for your line drawing algorithm I'd say an additional improvement is, instead of
add r8,r7,r7 ;e2 = 2 * err
cmp r8,r4 ;e2 >= dy

to use instead
CMP R4,R7,LSL#1
and change the conditions of the instructions afterwards.

the cmp r8,r3 instruction will become CMP R3,R7,LSL#1 and again you will have to change the conditions of the instructions afterwards.

[litwr]

Thank you very much! It is sad for me that I missed the idea that e2 can be optimized.

[anonymous]

Hi there

[anonymous]

I have just send you some comment and it has been notified as 'spam' so please read it in your spam folder ;-(

[litwr]

I don't understand. Your comment about spam is here so where should I seek your other comment? LJ doesn't have any such folder.

[anonymous]

OK so here is the post again, shortened

[litwr]

Thank you very much for your information and a very good demo. I was an owner of A500 so I can't call it cripple, it was quite good for 1987 but Commodore as usual was reluctant to upgrade hardware and A1200 in 1992 was rather late.
BTW there is another Zarchos at youtube - https://www.youtube.com/user/zarchos