in ARMv6, why the value of PC is current instruction plus 8? - arm

what if the stages of pipeline is not 3, such as in ARM1156T2-S (also is ARMv6), it has 9 stages:
Fe1 Fe2 De Iss Fe3 Sh ALU Sat WBex
the PC is still address of current instruction plus 8?

For compatibility.
MOV LR, PC
LDR PC, =myfunc
You don't want to break all the old code just because the pipeline was changed.

Related

No "beq" or "bne" instruction in ARMv7-M manual?

I'm working on a STM32l475 micro-controller which runs a Cortex-M4 processor and ARM/Thumb instruction sets. I see (from objdump) that there are beq.n and bne.n instructions generated in the binary of an ARM program (I added the -mthumb flag when compile the program). However, I don't find these branch instructions in the latest ARMv7-M manual.
Can anyone tell me the reason? And what are the instructions available in the manual that are equivalent to these two branch instructions?
beq and bne are conditional branches; in other words, they are conditional versions of the unconditional branch b. eq and ne are two different condition codes; they are described in section A7.3. beq means branch if equal and bne means branch if not equal.
The b branch instruction has two different encodings in Thumb mode. The encoding you're seeing is probably encoding T1 described in section A7.7.12:
B<c> <label>
15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
1 1 0 1 [-<cond>--] [--------imm8---------]
In this encoding, the condition code (like eq or ne) is encoded directly into the instruction, in bits 8-11. The disassembly from objdump displays the condition code in place of <c> above. So using the condition code table in section A7.3, you would encode beq as 11010000[imm8].

decode ARM BL instruction

I'm just getting started with the ARM architecture on my Nucleo STM32F303RE, and I'm trying to understand how the instructions are encoded.
I have running a simple LED-blinking program, and the first few disassembled application instructions are:
08000188: push {lr}
0800018a: sub sp, #12
235 __initialize_hardware_early ();
0800018c: bl 0x80005b8 <__initialize_hardware_early>
These instructions resolve to the following in the hex file (displayed weird in Eclipse -- each 32-bit word is in MSB order, but Eclipse doesn't seem to know it... but that's for another topic):
address 0x08000188: B083B500 FA14F000
Using the ARM Architecture Ref Manual, I've confirmed the first 2 instructions, push (0xB500) and sub (0xB083). But I can't make any sense out of the "bl" instruction.
The hex instruction is 0xFA14F000. The Ref Manual says it breaks down like this:
31.28 27 26 25 24 23............0
cond 1 0 1 L signed_immed_24
The first "F" (0xF......) makes sense: all conditions are set (ALways).
The "A" doesn't make sense though, since the L bit should be set (1011). Shouldn't it be 0xFB......?
And the signed_immed_24 doesn't make sense, either. The ref manual says:
- start with 0x14F000
- sign extend to 30 bits (signed 2's-complement), giving 0x0014F000
- shift left to form 32-bit value, giving 0x0053C000
- add to the PC, which is the current instruction + 8, giving 0x0800018c + 8 + 0x0053C000, or 0x0853C194.
So I get a branch address of 0x0853C194, but the disassembly shows 0x080005B8.
What am I missing?
Thanks!
-Eric
bl is two, separate, 16 bit instructions. The armv5 (and older) ARM ARM does a better job of documenting them.
111HHoffset11
From the ARM ARM
The first Thumb instruction has H == 10 and supplies the high part of
the branch offset. This instruction sets up for the subroutine call
and is shared between the BL and BLX forms.
The second Thumb instruction has H == 11 (for BL) or H == 01 (for
BLX). It supplies the low part of the branch offset and causes the
subroutine call to take place.
0xFA14 0xF000
0xF000 is the first instruction upper offset is zeros
0xFA14 is the second instruction offset is 0x214
If starting at 0x0800018c then it is 0x0800018C + 4 + (0x0000214<<1) = 0x080005B8. The 4 is the two instructions head for the current PC. And the offset is units of (16 bit) instructions.
I guess the armv7-m ARM ARM covers it as well, but is harder to read, and apparently features were added. But they do not affect you with this branch link.
The ARMv5 ARM ARM does a better job of describing what happens as well. you can certaily take these two separate instructions and move them apart
.byte 0x00,0xF0
nop
nop
nop
nop
nop
.byte 0x14,0xFA
and it will branch to the same offset (relative to the second instruction). Maybe the broke that in some cores, but I know in some (after armv5) it works.

Encoding and decoding ARM instructions to/ from binary

Task 1: Write the corresponding ARM assembly representation for the following instructions:
11101001_000111000001000000010000
11100100_110100111000000000011001
10010010_111110100100000011111101
11100001_000000010010000011111010
00010001_101011101011011111001100
Task 2: Write the instruction code for the following instructions:
STMFA R13!, {R1, R3, R5-R11}
LDR R11, [R3, R5, LSL #2]
MOVMI R6, #1536
LDR R1, [R0, #4]!
EORS R3, R5, R10, RRX
I have zero experience with this material and the professor has left us students out to dry. Basically I've found the various methods for decoding these instructions but I have three major doubts still.
I don't have any idea on how to get started on decoding binary to ARM Instructions which is the first part of the homework.
I can't find some of these suffixes for example on EORS what is the S? Is it the set condition bit? Is it set to 1 when there is an S in front of the instruction?
I don't what to make of having multiple registers in one instruction line. Example:
EORS R3,R5,R10,RRx
I don't understand what's going on there with so many registers.
Any nudge in the right direction is greatly appreciated. Also I have searched the ARM manual, they're not very helpful for someone with no understanding of what they're looking for. They do have the majority of instructions for coding and decoding but have little explanation for the things I asked above.
If you have the ARM v7 A+R architecture manual (DDI0406C) there is a good table-based decode/disassembly description in chapter A5. You start at table A5.1 and and depending on the value of different bits in the instruction word it refers to more and more specific tables leading to the instruction.
As an example, consider the following instruction:
0001 0101 1001 1111 0000 0000 0000 1000
According to the first table it is an unsigned load/store instruction since the condition is not 1111 and op1 is 010. The encoding of this is further expanded in A5.3
From this section we see that A=0, op1=11001, Rn=1111 (PC), and B=0. This implies that the instruction is LDR(literal). Checking the page describing this instruction and remembering that cond=0001 we see that the instruction isLDRNE R0, [PC, #4].
To do the reverse procedure you look up the instruction in the alphabetical list of instructions and follow the pattern.
Looking at a different part of (one of) the ARM architectural reference manuals (not the cortex-m (armv6m armv7m) you want the ARMv5 one or the ARMv7-AR one) going to look at a thumb instruction, but the ARM instructions work the same way and are a couple chapters prior.
it says thumb instruction set as the section/chapter, then shortly into that is a table that shows thumb instruction set encoding or you can just search for that. one if them is called Add/subtract register, there are a lot of hardcoded ones and zeros up front then bit 9 is opc, then rm, rn and rd bits.
arm toolchains are easy to come by for windows mac and linux or can easily build from sources (just need binutils). assembling this
.thumb
add r1,r2,r3
add r1,r2,r4
add r1,r2,r5
add r1,r2,r6
add r1,r2,r7
then disassembling gives
00000000 <.text>:
0: 18d1 adds r1, r2, r3
2: 1911 adds r1, r2, r4
4: 1951 adds r1, r2, r5
6: 1991 adds r1, r2, r6
8: 19d1 adds r1, r2, r7
from that chart in the ARM ARM the add register starts with hardcoded bits 000110 the instructions above start with 0x18 or 0x19 which both start with the 6 bits 000110 (00011000 or 00011001). In the alphabetical list of thumb instructions we look for the add instructions. find the three register one and in this case it has 7 bits which happen to match the bits we are decoding 0001100 so we are on the right track. the the last 9 bits are three sets of three
0x1951 is 0001100101010001 or 0001100 101 010 001, the last nine represent r5, r2, and r1. Looking at the syntax part of the instruction it shows add rd, rn, rm but the machine code has rm, rn, rd so we take the machine code and rearrange per the syntax and get add r1,r2,r5. Not bad it matches, now unfortunately the s here is confusing, this thumb instruction doesnt have an s bit there wasnt room so this one always updates the flags, so the nature of this instruction and toolchain and how I used it requires the add without the s on the assembly end and disassembles with the s. confusing, sorry. when using arm instructions the letter s works as expected.
Just repeat this with the ARM instructions in either direction. The immediate values are going to be the most challenging part of this. Or not depends on the specific instruction and immediate encoding.

How to determine the endian mode the processor is running in?

How do I determine the endian mode the ARM processor is running in using only assembly language.
I can easily see the Thumb/ARM state reading bit 5 of the CPSR, but I don't know if there a corresponding bit in the CPSR or elsewhere for endianness.
;silly example trying to execute ARM code when I may be in Thumb mode....
MRS R0,CPSR
ANDS R0,#0x20
BNE ThumbModeIsActive
B ARMModeIsActive
I've got access to the ARM7TDMI data sheet, but this document does not tell me how to read the current state.
What assembly code do I use to determine the endianness?
Let's assume I'm using an ARM9 processor.
There is no CPSR bit for endianness in ARMv4 (ARM7TDMI) or ARMv5 (ARM9), so you need to use other means.
If your core implements system coprocessor 15, then you can check the bit 7 of the register 1:
MRC p15, 0, r0, c1, c0 ; CP15 register 1
TST r0, #0x80 ; check bit 7 (B)
BNE big_endian
B little_endian
However, the doc (ARM DDI 0100E) seems to hint that this bit is only valid for systems where the endianness is configurable at runtime. If it's set by the pin, the bit may be wrong. And, of course, on most(all?) ARM7 cores, the CP15 is not present.
There is a platform-independent way of checking the endianness which does not require any hardware bits. It goes something like this:
LDR R0, checkbytes
CMP R0, 0x12345678
BE big_endian
BNE little_endian
checkbytes
DB 0x12, 0x34, 0x56, 0x78
Depending on the current endianness, the load will produce either 0x12345678 or 0x78563412.
ARMv6 and later versions let you check CPSR bit E (9) for endianness.
Before ARMv6 co-processor 15 register c1 bit 7 should tell which endianness core is using.
In both cases 1 is big-endian while 0 is little-endian.

Where can I learn about optimal math in C on a micro-controller like, say, ARMv7?

I'm trying to optimize some functions and I realized that I know next to nothing about how long certain things take.
I can ask all the questions here, but I'd rather just find a good article on the subject if anyone knows one.
I'm using IAR to write a program in C for an ATMEL SAM7S processor. I have a sort function that takes 500uS or so, and I wanted to see if I could speed it up. I could also just post it here but I was hoping to learn for myself.
Like, is it any faster to subtract two 16 bit integers than it is to subtract two 32 bit integers? And how long does an operation like that take? Just one cycle or more? How long does multiplication take compared to subtraction?
Anyone know a place to look? I tried googling for some stuff but I couldn't come up with any useful search terms.
If anyone has an ideas on my specific function, I can post details. I'm basically trying to match two analog values to the closest index in a table of calibrated values. Right now I iterate through the whole table and use least squares to determine the closest match. Its pretty straightforward and I'm not sure there is a faster way without applying some extra logic to my table. But if I at least knew how long certain things took, I could probably optimize it myself.
is it any faster to subtract two 16 bit integers than it is to subtract two 32 bit integers?
Not on an ARM architecture which has native 32-bit registers, no.
Anyone know a place to look?
The canonical place for instruction cycle timings would be the Tech Ref Manual for the particular architecture your chip implements, eg. ARM7TDMI; timings for simple alu ops here and yes, it is one cycle. This is not friendly doc to be reading if you're not already well familiar with the instruction set, though...
Right now I iterate through the whole table
You'll be much better off looking at algorithmic optimisations here (eg indexing the table, sorting by one co-ordinate to narrow it down, etc) than worrying about instruction-level micro-optimisations.
A good first stage could be to study the assembly language of the architecture you are coding for.
After you should be able to read the binary file generated by your compiler and finally compare what the computer will really have to do with two different implementation.
You can use the timers in your SAM7S. Read a timer on start, and read it after N number of searches and subtract to get the difference. Try different algorithms and see what you see.
As far as 16 bit math vs 32 bit math, yes there can be a huge difference, but you have to look at your architecture. A subtract operation between two registers will take the same one clock be it 16 bit or 32 bit. But coming from C code eventually the variables may land in memory and you have to know if you have a 16 bit or 32 bit data bus (yes ARM7s can have a 16 bit bus, look at the GameBoy Advance, thumb code runs significantly faster than ARM code on that processor). Takes twice as many cycles to read or write 32 bit numbers on a 16 but bus. You likely do NOT have a 16 bit bus though. Using 16 bit variables on a 32 bit processor causes the processor to have to add extra instructions to strip or extend the upper bits so that the math is correct for a 16 bit variable. Those extra instructions can cause performance hits, a simple subtract which might have been say 3 or 4 instructions worst case might now be 5 or 6 and that is noticeable if it is in a tight loop. Generally you want to use variables that match the processors register size, on a 32 bit ARM use 32 bit variables as much as possible even if you are only counting to 10.
Hopefully I am understanding the problem you are trying to solve here, if not let me know and I will edit/remove this response:
Depending on how many bits in your measurement the typical solution for what you are doing is to use a look up table. So that I can show an example lets say you are taking a 4 bit measurement that you want to calibrate. Call it 0 to 15. Calibration of the sensor generated a list of data points, lets say:
raw cal
0x03 16
0x08 31
0x14 49
I assume what you are doing runtime is something like this, if the sensor reads a 0x5 you would look through the list looking for entries your sensor reading matches or is between two of the cal points.
searching you will find it to be between 0x03 and 0x08 to get the calibrated result from the raw 0x05 measurement
cal= (((0x05-0x03)/(0x08-0x03))*(31-16)+16 = 22
You have a divide in there which is a HUGE performance killer on most processors, ARM7 in particular as it doesnt have a divide. Not sure about the multiply but you want to avoid those like the plague as well. And if you think about how many instructions all of that takes.
Instead what you do is take the algorithm you are using run-time, and in an ad-hoc program generate all the possible outputs from all the possible inputs:
0 7
1 10
2 13
3 16
4 19
5 22
6 25
7 28
8 31
9 34
10 37
11 40
12 43
13 46
14 49
15 52
Now turn that into a table in your run-time code:
unsigned char cal_table[16]={7,10,13,16,19,22,25,28,31,34,37,40,43,46,49,52};
and then runtime
cal = cal_table[raw&15];
The code to implement this looks something like:
ldr r3, =cal_table
and r0, r0, #15
ldrb r0, [r3, r0]
takes like 5 clocks to execute.
Just the math to find cal from raw after you have searched through the table:
cal= (((raw-xlo)/(xhi-xlo))*(yhi-ylo)+ylo);
looks something like this:
docal:
stmfd sp!, {r3, r4, r5, lr}
ldr r3, .L2
ldr r5, .L2+4
ldr lr, .L2+8
ldr ip, [r5, #0]
ldr r0, [r3, #0]
ldr r1, [lr, #0]
ldr r2, .L2+12
rsb r0, ip, r0
rsb r1, ip, r1
ldr r5, [r2, #0]
bl __aeabi_uidiv
ldr r4, .L2+16
ldr r3, .L2+20
ldr r4, [r4, #0]
rsb r5, r4, r5
mla r4, r0, r5, r4
str r4, [r3, #0]
ldmfd sp!, {r3, r4, r5, pc}
And the divide function is as bad if not worse. The look up table should make your code run dozens of times faster.
The problem with look up tables is you trade memory for performance so you have to have a table big enough to cover all the possible inputs. A 12 bit sensor would give you as many as 4096 entries in the look up table for example. If say you knew the measurement would never be below 0x100 you could make the table 0x1000 - 0x100 or 3840 entries deep and subtract 0x100 from the raw value before looking it up, trading an couple of instructions at run time to save a few hundred bytes of memory.
If the table would be too big you could try some other tricks like make a look up table of the upper bits, and the output of that might be a pre-computed offset into the cal table to start your search. So if you had a 12 bit ADC, but didnt have room for a 4096 entry look up table you could make a 16 entry look up table, take the upper 4 bits of the ADC output and use it to look in the table. The table would contain the entry in the cal table to start searching. Say your cal table had these entries:
....
entry 27 raw = 0x598 cal = 1005
entry 28 raw = 0x634 cal = 1600
entry 29 raw = 0x6AB cal = 1800
entry 30 raw = 0x777 cal = 2000
your 16 deep look up table would then have these entries
...
[6] = 27;
[7] = 29;
...
And how you would use it is
start = lut[raw>>8];
for(i=start;i<cal_tab_len;i++)
{
...
}
instead of
for(i=0;i<cal_tabl_len;i++)
{
}
It could potentially greatly shorten the time it takes to find the entry in the table for you to perform the math needed.
For the particular problem of taking a raw value and turning that into a calibrated value at runtime, there are many many similar shortcuts. I dont know of one book that would cover them all. Which path to take has a lot to do with your processor, memory system and availability, and the size and nature of your data. You generally want to avoid divides in particular and multiples sometimes if your processor does not support them (using very few clock cycles). Most processors do not. (Yes, the one or two processors most programmers target, do have a single cycle multiply and divide). Even for processors that have a single cycle multiply and divide they often have to be wrapped with a C library to decide if it is safe to perform the operation with the hardware instruction or if it has to be synthesized with a library. I mentioned above that for most variables you want to match the native register size of the processor. If you have fixed point multiplies or divides you will often want to use half the register size of the processor. A 32 bit processor, unless you take the time to examine the instructions in detail, you probably want to limit your multiples to 16 bit inputs with a 32 bit output and divides to 32 bit inputs with a 16 bit output and hope the optimizer helps you out.
Again, If I have assumed incorrectly what the problem you were trying to solve is please comment and I will edit/modify this response.

Resources