Bitmap management for ARM-7

I haven't found any suitable log-like assembly instruction... so I've wrote this. Are you up to you to make it better? Maybe it could be improved by reading...


.code 32
.text
.align 0

  .global dw_bmp_msb_pos
dw_bmp_msb_pos:
  movs r1, r0
  moveq r0, #0
  bxeq lr
  mov r0, #33
getmsb_cycle:
  sub r0, r0, #1
  movs r1, r1, lsl #1
  bxcs lr
  b getmsb_cycle

  .global dw_bmp_set
dw_bmp_set:
  ldr r2, [r0, #0]
  mov r3, #1
  mov r3, r3, lsl r1
  orr r2, r2, r3
  str r2, [r0, #0]
  bx lr

  .global dw_bmp_reset
dw_bmp_reset:
  ldr r2, [r0, #0]
  mov r3, #1
  mov r3, r3, lsl r1
  bic r2, r2, r3
  str r2, [r0, #0]
  bx lr

Not even x86 assembly is that ugly.

OP here.

Actually this is ARM7 assembly. It's really awesome except it's very difficult to understand at the beginning... nothing you can't do with a good reference!

The big stuff about it is the optimization phase, since this assembly is very expressive, although IMHO building an optimized compiler is not a trivial task.

I've tryied to write the same semantics with C and then I've checked out the assembly produced by gcc... it's much more cumbersome!

>>3
Did you forget to set -fomg-optimized?

>>3
Okay. You're an idiot.

OP speaking.

>>3
I've tried all levels of optimization. The point is that C doesn't have a statement that can test the overflow flag...

>>5
This should be proved. Why do you say that?

>>2
How is x86 assembly in any way ugly? Get the fuck out.

x86 > ARM

>>1
moveq r0, #0
Terrible!
x86 wins¹. Also, a binary search can be more -Omg-optimized.

mov r3, r3, lsl r1
orr r2, r2, r3
orr r2, r2, r3, lsl r1
Same for bic.
Why do you expect the bitmap to be in memory?

¹ http://siyobik.info/index.php?module=x86&id=20

>>9
[m]A6.7.24 CLZ
Count Leading Zeros returns the number of binary zero bits before the first binary one bit in a value.

Encoding T1 ARMv6T2, ARMv7
CLZ<c> <Rd>,<Rm>
151413121110 9 8 7 6 5 4 3 2 1 0 151413121110 9 8 7 6 5 4 3 2 1 0
1 1 1 1 1 0 1 0 1 0 1 1 Rm       1 1 1 1 Rd      1 0 0 0 Rm

if !Consistent(Rm) then UNPREDICTABLE;
d = UInt(Rd);  m = UInt(Rm);
if BadReg(d) || BadReg(m) then UNPREDICTABLE;
 
Assembler syntax
CLZ<c><q>  <Rd>, <Rm>
where:
<c><q> See Standard assembler syntax fields on page A6-7.
<Rd> Specifies the destination register.
<Rm> Specifies the register that contains the operand. Its number must be encoded twice in
encoding T1, in both the Rm and Rm2 fields.
Operation
if ConditionPassed() then
    EncodingSpecificOperations();
    result = CountLeadingZeroBits(R);
    R[d] = result<31:0>;
Exceptions
None.

>>10
Sigh, why are ARM embedding BBCode in their documentation?

Also, IIRC there's some clever way to do this in HAKMEM with bitops.

>>10
Architecture version
Version 5 and above.¹
Surpisingly, ARM7 is architecture version 3².

¹ http://www.arm.com/miscPDFs/14128.pdf
² http://en.wikipedia.org/wiki/ARM_architecture#ARM_cores

>>12
No-one uses ARM7 anymore, and the ARM7TDMI is architecture version 4T. Obviously.

No-one uses ARM7 anymore
>>1 does.

>>14
He doesn't. He just doesn't know his ass from a hole in the ground.

18 Name: Anonymous 2011-02-03 6:28

Don't change these.

Name:		Email:
Entire Thread Thread List

All trademarks and copyrights on this site are owned by their respective parties. All comments and posts are the responsibility of their own posters.
- Tinychan + 2ch Mode -