18 Aug
2014
18 Aug
'14
7:11 p.m.
A long time ago ... in a land far away :)
I did some assembler programming on the Acorn Archimedes (ARM 2/3) and worked
out a series of additions and subtractions that would perform very fast
multiplication of awkward numbers by known amounts.
Is there any point in doing this for C programs, or are modern compilers
sophisticated enough to do such things themselves?
If anyone is interested *7 is:
RSB R1, R0, R0, LSL #3
--
Will J Godfrey
http://www.musically.me.uk
Say you have a poem and I have a tune.
Exchange them and we can both have a poem, a tune, and a song.
18 Aug
18 Aug
7:19 p.m.
On 08/18/2014 07:11 PM, Will Godfrey wrote:
A long time ago ... in a land far away :)
I did some assembler programming on the Acorn Archimedes (ARM 2/3) and worked
out a series of additions and subtractions that would perform very fast
multiplication of awkward numbers by known amounts.
Is there any point in doing this for C programs, or are modern compilers
sophisticated enough to do such things themselves?
If anyone is interested *7 is:
RSB R1, R0, R0, LSL #3
gcc 4.8 for ARM seems to give that instruction too with -O2 here for
a simple function that returns *7 of its input argument.
With no optimization:
00000000 <mul7>:
0: e52db004 push {fp} ; (str fp, [sp, #-4]!)
4: e28db000 add fp, sp, #0
8: e24dd00c sub sp, sp, #12
c: e50b0008 str r0, [fp, #-8]
10: e51b2008 ldr r2, [fp, #-8]
14: e1a03002 mov r3, r2
18: e1a03183 lsl r3, r3, #3
1c: e0623003 rsb r3, r2, r3
20: e1a00003 mov r0, r3
24: e24bd000 sub sp, fp, #0
28: e49db004 pop {fp} ; (ldr fp, [sp], #4)
2c: e12fff1e bx lr
8:36 p.m.
On Mon, Aug 18, 2014 at 06:11:33PM +0100, Will Godfrey wrote:
A long time ago ... in a land far away :)
I did some assembler programming on the Acorn Archimedes (ARM 2/3) and worked
out a series of additions and subtractions that would perform very fast
multiplication of awkward numbers by known amounts.
Is there any point in doing this for C programs, or are modern compilers
sophisticated enough to do such things themselves?
If anyone is interested *7 is:
RSB R1, R0, R0, LSL #3
:-) Memories... I love ARM assembly...
Modern compilers will do such things. If they exploit other
features of the ARM (such as all instructions being conditional)
I don't know, but it's very well possible.
Ciao,
--
FA
A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)
19 Aug
19 Aug
9:09 a.m.
On Mon, 18 Aug 2014 18:36:35 +0000
Fons Adriaensen <fons(a)linuxaudio.org> wrote:
On Mon, Aug 18, 2014 at 06:11:33PM +0100, Will Godfrey
wrote:
Thanks for the responses guys. I think it's a case of let the compiler do the
work. Oh, and yes, I found ARM assembly a dream to work with, although I don't
know what it's like these days. There have been quite a lot of extensions I
believe.
--
Will J Godfrey
http://www.musically.me.uk
Say you have a poem and I have a tune.
Exchange them and we can both have a poem, a tune, and a song.
A long time ago ... in a land far away :)
I did some assembler programming on the Acorn Archimedes (ARM 2/3) and worked
out a series of additions and subtractions that would perform very fast
multiplication of awkward numbers by known amounts.
Is there any point in doing this for C programs, or are modern compilers
sophisticated enough to do such things themselves?
If anyone is interested *7 is:
RSB R1, R0, R0, LSL #3
:-) Memories... I love ARM assembly...
Modern compilers will do such things. If they exploit other
features of the ARM (such as all instructions being conditional)
I don't know, but it's very well possible.
Ciao,
3744
days inactive
3745
days old
linux-audio-dev@lists.linuxaudio.org
3 comments
3 participants
participants (3)
-
Cedric Roux -
Fons Adriaensen -
Will Godfrey