1. Embedding machine code
I wanted to use Juergen's 'bit.e' library for doing 32-bit manipulations
because he uses machine code to achieve get over Euphoria's lack of
32-bit integers.
However, the code doesn't work because it has lines like this ...
SHIFT_LEFT = define_c_func("", SHL_SPACE, {C_INT, C_INT}, C_INT)
but that gives the error message ...
c:\euphoria\include\dll.e:50 in function define_c_func()
type_check failure, lib is {}
Any ideas about how I should actually call these machine code routines?
What I really need is when I have an atom that exceeds a 32-bit format I
want to chop off the excess bits and only keep the low 32 bits.
--
Derek
2. Re: Embedding machine code
- Posted by akusaya at gmx.net
Feb 26, 2004
If I remember correctly you need to use eu 2.4 to use that...
D> I wanted to use Juergen's 'bit.e' library for doing 32-bit manipulations
D> because he uses machine code to achieve get over Euphoria's lack of
D> 32-bit integers.
D> However, the code doesn't work because it has lines like this ...
D> SHIFT_LEFT = define_c_func("", SHL_SPACE, {C_INT, C_INT}, C_INT)
D> but that gives the error message ...
D> c:\euphoria\include\dll.e:50 in function define_c_func()
D> type_check failure, lib is {}
D> Any ideas about how I should actually call these machine code routines?
D> What I really need is when I have an atom that exceeds a 32-bit format I
D> want to chop off the excess bits and only keep the low 32 bits.
3. Re: Embedding machine code
On Thu, 26 Feb 2004 08:31:20 +0000, Derek Parnell
<ddparnell at bigpond.com> wrote:
> SHIFT_LEFT = define_c_func("", SHL_SPACE, {C_INT, C_INT}, C_INT)
>
>but that gives the error message ...
>
>c:\euphoria\include\dll.e:50 in function define_c_func()
>type_check failure, lib is {}
>
>
>Any ideas about how I should actually call these machine code routines?
You need to use 2.4. Check that define_c_func in dll.e begins with
object lib not atom lib.
>What I really need is when I have an atom that exceeds a 32-bit format I
>want to chop off the excess bits and only keep the low 32 bits.
If speed isn't critical, this might show how, up to a point:
include machine.e
atom i
i=#0FFFFFFF
for j=1 to 40 do
i=i*2+1
? i
? int_to_bits(i,32)
? bits_to_int(int_to_bits(i,32))
if getc(0) then end if
end for
It seems to work up to 54 bits, I guess it depends on where you are
getting these atoms from. You may need to change the sign of the
result from bits_to_int if it is over 2^31-1.
Regards,
Pete
4. Re: Embedding machine code
Derek wrote:
> I wanted to use Juergen's 'bit.e' library for doing 32-bit manipulations
> because he uses machine code to achieve get over Euphoria's lack of
> 32-bit integers.
Thanks. 
> However, the code doesn't work because it has lines like this ...
>
>
> SHIFT_LEFT = define_c_func("", SHL_SPACE, {C_INT, C_INT}, C_INT)
>
> but that gives the error message ...
>
> c:\euphoria\include\dll.e:50 in function define_c_func()
> type_check failure, lib is {}
As others already wrote, 'bit.e' only works with Eu 2.4 or later, sorry. 
> Any ideas about how I should actually call these machine code routines?
Tommy Carlier wrote some sample code, that does something similar, using
the old way to define and call ASM functions:
http://www.listfilter.com/cgi-bin/esearch.exu?fromMonth=C&fromYear=8&toMonth=C&toYear=8&postedBy=Tommy+Carlier&keywords=rotate_left
> What I really need is when I have an atom that exceeds a 32-bit format I
> want to chop off the excess bits and only keep the low 32 bits.
Tommy uses poke4() to put the parameters to the machine code. Does
poke4() automatically only keep the low 32 bits? I don't know.
If you want, I can make a different version of 'bit.e', that also runs
on older Euphoria versions. Do you need all 9 functions in 'bit.e', or
only some of them?
Regards,
Juergen
5. Re: Embedding machine code
----- Original Message -----
From: "Juergen Luethje" <j.lue at gmx.de>
To: <EUforum at topica.com>
Subject: Re: Embedding machine code
>
>
> Derek wrote:
>
> > I wanted to use Juergen's 'bit.e' library for doing 32-bit manipulations
> > because he uses machine code to achieve get over Euphoria's lack of
> > 32-bit integers.
>
> Thanks.
>
> > However, the code doesn't work because it has lines like this ...
> >
> >
> > SHIFT_LEFT = define_c_func("", SHL_SPACE, {C_INT, C_INT}, C_INT)
> >
> > but that gives the error message ...
> >
> > c:\euphoria\include\dll.e:50 in function define_c_func()
> > type_check failure, lib is {}
>
> As others already wrote, 'bit.e' only works with Eu 2.4 or later, sorry.
Oooops! I've forgotten to change my 'EUDIR' environment symbol back to point to
the 2.4 version. Silly me!
I'm trying to implement the TEA (Tiny Encryption Algorithm -
http://www.ftp.cl.cam.ac.uk/ftp/papers/djw-rmn/djw-rmn-tea.html ) in Euphoria.
This is a very simple algo but it works on packets of 4-bytes using shifts, xors,
adds, and mults so I very quickly get bit overflows. I'm wanting to use machine
code to improve the speed as I'll be using it for real-time encoding/decoding of
traffic over a LAN.
--
Derek
