[retroforth] Re: ztype in assembler
- From: "Helmar Wodtke" <helmwo@xxxxxx>
- To: retroforth@xxxxxxxxxxxxx
- Date: Wed, 26 Jan 2005 22:58:16 +0100
retroforth@xxxxxxxxxxxxx schrieb am 26.01.05 22:49:27:
>
>
> On Wed, January 26, 2005 13:30, Helmar Wodtke said:
> > retroforth@xxxxxxxxxxxxx schrieb am 26.01.05 22:20:57:
> > With a shorter loop:
> >
> >> ; ----------------------------------------------------------------------
> >> code 'ztype',ztype
> >> push ebx ;
> >> push ecx ;
> >> upop ebx ;
> >> lea ecx, byte [ebx - 1]
> >> .a: inc ecx
> >> cmp byte [ecx], 0
> >> jnz .a
> >> sub ecx,ebx
> >> jmp type.a
> >> ; next
> >
> > Ron, expert :) Faster?
> >
>
> Dunno. I have a 'ztc' word which takes a zstr and returns a forth
> string/count
>
> Zstring ztc type
>
> ; ( zstr -- zstr len )
> code 'ztc', _ztc
> push eax ; save ptr
> mov edi, eax
> xor eax, eax
> mov ecx, -1
> repne scasb
> pop eax
> sub edi, eax
> dec edi
> upsh edi
> next
>
> I haven't tried to make it smaller or faster yet...
I dont know too... I've heard rumors that block-instructions are on some
processors much slower than explicit operations (I dont know why it should be
so...) Well, I would prefer your solution of having a "ztc": an explicit
"ztype" could help only to avoid some calling overhead, but the most overhead
should be produced by "emit".
Bis dann,
Helmar
helmwo@xxxxxx
-- Binary/unsupported file stripped by Ecartis --
-- Type: application/x-pkcs7-signature
-- File: smime.p7s
-- Desc: S/MIME Cryptographic Signature
Other related posts:
