Question about gforth speed : Forth

Hello,

during the last days i have been using gForth under Linux, and i have liked it a lot. One thing i liked is the concept of "primitive centric threaded code". So i have made a prototype of a forth virtual machine using this concept in assembly language for i386 Linux.

I expected my code to be faster than gForth because gForth is written in C, but to my surprise, it turns that gForth is even faster:

: prog 100000000 for i drop next ;
(3 seconds in assembly prototype)
: prog 100000000 for i drop next ;
(1.5 seconds in gForth)

I have no clue about what's happening here, does someone know what could be happening?

I show the assembly code for NEXT, DOCOLON, SEMICOLON:


macro NEXT: lodsd
jump eax

docol: lodsd
xchg esp,ebp
push esi
xchg esp,ebp
mov esi,eax
NEXT
semicol: xchg esp,ebp
pop esi
xchg esp,ebp
NEXT