Linux nasm square root algorithm poblem

Posted on 2004-04-07
Medium Priority
Last Modified: 2008-02-01
I am programming with the FPU under linux and the NASM assembler and linking using gcc

Here is the algorithm I am implementing:

take in x from the keyboard

 Initialize x_new = x
     x_old = x_new;
     x_new = (x_old + x / x_old) / 2;
   while |x_old - x_new|/x_new > e

I am having the problem with my compare and my conditional jump.
I have tested the values of the algorthm throughout and the values are correct to what I worked out on paper.
extern  printf
extern  scanf
global main
segment .data
prompt: db "Please enter a number: ",0
format: db "%lf", 0
output: db "%lf",10, 0
epsilon dq  0.0001
x2:     dq  2.0
segment .bss
x:      resq 1
x_new:  resq 1
x_old:   resq 1
x_temp:  resq 1
segment .text
        push prompt
        call printf
        add esp, 4
        push x
        push format
        call scanf
        add esp, 8
        fld qword [x]               ;fstack: x
        fstp qword [x_new]      ;fstack:
                                          ;x_new = x
.L1:    finit
        fld qword [x_new]       ;fstack: x_new
        fstp qword [x_old]       ;fstack:
                                         ;x_old = x_new
        finit                            ;clear the stack
                                         ;now lets make x_new = (x_old + (x/x_old))/2
        fld qword [x]              ;fstack: x
        fdivr qword [x_old]     ;fstack: x/x_old
        fadd qword [x_old]      ;fstack:(x_old + (x/x_old))
        fdiv qword [x2]           ;fstack:(x_old + (x/x_old))/2
        fstp qword [x_new]      ;fstack:
        finit                            ;clear the stack
                                          ;now lets make |x_old - x_new|/x_new
        fld qword [x_old]         ;fstack: x_old
        fsub qword [x_new]     ;fstack: x_old - x_new
        fabs                            ;fstack: |x_old - x_new|
        fdiv qword [x_new]      ;fstack: |x_old - x_new|/x_new
        fld qword [epsilon]     ;fstack: epsilon, |x_old - x_new|/x_new
        fcomi st1                       ;compare ST0 with ST1
        jl .L1
        push dword [x_new + 4]   ;outputting x_new after loop
        push dword [x_new]
        push output
        call printf
        add esp, 12

        mov eax, 0
Question by:SaumZ
Welcome to Experts Exchange

Add your voice to the tech community where 5M+ people just like you are talking about what matters.

  • Help others & share knowledge
  • Earn cash & points
  • Learn & ask questions
  • 3
  • 2

Expert Comment

ID: 10784584
I believe you have a reason for not using FSQRT instead.

Well, by definition "jl" jumps if "SF <> OF".
However, fcomi sets the following flags: ZF, PF, CF.
So change jl to jc. Voila.

Author Comment

ID: 10784909
I tried the jump with jc, and my code is still only doing one iteration. and outputting for example:

Please enter a number: 20

This number is correct, but only after one iteration.
this is the output that is at the bottom. of my code, where it outputs x_new

I also tested the number that I have from |x_old - x_new|/x_new
by putting it in a temp variable and outputting it and comparing to what I calculated using pencil and paper, and that is correct also.

Maybe the problem lies somewhere else? I am not sure at this point.

The reason for not just using fsqrt is that I am just learning the fpu programming, and I am trying to get myself used to programming in it by using different instructions with it.  

So, I may have logic errors someplace else in my code, but to me, it all looks correct.

Expert Comment

ID: 10786189
"jnc", sorry.

Accepted Solution

joghurt earned 500 total points
ID: 10786259
Well, another approach. There were too many finit's even for the first sight so I've tried to reorganize the things a bit.

fld         qword ptr [x_new]
fstp        qword ptr [x_old]
fld         qword ptr [x]
fdiv        qword ptr [x_old]
fadd        qword ptr [x_old]
fdiv        qword ptr [x2]
fstp        qword ptr [x_new]
fld         qword ptr [x_old]
fsub        qword ptr [x_new]
fdiv        qword ptr [x_new]
fcomp       qword ptr [epsilon]
fnstsw      ax  
test        ah,41h
je          .L1

I haven't used fcomi because it won't run on older machines. (And because my development environment couldn't compile P6 instructions. :-)
But you can give it a try.

Author Comment

ID: 10810023
My whole problem in the task was a fdivr in place of a fdiv in my origional code...
I will give you the points, thank you for your help.

Featured Post

Free Tool: Subnet Calculator

The subnet calculator helps you design networks by taking an IP address and network mask and returning information such as network, broadcast address, and host range.

One of a set of tools we're offering as a way of saying thank you for being a part of the community.

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

New style of hardware planning for Microsoft Exchange server.
We are witnesses that everyone is saying that our children shouldn't "play" with a technology because it is dangerous. This article is going to prove that they are wrong.
In this video we outline the Physical Segments view of NetCrunch network monitor. By following this brief how-to video, you will be able to learn how NetCrunch visualizes your network, how granular is the information collected, as well as where to f…
In this video, Percona Solution Engineer Dimitri Vanoverbeke discusses why you want to use at least three nodes in a database cluster. To discuss how Percona Consulting can help with your design and architecture needs for your database and infras…
Suggested Courses
Course of the Month14 days, 1 hour left to enroll

801 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question