Simple floating point arithmetic for simple-forth

Here is our latest incarnation of simple-forth, a forth interpreter written in C.
This time we add floating point arithmetic capabilities. There are two basic ways to incorporate floating point capabilities. One is using the same integer stack and the other is using a separate foating point stack.



We use a separate floating point stack with maximum 32 elements. in our newest version shown below for simplicity. The built-in integer stack is 32 bits wide. Floating point numbers in our interpreter are 64 bit wide. The current stack element index is indicated by FP. The basic floating point operations are f+, f-, f*, f/ these Forth words are the equivalent floating point words corresponding to the built-in basic + - * / operator words



:

/*
file     simple-forth-0.0.3.c
author   Dr. Ernesto P. Adorio
         UPDEPP (University of the Philippines,
         Extension Program in Pampanga
         Clarkfield, Pampanga
email    ernesto.adorio@gmail.com
version  0.0.1 January 14, 2012 basic interpreter.
         0.0.2 January 16, 2012 interactivity added.
         0.0.3 January 25, 2012 basic floating point.
*/

#include 
#include 
#include 
#include 



enum {E_STACKUNDERFLOW, E_FSTACKUNDERFLOW} ERRCODES;


enum {L_EMIT, L_DOT,L_DOTT, L_SPACE, L_ADD, L_SUB, L_MUL, L_DIV, L_DROP, L_INT32, L_CR,
      L_FNUM, L_FPLUS, L_FSUB, L_FMUL, L_FDIV, L_FDOT, L_FDOTT, L_FDROP,
      L_BYE, L_EOS,L_ERROR} OPCODES;

const char *stdwords[] = { "emit", ".",  ".t", "space", "+", "-", "*", "/", "drop", "int32", "cr",
                           "fnum", "f+", "f-", "f*" ,"f/", "f.","f.t","fdrop",
                           "bye",
                           "\n",
                         };
// "int32" and "fnum" are dummies. 


#define MAXSTKLEN 32
#define MAXFSTKLEN 32
#define MAXTOKENLEN 128

int  LENSTDWORDS = sizeof(stdwords)/sizeof(stdwords[0]);
char LINEBUFFER[256];
char *goodbye= "bye";
int  ERRCODE = 0;
int SP = -1;   /* stack pointer index */
int stack[MAXSTKLEN];

int  opcode;
char *tokstart;
char *tokend;

int FP = -1;   /* floating point stack index */
double fstack[MAXFSTKLEN];


int numbertype(char *s) 
{
   /*
   Returns 
     0 - not a number!
     1 - an integer
     2 - a floating point number
   */
   
   char *t = s;
   while (isspace(*t)) t++;
   if (*t == '\0') return 0;
    
   if (*t == '+' || *t == '-') t++;
   if (*t == '\0') return 0;
   while (isdigit(*t)) t++;
   if (*t == '\0') {
     return 1; // an integer!
   }
  
   if (*t == '.') t++;
   while (isdigit(*t)) t++;
   if (*t == '\0') return 2;
   
   if (*t == 'e') t++;
   if (*t == '\0') return 0; // error!
   if (*t == '+' || *t == '-') t++;
   if (*t == '\0') return 0; // error!
   while (isdigit(*t)) t++;
   if (*t == '\0') return 2; // a floating point number!
   return 0; // not an integer or floating point.
}





int eval(int opcode)
{
    /* Evaluate opcode */
    switch (opcode){
    case L_EMIT: 
      //@@printf("opcode L_EMIT %s %d\n", tokstart, SP);
      if (SP < 0) {
     ERRCODE = E_STACKUNDERFLOW;
      } else {
        printf ("%c", stack[SP--]);
      }
      break;

    case L_DOT:
      printf("opcode L_EMIT %s %d\n", tokstart, SP);
      if (SP < 0) {
     ERRCODE = E_STACKUNDERFLOW;
     printf("%s", "underflow!");
      } else {
        printf ("%d", stack[SP--]);
      }
      break;

    case L_DOTT:
      //@@printf("opcode L_EMIT %s %d\n", tokstart, SP);
      if (SP < 0) {
     ERRCODE = E_STACKUNDERFLOW;
     printf("%s", "underflow!");
      } else {
        printf ("%d", stack[SP]);
      }
      break;
      
    case L_SPACE:
      printf ("%c", 32);
      break;
      
    case L_ADD:
      //printf("opcode L_ADD %s\n", tokstart);
      if (SP <= 0) {
     ERRCODE = E_STACKUNDERFLOW;
     printf("%s", "underflow!");
     break;
      } 
      stack[SP-1] += stack[SP];
      SP--;
      break;
      
    case L_SUB:
      //printf("opcode L_SUB %s\n", tokstart);
      if (SP <= 0) {
     ERRCODE = E_STACKUNDERFLOW;
     printf("%s", "underflow!");
     break;
      } 
      stack[SP-1] -= stack[SP];
      SP--;
      break;
      
    case L_MUL:
      //printf("opcode L_MUL %s\n", tokstart);
      if (SP <= 0) {
     ERRCODE = E_STACKUNDERFLOW;
     printf("%s", "underflow!");
     break;
      } 
      stack[SP-1] *= stack[SP];
      SP--;
      break;
      
    case L_DIV:
      //printf("opcode L_DIV %s\n", tokstart);
      if (SP <= 0) {
     ERRCODE = E_STACKUNDERFLOW;
     printf("%s", "underflow!");
     break;
      } 
      stack[SP-1] /= stack[SP];
      SP--;
      break;
  
   case L_INT32:
      //printf("opcode L_INT32 %s\n",tokstart);
      stack[++SP]= atoi(tokstart);
      // printf("@@@ pushing an integer!");
      break;
      
    case L_DROP:
      //printf("opcode L_DROP %s \n", tokstart);
      SP=SP -1;
      break;
      
    case L_CR:
      // printf("opcode L_CR %s \n", tokstart);
      printf("\n");
      break;


/* simple floating point features */
    case L_FNUM:
      //printf("floating point number %s\n",tokstart);
      fstack[++FP]= atof(tokstart);
      break;
    
    case L_FPLUS:
      if (FP <= 0) {
     ERRCODE = E_FSTACKUNDERFLOW;
     printf("%s", "fstack underflow!");
     break;
      } 
      fstack[FP-1] += fstack[FP];
      FP--;
      break;

    case L_FSUB:
      if (FP <= 0) {
     ERRCODE = E_FSTACKUNDERFLOW;
     printf("%s", "fstack underflow!");
     break;
      } 
      fstack[FP-1] -= fstack[FP];
      FP--;
      break;

    case L_FMUL:
      if (FP <= 0) {
     ERRCODE = E_FSTACKUNDERFLOW;
     printf("%s", "fstack underflow!");
     break;
      } 
      fstack[FP-1] *= fstack[FP];
      FP--;
      break;

    case L_FDIV:
      if (FP <= 0) {
     ERRCODE = E_FSTACKUNDERFLOW;
     printf("%s", "fstack underflow!");
     break;
      } 
      fstack[FP-1] /= fstack[FP];
      FP--;
      break;

    case L_FDOT:
      printf("opcode FDOT %s\n", tokstart);
      if (FP < 0) {
     ERRCODE = E_FSTACKUNDERFLOW;
     printf("%s", "underflow!");
      } else {
        printf ("%f", fstack[FP--]);
      }
      break;

    case L_FDOTT:
      //@@printf("opcode L_EMIT %s %d\n", tokstart, SP);
      if (SP < 0) {
     ERRCODE = E_FSTACKUNDERFLOW;
     printf("%s", "underflow!");
      } else {
        printf ("%f", fstack[FP]);
      }
      break;
      
    case L_BYE:
      printf("Terminating... ");
      break;
    default:
      ;
      break;
    }  
   
}

void nexttoken()
{   
    tokstart = tokend;
    //@ printf("inside nexttoken [%s]", tokstart);
    
    /* ignore leading white spaces */
    while (isspace(*tokstart)) tokstart++;

    /* find terminating space of end of string */
    tokend = tokstart;
    while (!isspace(*tokend) && *tokend != '\0') tokend++;
    
    /* string terminator */
    if (*tokend != '\0') {
      *tokend = '\0';
      tokend ++;
    } 

    /* get opcode */
    opcode = -1; 
    if (tokend == tokstart) {
       opcode = L_EOS;
       return;
    }

    //@@@ printf("token [%s]\n", tokstart); 
    for (int i =0; i < LENSTDWORDS; i++) {
      if (strcmp(stdwords[i], tokstart) == 0){
 opcode = i;
        break;
      }
    }
    if (opcode != -1) {
       return;
    }

    int ntype = numbertype(tokstart); 
    if (ntype == 1){
      /* is this a number ?*/
      opcode = L_INT32;
    } else if (ntype == 2) {
      opcode = L_FNUM;
    } else if (opcode == -1){
      opcode = L_ERROR;
    };
}



int main(){
  while (1) {
    printf("> ");
    if (fgets(LINEBUFFER, 250, stdin)!= NULL) {
      tokend=LINEBUFFER;
    } else {
      tokend= goodbye;
    }
    while (1) {
      nexttoken();
      eval(opcode);
      if (opcode == L_BYE) {
 return 0;
      }
      if (opcode ==L_EOS) {
 break;
      }
      if (ERRCODE != 0) {
 printf("ERRORCODE [%d]", ERRCODE);
 break;
      };
    }
  }
  printf("\n");
  return 0;
}

Here is compilation and a simple execution run.

gcc simple-forth-0.0.3.c -g -std=c99 -o simple-forth
toto@toto-Aspire-4520:~/Blogs/my-other-life-as-programmer/forth$ ./simple-forth
> 23.34 -56.12 f* f. 
-1309.840800> bye
Terminating... 

Dont forget: a floating point number has a comma or an exponent in our simple-forth implementation.

Does a number string represents an integer or a floating value in C?

I got the shock of my blogging life when I cannot access an old post about the evils of the C function atof. Here is a code which return 1 for integers and 2 for floating point and 0 if not a valid integer or a valid float.

/*
author  Ernesto P. Adorio
        UPDEPP (University of the 
        Philippines, Extension Program
        in Pampanga, Clarkfield, Pampanga
email   ernesto. adorio@ gmail.com [remove spaces]
version 0.0.1 january 14, 2012
*/

#include 
#include 


int numbertype(char *s) 
{
   char *t = s;
   while (isspace(*t)) t++;
   if (*t == '\0') return 0;
    
   // number sign
   printf("ltrimmed [%s]", t);
   if (*t == '+' || *t == '-') t++;
   if (*t == '\0') return 0;
   printf("after sign [%s]", t);
   while (isdigit(*t)) t++;
   printf("after digits [%s]", t);
   if (*t == '\0') return 1; // an integer!
   
   // floating point??
   printf ("decimal point? [%s] ", t);  
   if (*t == '.') t++;
   printf ("fractional digits?");
   while (isdigit(*t)) t++;
   if (*t == '\0') return 2;
   
   // exponent part.
   printf ("testing exponent part %s", t); 
   if (*t == 'e') t++;
   printf ("after 'e' ");  
   if (*t == '\0') return 0; // error!
   printf ("sign after e?", t);  
   if (*t == '+' || *t == '-') t++;
   printf("%s", t);  
   if (*t == '\0') return 0; // error!
   while (isdigit(*t)) t++;
   if (*t == '\0') return 2; // a floating point number!
   return 0; // not an integer or floating point.
}

int main() 
{
  printf("numtype =[%d]", numbertype("-123.34")); 
};

Remove the deubgging printf statements when you use it for applications.

When the program is run, it returns a code of 2 to denote a floating point number, a 1 for an integer, and a 0 if not a number.

A basic Forth language interpreter (non-interactive)

Last time we showed a simple tokenizer for a Forth interpreter. We find it easier to combine the ideas in that post to create a simple Forth engine which process a simple string.

We will expound further on this until we successfully build a compiler for a non-standard Forth language.

Have fun working on this. The current version works on simple strings but specified on the source code itself. We will introduce interactivity and expand the capabilities of this simple Forth interpreter.

/*
file     simple-forth.c
author   Dr. Ernesto P. Adorio
         UPDEPP (University of the Philippines,
         Extension Program in Pampanga
         Clarkfield, Pampanga
email    ernesto.adorio@gmail.com
version  0.0.1 January 14, 2012
*/

#include 
#include 
#include 
#include 


#define MAXSTKLEN 32
#define MAXTOKENLEN 128

enum {L_EMIT, L_ADD, L_SUB, L_MUL, L_DIV, L_POP, L_INT32, L_CR, L_ERROR} OPCODES;
const char *stdwords[] = { 
".", 
"+",
"-",
"*",
"/",
"pop",
"int32",
"cr",
};


int LENSTDWORDS = 8;

int stack[MAXSTKLEN];
int SP = 0;   /* stack pointer index */

int main(){
  char tokens[] = "   123 34 + . cr 567 -3456 * . cr";
  char *tokstart = tokens;
  char *tokend = tokstart;
  int  opcode;

  /* @@@ printf("tokstart: %s\n", tokstart);*/

  tokend = tokens;

  while (1) {
    tokstart = tokend;
    /* ignore leading white spaces */
    while (isspace(*tokstart)) tokstart++;

    /* find terminating space of end of string */
    tokend = tokstart;
    while (!isspace(*tokend) && *tokend != '\0') tokend++;
    
    /* string terminator */
    if (*tokend != '\0') {
      *tokend = '\0';
      tokend ++;
    } else {
      break;
    } 

    /* get opcode */
    opcode = -1; 
    /*@@@ printf("token [%s]\n", tokstart); */
    for (int i =0; i < LENSTDWORDS; i++) {
      if (strcmp(stdwords[i], tokstart) == 0){
 opcode = i;
        break;
      }
    }
    if (opcode == -1){
      opcode = L_INT32;
    }


    /* Evaluate opcode */
    switch (opcode){
    case L_EMIT: 
      //printf("opcode L_EMIT %s \n", tokstart);
      printf ("%d", stack[SP--] );
      break;
    case L_ADD:
      //printf("opcode L_ADD %s\n", tokstart);
      stack[SP-1] += stack[SP];
      SP--;
      break;
    case L_SUB:
      //printf("opcode L_SUB %s\n", tokstart);
      stack[SP-1] += stack[SP];
      SP--;
      break;
    case L_MUL:
      //printf("opcode L_MUL %s\n", tokstart);
      stack[SP-1] *= stack[SP];
      SP--;
      break;
    case L_DIV:
      //printf("opcode L_DIV %s\n", tokstart);
      stack[SP-1] /= stack[SP];
      SP--;
      break;
    case L_INT32:
      //printf("opcode L_INT32 %s\n",tokstart);
      stack[++SP]= atoi(tokstart);
      break;
    case L_POP:
      //printf("opcode L_POP %s \n", tokstart);
      SP--;
      break;
    case L_CR:
      // printf("opcode L_CR %s \n", tokstart);
      printf("\n");
    }  
  }
  printf("\n");
  return 0;
}

Save the code to a file simple-forth.c, then compile using the following command line on the directory where the source file is stored. gcc simple-forth.c -std=c99 -o forth It should compile cleanly, then execute the executible by issuing ./forth When the program runs, it prints out

toto@toto-Aspire-4520:~/Blogs/my-other-life-as-programmer/forth$ ./forth
157
-1959552

Dennis Ritchie (September 9, 1941 – October 8, 2011: inventor of C programming language

One of my regrets in life is not having to meet in person Dennis Ritichie, the developer of the C programmng language.



I will never forget the time I have to understand pointers (I suspect my instructor did not understand much too!). C is like more a friendlier assembly language in English! C++, Java may be more "in" these days, but they owe their success to the pioneering path blazed by C.

A return to C programming: Hello world example

I took up C programming at Camp Aguinaldo in the mid 80's and got a certificate for it. but I never earned a centavo for my C programming skills in my later work since the oppurtunity to program was non-existent, instead I opted to enrol in an MS Applied Math, Major in Computer Science after work.

C may be considered a higher level assembly language as the power constructs in assembly are also available in C; concepts such as pointers, unrestricted gotos, and the like. In fact a C compiler worth its salt should be able to generate assembly output!

For the hello world example in C, we try to write in an undiciplined manner and see if we can get away it. As usual we write a function to do the actual writing of a message "hello world!" to be closer in spirit to the previous Javascript example. Save the following to a file "hello.c" without the quotes.

void showMessage(char *str)
{
   puts(str);
}

main()
{
    showMessage("Hello World");
}

Every C program has a main function. The difference between a function and procedure is that the former exlicitly returns a value while a procedure does not and should be declared of type void. Now lets try to run the hello.c code in the console: type

gcc -Wall hello.c

Notice the -Wall flag to output any warnings encountered in processing the source file.
Here is the output from my Ubuntu 10.04 terminal (console):


gcc -Wall hello.c
hello.c: In function ‘showMessage’:
hello.c:4: warning: implicit declaration of function ‘puts’
hello.c: At top level:
hello.c:8: warning: return type defaults to ‘int’
hello.c: In function ‘main’:
hello.c:10: warning: control reaches end of non-void function
toto@toto-laptop:~/Blogs/my-other-life-as-programmer/C$


In good C programs, all functions must be declared and return types should be explicit. Here the compiler complains that puts was implicitly declared and that the main function did not return any value! and yet in spite of these warnings, you will find that the gcc outputte an executible file hello. To run this ececutible, issue ./hello. The program outputs


toto@toto-laptop:~/Blogs/my-other-life-as-programmer/C$ ./a.out
Hello World
toto@toto-laptop:~/Blogs/my-other-life-as-programmer/C$


Now a good C programmer tries to remove all warnings. So lets do some modifications to be in good terms with the gcc compiler.

#include 

void showMessage(char *str)
{
   puts(str);
}

int main() # explicit declaration.
{
    showMessage("Hello World");
    return 0;
}

We have included the standard header "stdio.h", gcc knows where to find it (in "/usr/include") and this erases the warning on puts. We also explicitly declare the main function as type int and specifically return a zero (the usual value for successful execution.) The compilation is now successful without any complaints from the compiler.


toto@toto-laptop:~/Blogs/my-other-life-as-programmer/C$ gcc -Wall hello.c
toto@toto-laptop:~/Blogs/my-other-life-as-programmer/C$


So the minor corrections removed all warnings. Now lets do some extra activities for fun.

1. Create assembler listing
Type gcc -S hello.c this outputs hello.s

        .file   "hello.c"

        .text
.globl showMessage
        .type   showMessage, @function
showMessage:
.LFB0:
        .cfi_startproc
        pushq   %rbp
        .cfi_def_cfa_offset 16
        movq    %rsp, %rbp
        .cfi_offset 6, -16
        .cfi_def_cfa_register 6
        subq    $16, %rsp
        movq    %rdi, -8(%rbp)
        movq    -8(%rbp), %rax
        movq    %rax, %rdi
        call    puts
        leave
        ret
        .cfi_endproc
.LFE0:
        .size   showMessage, .-showMessage
        .section        .rodata
.LC0:
        .string "Hello World"
        .text
.globl main
        .type   main, @function
main:
.LFB1:
        .cfi_startproc
        pushq   %rbp
        .cfi_def_cfa_offset 16
        movq    %rsp, %rbp
        .cfi_offset 6, -16
        .cfi_def_cfa_register 6
        movl    $.LC0, %edi
        call    showMessage
        movl    $0, %eax
        leave
        ret
        .cfi_endproc
.LFE1:
        .size   main, .-main
        .ident  "GCC: (Ubuntu/Linaro 4.4.4-14ubuntu5) 4.4.5"
        .section        .note.GNU-stack,"",@progbits

The assembler file is one reason most programmers (run away) do not like assembly language! ha ha. Unless of course if the paying job calls for using assmbly (or assembler) langauge. ! But this feature of the gcc to ouput assembler will come in handy when we program a Forth language translator.
2. Seek help in using gcc
Try man gcc. You will get an eyeful to the useful flags. Better yet, try the free forums or buy a book specifically on the gcc.