Software Vault: The Gold Collection/Software Vault - The Gold Collection (American Databankers) (1993).ISO/cdr47/asmlib30.zip/FLOAT.DOC

Text File | 1993-04-08 | 21KB | 777 lines

***********************  FLOATING-POINT DATA  *******************************

ASMLIB floating-point data subroutines (c) copyright 1991 Douglas Herr
all rights reserved

ASMLIB recognizes two floating point formats: IEEE-standard single-
precision, and IEEE-standard double-precision.  Single-precision
floating point numbers are 4 bytes long, and are referred to as float4
or simply f4, while double-precision floating point numbers are 8 bytes
long, which I describe as float8 or f8.



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

ADDF4:     add two float4 values
Source:    addf4.asm

80x87 not required

Call with: DS:[SI] pointing to value0, ES:[DI] pointing to value1
           both values must be float4 values in IEEE format
Returns:   if CF = 0, value0 is replaced by the sum
           if CF = 1, the sum would result in an overflow; neither
           value is changed
Uses:      CF

Example:

include asm.inc

extrn addf4:proc

.data
value0  dd 123.456
value1  dd -67.4

.code
; program fragment assumes DS:@data
         .
         .
         .
        push    ds
        pop     es           ; ES = DS
        assume  es:@data
        lea     si,value0
        lea     di,value1
        call    addf4

; note that this can be used for subtraction by changing the sign
; of one of the values
;
; changing the sign is trivial: if DS:[SI] points to a float4 value,
; the sign is changed with
;
;      xor    byte ptr 3[si],10000000b


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

ADDF8:     add two float8 values
Source:    addf8.asm

80x87 not required

Call with: DS:[SI] pointing to value0, ES:[DI] pointing to value1
           both values must be float8 values in IEEE format
Returns:   if CF = 0, value0 is replaced by the sum
           if CF = 1, the sum would result in an overflow; neither
           value is changed
Uses:      CF

Example:

include asm.inc

extrn addf8:proc

.data

value0  dq 123.456
value1  dq -67.4

.code
; program fragment assumes DS:@data
         .
         .
         .
        push    ds
        pop     es           ; ES = DS
        assume  es:@data
        lea     si,value0
        lea     di,value1
        call    addf8

; note that this can be used for subtraction by changing the sign
; of one of the values
;
; changing the sign is trivial: if DS:[SI] points to a float8 value,
; the sign is changed with
;
;      xor    byte ptr 7[si],10000000b



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

CMPF4:      compare two float4 values
Source:     cmpf4.asm

CMPF8:      compare two float8 values
Source:     cmpf8.asm

80x87 NOT required

Call with:  DS:[SI] pointing to value 0
            ES:[DI] pointing to value 1
Returns:    if ZF = 1, values are equal
            if SF = 1, value 1 is larger
Uses:       zero flag (ZF), sign flag (SF)
Example:

include  asm.inc

extrn    cmpf4:proc

.data
v0       dd 12.345
v1       dd 17.04

.code
; program fragment assumes DS:@data
         .
         .
         .
         push   ds
         pop    es               ; in this example, both values are in DGROUP
         assume es:@data
         lea    si,v0            ; point DS:[SI] to value 0
         lea    di,v1            ; point ES:[DI] to value 1
         call   cmpf4            ; compare
         js     value1           ; value 1 is larger if SF = 1
         je     equal            ; both float4 values are identical if ZF = 1
                                 ; otherwise value 0 is larger


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

DIVI2:      divides integer in AX by integer in DX, returning float4 quotient
Source:     divi2.asm (i2tof4.asm, invf4.asm, mulf4.asm)

80x87 not required

Call with:  AX, DX = I2 values; DX <> 0
            DS:[BX] pointing to 4-byte buffer for quotient
            DivI2 does not check for divide-by-zero errors
            make sure DX <> 0 and you'll be OK
Returns:    quotient at DS:[BX], f4 format
Uses:       nothing; all flags and registers are saved
Example:

include asm.inc

extrn   divi2:proc

.data
f4data  dd ?                     ; 4 bytes for resulting quotient

.code
; program fragment assumes DS:@data
        .
        .
        .
        mov    ax,17             ; I want to calculate 17/47
        mov    dx,47             ; don't ask me why
        lea    bx,f4data         ; point to 4-byte data area
        call   divi2


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

F4TOF8:     converts a float4 value to float8 format
Source:     f4tof8.asm

80x87 not required

Call with:  DS:[SI] pointing to float4 value
            ES:[DI] pointing to 8-byte buffer for float8 value
Returns:    CF = 0 if successful
            CF = 1 if the float4 value is invalid
Uses:       CF; all other flags and registers are saved
Example:

include asm.inc

extrn  f4tof8:proc

.data
float4      dd 123.456
float8      dq ?

.code
; program fragment assumes DS:@data
             .
             .
             .
            push    ds
            pop     es           ; ES = DS
            assume  es:@data
            lea     si,float4           ; DS:[SI] points to float4 value
            lea     di,float8           ; ES:[DI] points to float8 buffer
            call    f4tof8
            jc      oops                ; go to error handler if problem




░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

F4TOI2:     copies the integer portion of a float4 value to AX
Source:     f4toi2.asm

80x87 not required

Note:       only the integer portion of the value is converted; decimals
            are truncated
Call with:  DS:[SI] pointing to a float4 value
Returns:    if CF = 0, AX = integer value
            if CF = 1, float4 value is too large (if positive)
                       or is too small (if negtive)
Uses:       AX, CF; all other flags and registers are saved
Example:

include asm.inc

extrn   f4toi2:proc

.data
float4  dd  1234.567

.code
; program fragment assumes DS:@data
        .
        .
        .
        lea  si,float4
        call f4toi2
        jc   oops                     ; gotta fix something if error



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

F4TOI4:     copies the long integer portion of a float4 value to DX:AX
Source:     f4toi4.asm

80x87 not required

Note:       only the integer portion of the value is converted; decimals
            are truncated
Call with:  DS:[SI] pointing to a float4 value
Returns:    if CF = 0, DX:AX = long integer value
            if CF = 1, float4 value is too large (if positive)
                       or is too small (if negtive)
Uses:       AX, DX, CF; all other flags and registers are saved
Example:

include asm.inc

extrn   f4toi4:proc

.data
float4  dd  1234.567

.code
; program fragment assumes DS:@data
        .
        .
        .
        lea    si,float4
        call   f4toi4
        jc     oops                   ; gotta fix something if error




░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

F8TOF4:     converts a float8 value to float4 format
Source:     f8tof4.asm

80x87 not required

Call with:  DS:[SI] pointing to float8 value
            ES:[DI] pointing to 4-byte buffer for float4 value
Returns:    CF = 0 if no error
            CF = 1 float8 number overflowed to infinity
                   or underflowed to zero
Uses:       CF; all other flags and registers are saved
Example:

include asm.inc

extrn  f8tof4:proc

.data
float8      dq 123.456
float4      dd ?

.code
; program fragment assumes DS:@data
             .
             .
             .
            push    ds
            pop     es           ; ES = DS
            assume  es:@data
            lea     si,float8           ; DS:[SI] points to float8 value
            lea     di,float4           ; ES:[DI] points to float4 buffer
            call    f8tof4
            jc      oops                ; go to error handler if problem




░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

I2TOF4:     convert integer value to float4 IEEE format
Source:     i2tof4.asm

80x87 not required

Call with:  DS:[BX] pointing to 4-byte buffer for float4 value
            AX = signed integer value to convert
Returns:    converted value in float4 IEEE format at DS:[BX]
Uses:       nothing; all registers and flags are saved
Example:

include asm.inc

extrn   i2tof4:proc

.data
float4  dd ?                   ; initial value undefined

.code
; program fragment assumes DS:@data
        .
        .
        .
        mov    ax,12345        ; signed integer value in AX
        lea    bx,float4       ; this is where I want the resulting value
        call   i2tof4



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

I4TOF4:     convert long integer value to float4 IEEE format
Source:     i4tof4.asm

80x87 not required

Call with:  DS:[BX] pointing to 4-byte buffer for float4 value
            DX:AX = signed long integer value to convert
                    (high word in DX)
Returns:    converted value in float4 IEEE format at DS:[BX]
Uses:       nothing; all registers and flags are saved
Example:

include asm.inc

extrn   i4tof4:proc

.data
float4  dd ?                          ; initial value undefined
longint dd -123456789

.code
; program fragment assumes DS:@data
        .
        .
        .
        mov    ax,word ptr longint    ; low word of signed long integer
        mov    dx,word ptr longint+2  ; high word of signed long integer
        lea    bx,float4              ; point to buffer for converted value
        call   i4tof4



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

INVF4:      calculate the inverse of a float4 value
Source:     invf4.asm

80x87 not required

Call with:  DS:[SI] = address of float4 value
Returns:    DS:[SI] = address of 1/value
            previous value is overwritten
Uses:       nothing
Example:

include asm.inc

extrn   invf4:proc

.code
; program fragment assumes DS:@data
        .
        .
        .
        lea   si,float4          ; point to float4 value
        call  invf4              ; original value replaced with inverse



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

MAXF4, MAXF4b: find maximum value in single-precision real-number array
Source:        small & medium models: maxf4.asm
               huge model: maxf4.asm (lowDS2hi.asm)

MINF4, MINF4b: find minimum value in single-precision real-number array
Source:        small & medium models: minf4.asm
               huge model: minf4.asm (lowDS2hi.asm)

80x87 not required

Call with:  ES:[DI] pointing to array element at start of search
            CX = number of array elements to search
            For maxf4b or minf4b, call with BX = byte increment between
            array elements.  Maxf4 and minf4 assume byte increment = 4.
Returns:    AX = array element number with maximum or minimum value
            note that the offset of that value is DI + (AX shl 2).
            With maxf4b and minf4b, the offset of the value is
            DI + (AX * BX).
            AX = -1 if maxf4 was called with CX = 0
Uses:       AX
Example:

include asm.inc

extrn  maxf4:proc

.data
floatdata   dd 1500 dup(0)

.code
; program fragment assumes DS:@data
; program provides values for the array
       .
       .
       .
       push   ds
       pop    es
       assume es:@data
       lea    di,floatdata      ; ES:[DI] points to the data array
       mov    cx,1500           ; search entire array
       call   maxf4             ; returns with AX = array element



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

MAXF8, MAXF8b: find maximum value in double-precision real-number array
Source:        maxf8.asm

MINF8, MINF8b: find minimum value in double-precision real-number array
Source:        minf8.asm

80x87 not required

Call with:  ES:[DI] pointing to array element at start of search
            CX = number of array elements to search
            For maxf8b or minf8b, call with BX = byte increment between
            array elements.  Maxf8 and minf8 assume byte increment = 8.
Returns:    AX = array element number with maximum or minimum value
            note that the offset of that value is DI + (AX shl 3)
            With maxf8b and minf8b, the offset of the value is
            DI + (AX * BX).
            AX = -1 if maxf8 was called with CX = 0
Uses:       AX
Example:

include asm.inc

extrn  maxf8:proc

.data
floatdata   dq 1500 dup(0)

.code
; program fragment assumes DS:@data
; program provides values for the array
       .
       .
       .
       push   ds
       pop    es
       assume es:@data
       lea    di,floatdata      ; ES:[DI] points to the data array
       mov    cx,1500           ; search entire array
       call   maxf8             ; returns with AX = array element
       mov    cl,3
       shl    ax,cl
       add    di,ax             ; ES:[DI] points to maximum value



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

MULF4:      multiplies two float4 values
Source:     mulf4.asm

80x87 not required

Call with:  DS:[SI] pointing to value0, ES:[DI] pointing to value1
Returns:    if CF = 0, no error; product replaces value0 at DS:[SI]
            if CF = 1, the product is either zero or infinity (value0 is lost)
Uses:       CF; all other flags and registers are saved
Example:

include asm.inc

extrn  mulf4:proc

.data
value0   dd 1.234
value1   dd 0.5

.code
; program fragment assumes DS:@data
         .
         .
         .
         push   ds
         pop    es
         assume es:@data
         lea    si,value0
         lea    di,value1
         call   mulf4


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

MULF4TEN:   multiply a float4 value by 10
Source:     mulf4ten.asm (mulf4.asm)

80x87 not required

Call with:  DS:[SI] pointing to float4 value
            MulF4Ten is front-end code for MulF4, which saves ES and DI,
            points ES:[DI] to f4 data equalling 10.0, calls MulF4, then
            restores ES and DI.
Returns:    if CF = 0, float4 value at DS:[SI] multiplied by 10.0
            if CF = 1, the product is either 0 or infinity
Uses:       CF; all other flags and registers are saved
Example:

include asm.inc

extrn  mulf4ten:proc

.data
value0   dd 1.234

.code
; program fragment assumes DS:@data
         .
         .
         .
         lea    si,value0
         call   mulf4ten


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

MULF8:      multiplies two float8 values
Source:     mulf8.asm (mpmul.asm)

80x87 not required

Call with:  DS:[SI] pointing to value0, ES:[DI] pointing to value1
Returns:    if CF = 0, no error; product replaces value0 at DS:[SI]
            if CF = 1, the product is either zero or infinity (value0 is lost)
Uses:       CF; all other flags and registers are saved
Example:

include asm.inc

extrn  mulf8:proc

.data
value0   dq 1.2347349
value1   dq 0.5

.code
; program fragment assumes DS:@data
         .
         .
         .
         push   ds
         pop    es
         assume es:@data
         lea    si,value0
         lea    di,value1
         call   mulf8


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

MULF8TEN:   multiply a float8 value by 10
Source:     mulf8ten.asm (mulf8.asm, mpmul.asm)

80x87 not required

Call with:  DS:[SI] pointing to float8 value
            MulF8Ten is front-end code for MulF8, which saves ES and DI,
            points ES:[DI] to f8 data equalling 10.0, calls MulF8, then
            restores ES and DI.
Returns:    if CF = 0, float8 value at DS:[SI] multiplied by 10.0
            if CF = 1, the product is either 0 or infinity
Uses:       CF; all other flags and registers are saved
Example:

include asm.inc

extrn  mulf8ten:proc

.data
value0   dq 1.234

.code
; program fragment assumes DS:@data
         .
         .
         .
         lea    si,value0
         call   mulf8ten


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

SCALEF4:    scale a float4 value by an integer power of 2
Source:     scalef4.asm

80x87 not required

Call with:  DS:[SI] pointing to the float4 data
            AL = scale factor (may be positive or negative)
Returns:    CF = 0 if success: float4 value was properly scaled
            CF = 1 if error: resulting float4 value would have overflowed
                   float4 value unchanged in this case
Uses:       CF; all other flags and all registers saved
Example:

include asm.inc

.code
; program fragment assumes DS:@data
        .
        .
        .
        lea   si,float4data
        mov   al,3             ; scale by (2^3) effectively multiplies by 8
        call  scalef4
        jc    oops             ; take remedial action if error


░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

SCALEF8:    scale a float8 value by an integer power of 2
Source:     scalef8.asm

80x87 not required

Call with:  DS:[SI] pointing to the float8 data
            AL = scale factor (may be positive or negative)
Returns:    CF = 0 if success: float8 value was properly scaled
            CF = 1 if error: resulting float8 value would have overflowed
                   float8 value unchanged in this case
Uses:       CF; all other flags and all registers saved
Example:

include asm.inc

.code
; program fragment assumes DS:@data
        .
        .
        .
        lea   si,float8data
        mov   al,3             ; scale by (2^3) effectively multiplies by 8
        call  scalef8
        jc    oops             ; take remedial action if error



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

SORTF4HI:    sort an array of single-precision real numbers, highest first
Source:      sortf4hi.asm

SORTF4LO:    sort an array of single-precision real numbers, lowest first
Source:      sortf4lo.asm

80x87 not required

Call with:  ES:[DI] pointing to the first of the array elements to be
            sorted, CX = number of array elements
Returns:    nothing
Uses:       nothing; all registers and flags are saved
Example:

include asm.inc

extrn  sortf4hi:proc

.data
floatdata   dd 1500 dup(0)

.code
; program fragment assumes DS:@data
; program provides values for the array
        .
        .
        push    ds
        pop     es
        assume  es:@data
        lea     di,floatdata
        mov     cx,1500           ; sort entire array
        call    sortf4hi          ; highest value first



░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

SUBF4:       subtract one float4 value from another
Source:      subf4.asm (addf4.asm)

SUBF8:       subtract one float8 value from another
Source:      subf8.asm (addf8.asm)

80x87 not required

Call with:   DS:[SI] pointing to value0
             ES:[DI] pointing to value1
             the number at ES:[DI] is subtracted from the number at DS:[SI]
             with the result replacing the number at DS:[SI]
Returns:     if CF = 0, DS:[SI] points to value0 - value1
             if CF = 1, the operation would result in an overflow;
                neither value is changed
Uses:        nothing; all registers and flags are saved
Example:

include asm.inc

extrn subf4:proc

.data
value0  dd 123.456
value1  dd -67.4

.code
; program fragment assumes DS:@data
         .
         .
         .
        push    ds
        pop     es           ; ES = DS
        assume  es:@data
        lea     si,value0
        lea     di,value1
        call    subf4