Re: Fortran decimal anyone?
From: David Frank (dave_frank_at_hotmail.com)
Date: 05/14/04
- Next message: aggie: "Data conversion (from real to character)"
- Previous message: Gerry Thomas: "Re: Searching zeros of complex function"
- In reply to: Mike Cowlishaw: "Re: Fortran decimal anyone?"
- Next in thread: Dave Thompson: "Re: Fortran decimal anyone?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Fri, 14 May 2004 12:18:15 GMT
"Mike Cowlishaw" <mfc@uk.ibm.com> wrote in message
news:c825v0$c6c$1@news.btv.ibm.com...
> David Frank wrote:
>> Mike,
>> re: your Telco Million call benchmark
>>
>> I will chime in for my fav language Fortran,
>> perhaps you can supply missing info in below scorecard?
>>
>> 1. Fortran 25 statements 3 sec to execute (David Frank)
>> [etc.]
>
> Hi .. yes, by using integers and manual scaling, it is easy to
> write a special-purpose piece of code for this that is very
> short and fast.
>
> But that's missing the point of the benchmark The benchmark
> is intended to estimate the overhead of the calculations when
> they are done in a straightforward and maintainable way.
>
> The actual total speed is uninteresting; as someone pointed
> out, the full application would be more complex and work
> on much larger quantities of data. What the benchmark
> does appear to do is match the behaviour of those large
> real applications in that one respect.
>
> Mike
>
re: straightforward and maintainable way..
That of course, is in the eye of the beholder.
Looking at your TELCO.C file below
You use 2 function statements each time a multiply is performed,
PLUS statements to invoke rounding required.
PLUS functions have 4 arguments, thats more straightforward?
PLUS none of your languages support overloading of the = + - * /
operators vs. fortran's ability to do p = time*brate syntax..
A plethora of keystrokes in order to get the most basic math done
is apt to be LESS-straightforward/maintainable in my view.
TELCO.c
<snip over 100 statements to get to the compute loop>
/* Start of the by-number loop */
for (numbers=0; ; numbers++) {
// get next 8-byte packed decimal number
readlen=fread(inpack, 1, INNUMLEN, inp);
if (readlen<INNUMLEN) { // error or EOF
frc=ferror(inp); // check for error
// [note: ferror is single thread, but only this thread knows INP]
if (frc!=0) printf("Error: cannot read file '%s'", filein);
break; // [no message if EOF]
}
if (calc!=0) {
calltype=((unsigned)inpack[INNUMLEN-1]>>4) & 0x01; // last bit
// have packed decimal number; convert to working format
decPackedToNumber(inpack, INNUMLEN, &inscale, &n);
if (calltype==0) // p=r[c]*n
decNumberMultiply(&p, &baserate, &n, &set);
else
decNumberMultiply(&p, &distrate, &n, &set);
set.round=DEC_ROUND_HALF_EVEN; // round prices
decNumberRescale(&p, &p, &mtwo, &set); // to x.xx
if (tax!=0) {
set.round=DEC_ROUND_DOWN; // truncate taxes
decNumberMultiply(&b, &p, &basetax, &set); // b=p*0.0675
decNumberRescale(&b, &b, &mtwo, &set); // to x.xx
decNumberAdd(&sumB, &sumB, &b, &set); // sumB=sumB+b
decNumberAdd(&t, &p, &b, &set); // t=p+b
if (calltype!=0) {
decNumberMultiply(&d, &p, &disttax, &set); // d=p*0.0341
decNumberRescale(&d, &d, &mtwo, &set); // to x.xx
decNumberAdd(&sumD, &sumD, &d, &set); // sumD=sumD+d
decNumberAdd(&t, &t, &d, &set); // t=t+d
}
} else {
t=p;
}
decNumberAdd(&sumT, &sumT, &t, &set); // sumT=sumT+t
decNumberToString(&t, string);
// printf("[%d] %d: %s\n", calltype, numbers, string);
} else {
strcpy(string, "0.77");
}
// [Adding CRLF is part of I/O, as other languages have WriteLine]
strcat(string, "\r\n"); // Add CRLF
len=strlen(string);
urc=fwrite(string, 1, len, oup);
if (urc!=len) {
printf("Error: file '%s' could not be written", fileou);
break;
}
- Next message: aggie: "Data conversion (from real to character)"
- Previous message: Gerry Thomas: "Re: Searching zeros of complex function"
- In reply to: Mike Cowlishaw: "Re: Fortran decimal anyone?"
- Next in thread: Dave Thompson: "Re: Fortran decimal anyone?"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
