Re: May the numbers speak - supplement

On Thu, 12 Jun 2025 22:01:07 +0000, Paul Rubin wrote:

mhx@iae.nl (mhx) writes:

iSPICE> TEST TEST TEST TEST
\ dtimescan   : 2515 clock ticks elapsed, 45296
\ timestrscan : 419 clock ticks elapsed, 45296
\ HMS         : 419 clock ticks elapsed, 45296
It makes no difference. Whatever is holding it down must
be quite severe.

>
Is dtimescan the one that uses double word arithmetic for 16 bit
processors?  It's doing more stuff, I would think.  Weird that it
catches up after a few tries.  Cache warming?

The TICKS? word apparently can't be trusted under Windows 11. This
is better:
#2000000 VALUE #iters
CREATE num ," 12:34:56"
: .SECS  MS? #1000000 #iters 3 * */ ." / " . ." ns per iteration." ;
: TEST2 ( --  )
  CR
  CR ." \ dtimescan   : " TIMER-RESET num C@+
     #iters 0 ?DO  2DUP dtimescan   2DROP
                   2DUP dtimescan   2DROP
                   2DUP dtimescan   2DROP  LOOP dtimescan        UD. SECS
  CR ." \ timestrscan : " TIMER-RESET num C@+
     #iters 0 ?DO  2DUP timestrscan 2DROP
                   2DUP timestrscan 2DROP
                   2DUP timestrscan 2DROP  LOOP timestrscan      UD. SECS
  CR ." \ HMS         : " TIMER-RESET num C@+
     #iters 0 ?DO  2DUP hms          DROP
                   2DUP hms          DROP
                   2DUP hms          DROP  LOOP hms                . SECS
  CR ." \ HMS2        : " TIMER-RESET num C@+
     #iters 0 ?DO  2DUP hms2        3DROP
                   2DUP hms2        3DROP
                   2DUP hms2        3DROP  LOOP hms2  ROT . SWAP . . SECS ;
CLS TEST2 TEST2 TEST2 TEST2
\ dtimescan   : 45296 / 15 ns per iteration.
\ timestrscan : 45296 / 13 ns per iteration.
\ HMS         : 45296 / 5 ns per iteration.
\ HMS2        : 12 34 56 / 53 ns per iteration.
\ dtimescan   : 45296 / 15 ns per iteration.
\ timestrscan : 45296 / 14 ns per iteration.
\ HMS         : 45296 / 5 ns per iteration.
\ HMS2        : 12 34 56 / 53 ns per iteration.
\ dtimescan   : 45296 / 16 ns per iteration.
\ timestrscan : 45296 / 14 ns per iteration.
\ HMS         : 45296 / 5 ns per iteration.
\ HMS2        : 12 34 56 / 53 ns per iteration.
\ dtimescan   : 45296 / 15 ns per iteration.
\ timestrscan : 45296 / 14 ns per iteration.
\ HMS         : 45296 / 5 ns per iteration.
\ HMS2        : 12 34 56 / 53 ns per iteration. ok
HMS is three times faster than  dtimescan  and  timestrscan,  while
HMS2 is 3.5 times slower (as expected).
In practice, I'd use some variant of hms2.
-marcel