Re: [clock] behaves just slow (Re: [clock scan] behaves inconsistently in Tcl8.5 (was: Re: [clock scan day] behaves differentyl in Tcl8.5 than in Tcl8.4))

Kaitzschu wrote:
> % time {clock seconds} 1000
> 84 microseconds per iteration
> % time {clock scan "+1 day"} 1000
> 858 microseconds per iteration
> % time {clock format [clock scan "+1 day"]} 1000
> 1571 microseconds per iteration
> % info patchlevel
> 8.4.9
>
> % time {clock seconds} 100
> 475.35 microseconds per iteration
> % time {clock add [clock seconds] +1 day} 100
> 40774.31 microseconds per iteration
> % time {clock format [clock add [clock seconds] +1 day]} 100
> 117802.13 microseconds per iteration
> % info patchlevel
> 8.5a3

Using the latest sources, things appear to not be so bad as that:

% info patch
8.4.11
% time {clock seconds} 1000
3.471 microseconds per iteration
% time {clock scan "+1 day"} 1000
27.222 microseconds per iteration
% time {clock format [clock scan "+1 day"]} 1000
54.106 microseconds per iteration

% info patch
8.5a4
% time {clock seconds} 1000
6.163 microseconds per iteration
% catch {clock add}
1
% time {clock add [clock seconds] +1 day} 1000
898.758 microseconds per iteration
% time {clock format [clock add [clock seconds] +1 day]} 1000
2277.155 microseconds per iteration

And there's still the entire beta period of Tcl 8.5 where performance
and bug fix improvements can happen. While it's near certain we won't
get back to 8.4 performance levels, there may be room for more
improvement.

Please file a bug report on the matter, so that maintainers will not
overlook your performance concerns.

--
| Don Porter Mathematical and Computational Sciences Division |
| donald.porter@xxxxxxxx Information Technology Laboratory |
| http://math.nist.gov/~DPorter/ NIST |
|______________________________________________________________________|
.

Relevant Pages