8:37 a.m.
Hi,
this is weird:
<hansl>
s0 = NA
print s0
a = isnan(s0) # gives 0!?
</hansl>
Latest snapshot on Win8.
BTW, gretl now feels extremely sluggish, and the Windows task manager
shows it to consume very little memory (<30MB). I will try to downgrade
to earlier snapshots or the latest release to check this.
thanks,
sven
4:53 p.m.
Am 25.11.2014 um 15:59 schrieb Allin Cottrell:
On Tue, 25 Nov 2014, Sven Schreiber wrote:
> this is weird:
>
> <hansl>
> s0 = NA
> print s0
> a = isnan(s0) # gives 0!?
> </hansl>
s0 is NA, it's not NaN.
ah, right -- after a long-ago discussion about NA and NaN I had sort of
made a note to myself that gretl doesn't really distinguish between the
two, but obviously that wasn't the whole truth. Plain old missing(s0)
works fine here.
thanks,
sven
10:44 a.m.
Am 25.11.2014 um 14:37 schrieb Sven Schreiber:
Latest snapshot on Win8.
BTW, gretl now feels extremely sluggish, and the Windows task manager
shows it to consume very little memory (<30MB). I will try to downgrade
to earlier snapshots or the latest release to check this.
Well it's certainly as fast as ever on a Win7 machine... (with memory at
about 40MB using more or less the same dataset). Probably some
idiosyncratic thing going on with that 32bit-Win8-machine.
-s
4 p.m.
On Wed, 26 Nov 2014, Sven Schreiber wrote:
Am 25.11.2014 um 14:37 schrieb Sven Schreiber:
>
> Latest snapshot on Win8.
>
> BTW, gretl now feels extremely sluggish, and the Windows task manager
> shows it to consume very little memory (<30MB). I will try to downgrade
> to earlier snapshots or the latest release to check this.
>
Well it's certainly as fast as ever on a Win7 machine... (with memory at
about 40MB using more or less the same dataset). Probably some
idiosyncratic thing going on with that 32bit-Win8-machine.
Thanks for reporting back. I'm having difficulty thinking of
anything about the gretl build for Windows that would account for a
general slowdown or change in memory use.
Unless... were you looking at matrix intensive operations? About 6
weeks ago I substituted openblas for the netlib reference lapack +
blas in the 32-bit Windows build. (We've been using openblas with
the 64-bit build for much longer.)
This _ought_ to speed things up but I suppose that under very
unfavourable conditions it could be counter-productive. I'm not able
to test on 32-bit Windows myself; I tested the new 32-bit openblas
build only on 64-bit Windows in WoW64 mode.
If you continue to see suspicious performance on 32-bit Win8, please
try running the matrix_perf package and send me the results.
Allin
3631
days inactive
3632
days old
4 comments
2 participants
participants (2)
-
Allin Cottrell -
Sven Schreiber