Re: [Gretl-devel] user-defined functions and smpl
by Allin Cottrell
On Thu, 20 Jan 2011, Summers, Peter wrote:
> One case that I can think of where your proposal would make
> sense is in the handling of lags. In your example, if I set the
> sample range to 90:1-05:4 and then call a function that
> estimates a VAR(4), I think I'd want the function to be able to
> access the 89:1-89:4 observations, rather than estimating from
> 91:1.
In fact it can already do so: "reach-back" to pre-sample values is
allowed in general when estimating autoregressive models.
But here's a case where the proposal could make a difference:
inside the function you want to define some new variables and
include them in a VAR. The new variables would be defined only
over the current sample range, so (accounting for lags) the start
of VAR estimation would be pushed forwards. But you could avoid
this, in the function, if you were allowed to push the sample
start back for generating the new variables, then reset it for
estimation.
Allin.
13 years, 3 months
Re: [Gretl-devel] Can't open datafile
by Allin Cottrell
I've now modified the configure script in CVS, and I think that
things should now be right for the case where the compiler
supports sse2 but the CPU does not. That is, you shouldn't have to
manually --disable-sse2 to get a working binary.
However, I can't check that definitively on my own machine.
Allin
13 years, 3 months
Re: [Gretl-devel] RNG (was: Can't open datafile)
by Allin Cottrell
On Wed, 19 Jan 2011, Allin Cottrell wrote:
> On Wed, 19 Jan 2011, Riccardo (Jack) Lucchetti wrote:
>
> > On Tue, 18 Jan 2011, Allin Cottrell wrote:
> >
> > > I've now modified the configure script in CVS, and I think that
> > > things should now be right for the case where the compiler
> > > supports sse2 but the CPU does not. That is, you shouldn't have to
> > > manually --disable-sse2 to get a working binary.
> > >
> > > However, I can't check that definitively on my own machine.
> >
> > Just crossed my mind: what happens to random number generation on machines
> > which don't support sse2 (which I assume is the case on non-x86
> > architectures)? I have no time to check this right now, but I would assume
> > we fall back on the old Mersenne Twister. But if this is the case, we'd
> > have that scripts with the same seed would produce architecture-specific
> > results. Do we want this?
>
> The SFMT doesn't actually require sse2. I think it will just run
> slower in the absence of sse2, but it's probably worth checking
> to see if you get the same sequence, for a given seed, with
> and without sse2.
Just checked: you get the same sequence from SFMT with and without
sse2, so we're OK.
Allin
13 years, 3 months
Re: [Gretl-devel] RNG (was: Can't open datafile)
by Allin Cottrell
On Wed, 19 Jan 2011, Riccardo (Jack) Lucchetti wrote:
> On Tue, 18 Jan 2011, Allin Cottrell wrote:
>
> > I've now modified the configure script in CVS, and I think that
> > things should now be right for the case where the compiler
> > supports sse2 but the CPU does not. That is, you shouldn't have to
> > manually --disable-sse2 to get a working binary.
> >
> > However, I can't check that definitively on my own machine.
>
> Just crossed my mind: what happens to random number generation on machines
> which don't support sse2 (which I assume is the case on non-x86
> architectures)? I have no time to check this right now, but I would assume
> we fall back on the old Mersenne Twister. But if this is the case, we'd
> have that scripts with the same seed would produce architecture-specific
> results. Do we want this?
The SFMT doesn't actually require sse2. I think it will just run
slower in the absence of sse2, but it's probably worth checking
to see if you get the same sequence, for a given seed, with
and without sse2.
Allin
13 years, 3 months
Re: [Gretl-devel] Can't open datafile
by Allin Cottrell
On Wed, 19 Jan 2011, Alexander B. Gedranovich wrote:
> I got
>
> SSE2 not supported
Thanks, maybe we can make use of that approach.
Allin
13 years, 3 months
Re: [Gretl-devel] Can't open datafile
by Allin Cottrell
On Tue, 18 Jan 2011, Alexander B. Gedranovich wrote:
> It still thinks sse2 is supported.
Hmm. Could you maybe try the attached little C program and see if
it gets the answer right?
gcc sse2check.c
./a.out
should do it.
Allin
13 years, 3 months
Re: [Gretl-devel] Can't open datafile
by Allin Cottrell
yOn Tue, 18 Jan 2011, Alexander B. Gedranovich wrote:
> OK, thank you.
>
> On 01/18/2011 11:15 PM, Allin Cottrell wrote:
>
> Thanks, very helpful. The problem seems to be that your installed
> gcc is out of sync with your processor...
If you have time and patience, you could try this:
* Edit gretl's configure.in: look for the string
AC_COMPILE_IFELSE and change it to AC_RUN_IFELSE.
* Run autoconf in the top-level gretl directory.
* Then try ./configure (without --disable-sse2).
I guess one of three things will happen:
* The ideal outcome: configure finds that sse2 is not supported,
but completes successfully.
* It still thinks sse2 is supported.
* The configure script crashes on Illegal Instruction.
If we get the ideal outcome, this is an easy change to make.
Allin
13 years, 3 months
Re: [Gretl-devel] Can't open datafile
by Allin Cottrell
On Tue, 18 Jan 2011, Alexander B. Gedranovich wrote:
> 1) In attachment are outputs from
>
> cat /proc/cpuinfo | grep sse
> ./configure
> make
Thanks, very helpful. The problem seems to be that your installed
gcc is out of sync with your processor. The configure test for
SSE2 intrinsics is passing OK and gcc is happily allowing you to
compile with the -msse2 flag, but in fact the cpu has no sse2
support, so naturally the gretl binary is going to issue illegal
instructions.
I suppose (hope) this case is quite rare, in which case it can be
handled by --disable-sse2, but maybe we can put a more thorough
test for sse2 into the configure script.
Allin
13 years, 3 months
Re: [Gretl-devel] Can't open datafile
by Allin Cottrell
On Tue, 18 Jan 2011, Alexander B. Gedranovich wrote:
> Hello, Allin.
>
> You was right: --disable-sse2 had helped.
>
> Additional information:
> ==============================
> ales@opensuse:~>uname -a
> Linux opensuse 2.6.34.7-0.7-default #1 SMP 2010-12-13 11:13:53 +0100 i686
> athlon i386 GNU/Linux
>
> ales@opensuse:~> gcc -dumpversion
> 4.5
> ==============================
>
> ./configure & make outputs are attached.
Thanks, Alexander. It seems, though, that the configure output you
attached is from a run where you did --disable-sse2 (and got a
working executable) -- is that right?
What I'm really interested in is what's going on when you don't
explicitly disable sse2 (and get a non-working binary). The result
of
cat /proc/cpuinfo | grep sse
might also be of interest.
Allin
13 years, 3 months
Re: [Gretl-devel] Can't open datafile
by Allin Cottrell
On Mon, 17 Jan 2011, Allin Cottrell wrote:
> On Tue, 18 Jan 2011, Alexander B. Gedranovich wrote:
> > Illegal instruction
>
> What type of CPU, and what version of gcc? ...
As a temporary thing, you could see what happens if you add the
flag --disable-sse2 when configuring gretl.
Allin Cottrell
13 years, 3 months