whetstone test

whetstone test

[Sergei Gavrikov]

Hi,

What is your opinion, Is it a good idea to have `whetstone' test for
eCos? http://en.wikipedia.org/wiki/Whetstone_(benchmark)

The source can be grabbed from here: http://www.netlib.org/benchmark/ I
use http://www.netlib.org/benchmark/whetstone.c (FYI: no warnings with
nowadays toolchains).

If your answer will be, 'Yes, it is', my next question: where to have
the test? I placed it together with `dhrystone' test under eCos kernel
tests with a minimal tweak for CYGPKG_KERNEL_TESTS CDL option:

  . ((CYGPKG_LIBM) ? " tests/whetstone" : "")

Of course, I would place the test together with the libm  tests if that
is more suitable place. What is better?

NOTE: Mostly, I use the test to know how evolved the eCos toolchains.

Sergei

Re: whetstone test

[John Dallaway]

Hi Sergei

Sergei Gavrikov wrote:

> What is your opinion, Is it a good idea to have `whetstone' test for
> eCos? http://en.wikipedia.org/wiki/Whetstone_(benchmark)

What is the license and copyright situation with this benchmark code?
The source code at:

  http://freespace.virgin.net/roy.longbottom/whets.c

does not include this information.

John Dallaway

Re: whetstone test

[Sergei Gavrikov]

On Fri, 26 Nov 2010, John Dallaway wrote:
> Sergei Gavrikov wrote:
> 
> > What is your opinion, Is it a good idea to have `whetstone' test for
> > eCos? http://en.wikipedia.org/wiki/Whetstone_(benchmark)
> 
> What is the license and copyright situation with this benchmark code?
> The source code at:
> 
>   http://freespace.virgin.net/roy.longbottom/whets.c
> 
> does not include this information.

Hi John,

I confess that I have not thought about it :-( Here are just now checked
out where it can be (my version) http://en.wikipedia.org/wiki/Netlib
My mistake. I will look around other implementations.

Thank you,
Sergei

Re: whetstone test

[Sergei Gavrikov]

On Fri, 26 Nov 2010, Sergei Gavrikov wrote:
> On Fri, 26 Nov 2010, John Dallaway wrote:
> > Sergei Gavrikov wrote:
> > 
> > > What is your opinion, Is it a good idea to have `whetstone' test for
> > > eCos? http://en.wikipedia.org/wiki/Whetstone_(benchmark)
> > 
> > What is the license and copyright situation with this benchmark code?
> > The source code at:
> > 
> >   http://freespace.virgin.net/roy.longbottom/whets.c
> > 
> > does not include this information.
> 
> I confess that I have not thought about it :-( Here are just now checked
> out where it can be (my version) http://en.wikipedia.org/wiki/Netlib
> My mistake. I will look around other implementations.

Hi John,

I found a remake of whetstone.c test (original one came from UNIX world
as I could understand) on one NetBSD FTP mirror. I would not want to
paste URL address now, but it leaves under 'NetBSD/packages/distfiles'
directory. Can it be "a pass"? Also there is another test 'flops.c' (it
calculates the MFLOPS value) in the NetBSD packages.

Both of NetBSD sources can be dig with a 'FileWatcher' Search Engine:
http://www.filewatcher.com

Can you, please, check copyright header for that whetstone.c (1.2) from
NetBSD packages?

IANAL and I would withdraw my questions if NetBSD trunk is not (C)-pass.

Thank you,
Sergei

Re: whetstone test

[John Dallaway]

Hi Sergei

Sergei Gavrikov wrote:

> Can you, please, check copyright header for that whetstone.c (1.2) from
> NetBSD packages?
> 
> IANAL and I would withdraw my questions if NetBSD trunk is not (C)-pass.

IANAL either.

I note that we _do_ have a precedent for including benchmark code with
no stated copyright and licensing terms. Namely, the dhrystone test:

http://ecos.sourceware.org/cgi-bin/cvsweb.cgi/ecos/packages/kernel/current/tests/dhrystone.c?rev=1&content-type=text/x-cvsweb-markup&cvsroot=ecos

In my opinion, including a whetstone test adapted from the original
whets.c code from http://www.roylongbottom.org.uk/whets.c would be
acceptable (avoiding other versions with company copyright notices). Do
any of the other maintainers have an opinion on this?

I would be inclined to place the test alongside the dhrystone test in
the kernel package with similar pre-processor tests for correct
configuration of eCos, similar implementation of dtime() and similar use
of __ECOS__ to annotate eCos-specific changes. I would also suggest
#ifdefs allowing the test to build both with and without libm since only
two of the test sections (N5, N8) appear to require libm.

John Dallaway
eCos maintainer

Re: whetstone test

[Sergei Gavrikov]

Hi John,

John Dallaway wrote:

> Hi Sergei
> 
> Sergei Gavrikov wrote:
> 
> > Can you, please, check copyright header for that whetstone.c (1.2) from
> > NetBSD packages?
> > 
> > IANAL and I would withdraw my questions if NetBSD trunk is not (C)-pass.
> 
> IANAL either.
> 
> I note that we _do_ have a precedent for including benchmark code with
> no stated copyright and licensing terms. Namely, the dhrystone test:
> 
> http://ecos.sourceware.org/cgi-bin/cvsweb.cgi/ecos/packages/kernel/current/tests/dhrystone.c?rev=1&content-type=text/x-cvsweb-markup&cvsroot=ecos

Yep, I did take a look on the dhrystone's header in CVS at that moment
and I thought, Why not have the same reservation about whetstone test,
but still...

> In my opinion, including a whetstone test adapted from the original
> whets.c code from http://www.roylongbottom.org.uk/whets.c would be
> acceptable (avoiding other versions with company copyright notices). Do
> any of the other maintainers have an opinion on this?

> I would be inclined to place the test alongside the dhrystone test in
> the kernel package with similar pre-processor tests for correct
> configuration of eCos, similar implementation of dtime() and similar use
> of __ECOS__ to annotate eCos-specific changes. I would also suggest
> #ifdefs allowing the test to build both with and without libm since only
> two of the test sections (N5, N8) appear to require libm.

Thank you for your expert opinion.

If other maintainers would have no objection on the source of the test I
would adapt its source code for eCos environment/needs, test and post a
result patch to eCos Bugzilla registrar.

Thank you,
Sergei

> John Dallaway
> eCos maintainer

Re: whetstone test

[Sergei Gavrikov]

On Tue, 30 Nov 2010, Sergei Gavrikov wrote:
> John Dallaway wrote:
> > Sergei Gavrikov wrote:
> > > Can you, please, check copyright header for that whetstone.c (1.2)
> > > from NetBSD packages?
> > > 
> > > IANAL and I would withdraw my questions if NetBSD trunk is not
> > > (C)-pass.
> > 
> > IANAL either.
> > 
> > I note that we _do_ have a precedent for including benchmark code
> > with no stated copyright and licensing terms. Namely, the dhrystone
> > test:
> > 
> > http://ecos.sourceware.org/cgi-bin/cvsweb.cgi/ecos/packages/kernel/current/tests/dhrystone.c?rev=1&content-type=text/x-cvsweb-markup&cvsroot=ecos
> 
> Yep, I did take a look on the dhrystone's header in CVS at that moment
> and I thought, Why not have the same reservation about whetstone test,
> but still...
> 
> > In my opinion, including a whetstone test adapted from the original
> > whets.c code from http://www.roylongbottom.org.uk/whets.c would be
> > acceptable (avoiding other versions with company copyright notices).
> > Do any of the other maintainers have an opinion on this?
> 
> > I would be inclined to place the test alongside the dhrystone test
> > in the kernel package with similar pre-processor tests for correct
> > configuration of eCos, similar implementation of dtime() and similar
> > use of __ECOS__ to annotate eCos-specific changes. I would also
> > suggest #ifdefs allowing the test to build both with and without
> > libm since only two of the test sections (N5, N8) appear to require
> > libm.
> 
> Thank you for your expert opinion.
> 
> If other maintainers would have no objection on the source of the test
> I would adapt its source code for eCos environment/needs, test and
> post a result patch to eCos Bugzilla registrar.

Hi All,
hi John,

As we are not lawyers :-), I followed the established precedent by the
'eCos/Red Hat' for dhrystone test (as John suggested). I do not know
whether it was appropriate here, well, I set the next preamble:

//=============================================================================
//####UNSUPPORTEDBEGIN####
//
// -------------------------------------------
// This source file has been contributed to eCos. It may have been changed
// slightly to provide an interface consistent with those of other files.
//
// The functionality and contents of this file is supplied "AS IS"
// without any form of support and will not necessarily be kept up
// to date by eCos Maintainers.
//
// The style of programming used in this file may not comply with the
// eCos programming guidelines. Please do not use as a base for other
// files.
//
// All inquiries about this file, or the functionality provided by it,
// should be directed to the 'ecos-discuss' mailing list (see
// http://ecos.sourceware.org/ecos/intouch.html for details).
//
// -------------------------------------------
//
//####UNSUPPORTEDEND####
//=============================================================================
// Converted DOS line terminators to UNIX ones, changed a little to avoid
// compilation warnings.
// The file was found at:
// HTTP www.roylongbottom.org.uk/whets.c
//=============================================================================

The fixes against dos2unix'ed original (including above banner) were:

whets.c |  162 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 160 insertions(+), 2 deletions(-)

The test expands a set of eCos kernel tests (my testing was done for
ARM/i386 targets).

If your opinion - *We should not create new precedents*, then I simply
will build EPK and frozen it. If the test can be a useful addition, I
will post it to eCos Bugzilla.

Thank you for your point of view, thanks John for his preview, suggests
and the link. And of course, thanks to Roy for his benchmark program.

Sergei