Niggli-reduced cell C++ implementation

older
Handling unmerged reflections from...

Martin Uhrin

22 Mar 2012 22 Mar '12

6:24 p.m.

Dear Cctbx community, Firstly I'd like to say thank you to Rolf, Nicholas and Paul for their expertly thought through implementation of the reduced cell algorithm. I've found it to be extremely useful for my work. My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm. My understanding is that only the reduced (Buerger) unit cell algorithm is implemented in C++ [1] which guarantees shortest lengths but not unique angles. From my understanding the Krivy-Gruber would also guarantee me uniqueness of unit cell angles, however this is only implemented in Python [2]. Sorry to be so verbose, I just wanted to check that I was on the right page. Would it be possible for me to implement the Krivy-Gruber in C++ by adding in the epsilon_relative to the parameter and following the procedure found in the python version? Many thanks, -Martin [1] http://cctbx.sourceforge.net/current/c_plus_plus/classcctbx_1_1uctbx_1_1fast... [2] http://cctbx.sourceforge.net/current/python/cctbx.uctbx.krivy_gruber_1976.ht... -- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

Attachments:

attachment.html (text/html — 2.0 KB)

Show replies by date

David Lonie

22 Mar 22 Mar

6:39 p.m.

Hi Martin, I might be able to save you some effort here ;-) On Thu, Mar 22, 2012 at 5:24 PM, Martin Uhrin wrote:

...

My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm.

I've put my C++ implementation of the Krivy-Gruber algorithm described in that paper (doi:10.1107/S010876730302186X) under the BSD license in this project: http://github.com/dlonie/XtalComp The actual algorithm starts here: http://github.com/dlonie/XtalComp/blob/master/xtalcomp.cpp#L1538 It uses a linear algebra library that ships with XtalComp, but could be easily adapted to remove that dependency. I've also implemented it here using the Eigen linear algebra library: https://github.com/dlonie/avogadro/blob/master/libavogadro/src/extensions/cr... The code in this last link is technically licensed under GPLv2+, but I'll gladly provide it under a BSD-ish license if you're interested. Hope this helps, Dave

Ralf Grosse-Kunstleve

7 p.m.

Hi Martin, You're very welcome to add a C++ version of the Krivy-Gruber algorithm to cctbx if that's what you had in mind. I'm not sure what's better, generalizing the fast-minimum-reduction code, or just having an independent implementation. Ralf On Thu, Mar 22, 2012 at 2:24 PM, Martin Uhrin wrote:

...

Dear Cctbx community,

Firstly I'd like to say thank you to Rolf, Nicholas and Paul for their expertly thought through implementation of the reduced cell algorithm. I've found it to be extremely useful for my work.

My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm. My understanding is that only the reduced (Buerger) unit cell algorithm is implemented in C++ [1] which guarantees shortest lengths but not unique angles. From my understanding the Krivy-Gruber would also guarantee me uniqueness of unit cell angles, however this is only implemented in Python [2]. Sorry to be so verbose, I just wanted to check that I was on the right page.

Would it be possible for me to implement the Krivy-Gruber in C++ by adding in the epsilon_relative to the parameter and following the procedure found in the python version?

Many thanks, -Martin

[1] http://cctbx.sourceforge.net/current/c_plus_plus/classcctbx_1_1uctbx_1_1fast... [2] http://cctbx.sourceforge.net/current/python/cctbx.uctbx.krivy_gruber_1976.ht...

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

Martin Uhrin

23 Mar 23 Mar

7:35 a.m.

Dear David and Rolf, thank you for your encouragement. David: I'm more than happy to port your implementation to cctbx if you're happy with this. Of course I don't want to step on your toes so if you'd rather do it yourself (or not at all) that's cool. There may be some licensing issues to sort out as it looks like cctbx has a custom (non viral) license but the BSD license is likely compatible. On first impression I think a new class would be the way to go but I'd have to look at the two algorithms in greater detail to be sure. All the best, -Martin On 22 March 2012 22:00, Ralf Grosse-Kunstleve wrote:

...

Hi Martin, You're very welcome to add a C++ version of the Krivy-Gruber algorithm to cctbx if that's what you had in mind. I'm not sure what's better, generalizing the fast-minimum-reduction code, or just having an independent implementation. Ralf

On Thu, Mar 22, 2012 at 2:24 PM, Martin Uhrin wrote:

...
Dear Cctbx community,

Firstly I'd like to say thank you to Rolf, Nicholas and Paul for their expertly thought through implementation of the reduced cell algorithm. I've found it to be extremely useful for my work.

My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm. My understanding is that only the reduced (Buerger) unit cell algorithm is implemented in C++ [1] which guarantees shortest lengths but not unique angles. From my understanding the Krivy-Gruber would also guarantee me uniqueness of unit cell angles, however this is only implemented in Python [2]. Sorry to be so verbose, I just wanted to check that I was on the right page.

Would it be possible for me to implement the Krivy-Gruber in C++ by adding in the epsilon_relative to the parameter and following the procedure found in the python version?

Many thanks, -Martin

[1] http://cctbx.sourceforge.net/current/c_plus_plus/classcctbx_1_1uctbx_1_1fast... [2] http://cctbx.sourceforge.net/current/python/cctbx.uctbx.krivy_gruber_1976.ht...

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

David Lonie

11 a.m.

Hi Martin, On Fri, Mar 23, 2012 at 6:35 AM, Martin Uhrin wrote:

...

I'm more than happy to port your implementation to cctbx if you're happy with this. Of course I don't want to step on your toes so if you'd rather do it yourself (or not at all) that's cool.

There may be some licensing issues to sort out as it looks like cctbx has a custom (non viral) license but the BSD license is likely compatible.

Sure, feel free to do so -- I'd have done it myself by now, but free time is scarce these days. It looks like cctbx is also licensed under the "New" BSD license, so there should be no worries about that. Dave

Ralf Grosse-Kunstleve

2:12 p.m.

Hi Martin, Let me know if you need svn write access to check in your changes. All I need is your sourceforge user id. Ralf On Fri, Mar 23, 2012 at 3:35 AM, Martin Uhrin wrote:

...

Dear David and Rolf,

thank you for your encouragement.

David: I'm more than happy to port your implementation to cctbx if you're happy with this. Of course I don't want to step on your toes so if you'd rather do it yourself (or not at all) that's cool.

There may be some licensing issues to sort out as it looks like cctbx has a custom (non viral) license but the BSD license is likely compatible.

On first impression I think a new class would be the way to go but I'd have to look at the two algorithms in greater detail to be sure.

All the best, -Martin

On 22 March 2012 22:00, Ralf Grosse-Kunstleve wrote:

...
Hi Martin, You're very welcome to add a C++ version of the Krivy-Gruber algorithm to cctbx if that's what you had in mind. I'm not sure what's better, generalizing the fast-minimum-reduction code, or just having an independent implementation. Ralf

On Thu, Mar 22, 2012 at 2:24 PM, Martin Uhrin wrote:

...
Dear Cctbx community,

Firstly I'd like to say thank you to Rolf, Nicholas and Paul for their expertly thought through implementation of the reduced cell algorithm. I've found it to be extremely useful for my work.

My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm. My understanding is that only the reduced (Buerger) unit cell algorithm is implemented in C++ [1] which guarantees shortest lengths but not unique angles. From my understanding the Krivy-Gruber would also guarantee me uniqueness of unit cell angles, however this is only implemented in Python [2]. Sorry to be so verbose, I just wanted to check that I was on the right page.

Would it be possible for me to implement the Krivy-Gruber in C++ by adding in the epsilon_relative to the parameter and following the procedure found in the python version?

Many thanks, -Martin

[1] http://cctbx.sourceforge.net/current/c_plus_plus/classcctbx_1_1uctbx_1_1fast... [2] http://cctbx.sourceforge.net/current/python/cctbx.uctbx.krivy_gruber_1976.ht...

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

Martin Uhrin

24 Apr 24 Apr

2:57 p.m.

Dear cctbxers, I've finally found the time to play around with a C++ version of the KG algorithm and I've come across a result I don't understand. I've tried both David's C++ and the cctbx python niggli_cell() implementations and they both give the roughly the same answer. I'm reducing the following cell with two, equivalent, representations (a, b, c, alpha, beta, gamma): Before: 1: 4.630811 4.630811 4.630811 90 90 90 2: 3.27448 5.67156 5.67156 99.5941 106.779 90 After: 1: 4.63081 4.63081 4.63081 90 90 90 2: 3.27448 5.67154 5.67156 99.5941 90 106.778 Looking at the trace, cell 1 undergoes step 3 and finishes while cell 2 undergoes steps 2, 3, 7 and 4. Does anyone know why these haven't converged to the same cell? Many thanks, -Martin On 23 March 2012 17:12, Ralf Grosse-Kunstleve wrote:

...

Hi Martin, Let me know if you need svn write access to check in your changes. All I need is your sourceforge user id. Ralf

On Fri, Mar 23, 2012 at 3:35 AM, Martin Uhrin wrote:

...
Dear David and Rolf,

thank you for your encouragement.

David: I'm more than happy to port your implementation to cctbx if you're happy with this. Of course I don't want to step on your toes so if you'd rather do it yourself (or not at all) that's cool.

There may be some licensing issues to sort out as it looks like cctbx has a custom (non viral) license but the BSD license is likely compatible.

On first impression I think a new class would be the way to go but I'd have to look at the two algorithms in greater detail to be sure.

All the best, -Martin

On 22 March 2012 22:00, Ralf Grosse-Kunstleve
...
wrote:

...
Hi Martin, You're very welcome to add a C++ version of the Krivy-Gruber algorithm to cctbx if that's what you had in mind. I'm not sure what's better, generalizing the fast-minimum-reduction code, or just having an independent implementation. Ralf

On Thu, Mar 22, 2012 at 2:24 PM, Martin Uhrin
...
wrote:

...
Dear Cctbx community,

Firstly I'd like to say thank you to Rolf, Nicholas and Paul for their expertly thought through implementation of the reduced cell algorithm. I've found it to be extremely useful for my work.

My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm. My understanding is that only the reduced (Buerger) unit cell algorithm is implemented in C++ [1] which guarantees shortest lengths but not unique angles. From my understanding the Krivy-Gruber would also guarantee me uniqueness of unit cell angles, however this is only implemented in Python [2]. Sorry to be so verbose, I just wanted to check that I was on the right page.

Would it be possible for me to implement the Krivy-Gruber in C++ by adding in the epsilon_relative to the parameter and following the procedure found in the python version?

Many thanks, -Martin

[1] http://cctbx.sourceforge.net/current/c_plus_plus/classcctbx_1_1uctbx_1_1fast... [2] http://cctbx.sourceforge.net/current/python/cctbx.uctbx.krivy_gruber_1976.ht...

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

Ralf Grosse-Kunstleve

4:42 p.m.

Hi Martin, Based on iotbx.lattice_symmetry --unit_cell="4.630811 4.630811 4.630811 90 90 90" and iotbx.lattice_symmetry --unit_cell="3.27448 5.67156 5.67156 99.5941 106.779 90" the first unit cell is (obviously) cubic, the second is only monoclinic. Even with iotbx.lattice_symmetry --unit_cell="3.27448 5.67156 5.67156 99.5941 106.779 90" --delta=20 it only comes back as orthorhombic. Is this what you expect? Ralf On Tue, Apr 24, 2012 at 10:57 AM, Martin Uhrin wrote:

...

Dear cctbxers,

I've finally found the time to play around with a C++ version of the KG algorithm and I've come across a result I don't understand. I've tried both David's C++ and the cctbx python niggli_cell() implementations and they both give the roughly the same answer.

I'm reducing the following cell with two, equivalent, representations (a, b, c, alpha, beta, gamma):

Before:

1: 4.630811 4.630811 4.630811 90 90 90 2: 3.27448 5.67156 5.67156 99.5941 106.779 90

After:

1: 4.63081 4.63081 4.63081 90 90 90 2: 3.27448 5.67154 5.67156 99.5941 90 106.778

Looking at the trace, cell 1 undergoes step 3 and finishes while cell 2 undergoes steps 2, 3, 7 and 4.

Does anyone know why these haven't converged to the same cell?

Many thanks, -Martin

On 23 March 2012 17:12, Ralf Grosse-Kunstleve wrote:

...
Hi Martin, Let me know if you need svn write access to check in your changes. All I need is your sourceforge user id. Ralf

On Fri, Mar 23, 2012 at 3:35 AM, Martin Uhrin wrote:

...
Dear David and Rolf,

thank you for your encouragement.

David: I'm more than happy to port your implementation to cctbx if you're happy with this. Of course I don't want to step on your toes so if you'd rather do it yourself (or not at all) that's cool.

There may be some licensing issues to sort out as it looks like cctbx has a custom (non viral) license but the BSD license is likely compatible.

On first impression I think a new class would be the way to go but I'd have to look at the two algorithms in greater detail to be sure.

All the best, -Martin

On 22 March 2012 22:00, Ralf Grosse-Kunstleve < [email protected]> wrote:

...
Hi Martin, You're very welcome to add a C++ version of the Krivy-Gruber algorithm to cctbx if that's what you had in mind. I'm not sure what's better, generalizing the fast-minimum-reduction code, or just having an independent implementation. Ralf

On Thu, Mar 22, 2012 at 2:24 PM, Martin Uhrin < [email protected]> wrote:

...
Dear Cctbx community,

Firstly I'd like to say thank you to Rolf, Nicholas and Paul for their expertly thought through implementation of the reduced cell algorithm. I've found it to be extremely useful for my work.

My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm. My understanding is that only the reduced (Buerger) unit cell algorithm is implemented in C++ [1] which guarantees shortest lengths but not unique angles. From my understanding the Krivy-Gruber would also guarantee me uniqueness of unit cell angles, however this is only implemented in Python [2]. Sorry to be so verbose, I just wanted to check that I was on the right page.

Would it be possible for me to implement the Krivy-Gruber in C++ by adding in the epsilon_relative to the parameter and following the procedure found in the python version?

Many thanks, -Martin

[1] http://cctbx.sourceforge.net/current/c_plus_plus/classcctbx_1_1uctbx_1_1fast... [2] http://cctbx.sourceforge.net/current/python/cctbx.uctbx.krivy_gruber_1976.ht...

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

Martin Uhrin

5:42 p.m.

Dear Ralf, you're quite right! I was looking at my basis along with the lattice which lead me to think that the two systems should be equivalent when indeed the lattices are not! Thanks for pointing this out. I'll get back to you once I'm ready with a version for the repository. Best, -Martin On 24 April 2012 20:42, Ralf Grosse-Kunstleve wrote:

...

Hi Martin,

Based on

iotbx.lattice_symmetry --unit_cell="4.630811 4.630811 4.630811 90 90 90"

and

iotbx.lattice_symmetry --unit_cell="3.27448 5.67156 5.67156 99.5941 106.779 90"

the first unit cell is (obviously) cubic, the second is only monoclinic. Even with

iotbx.lattice_symmetry --unit_cell="3.27448 5.67156 5.67156 99.5941 106.779 90" --delta=20

it only comes back as orthorhombic.

Is this what you expect?

Ralf

On Tue, Apr 24, 2012 at 10:57 AM, Martin Uhrin wrote:

...
Dear cctbxers,

I've finally found the time to play around with a C++ version of the KG algorithm and I've come across a result I don't understand. I've tried both David's C++ and the cctbx python niggli_cell() implementations and they both give the roughly the same answer.

I'm reducing the following cell with two, equivalent, representations (a, b, c, alpha, beta, gamma):

Before:

1: 4.630811 4.630811 4.630811 90 90 90 2: 3.27448 5.67156 5.67156 99.5941 106.779 90

After:

1: 4.63081 4.63081 4.63081 90 90 90 2: 3.27448 5.67154 5.67156 99.5941 90 106.778

Looking at the trace, cell 1 undergoes step 3 and finishes while cell 2 undergoes steps 2, 3, 7 and 4.

Does anyone know why these haven't converged to the same cell?

Many thanks, -Martin

On 23 March 2012 17:12, Ralf Grosse-Kunstleve
...
wrote:

...
Hi Martin, Let me know if you need svn write access to check in your changes. All I need is your sourceforge user id. Ralf

On Fri, Mar 23, 2012 at 3:35 AM, Martin Uhrin
...
wrote:

...
Dear David and Rolf,

thank you for your encouragement.

David: I'm more than happy to port your implementation to cctbx if you're happy with this. Of course I don't want to step on your toes so if you'd rather do it yourself (or not at all) that's cool.

There may be some licensing issues to sort out as it looks like cctbx has a custom (non viral) license but the BSD license is likely compatible.

On first impression I think a new class would be the way to go but I'd have to look at the two algorithms in greater detail to be sure.

All the best, -Martin

On 22 March 2012 22:00, Ralf Grosse-Kunstleve < [email protected]> wrote:

...
Hi Martin, You're very welcome to add a C++ version of the Krivy-Gruber algorithm to cctbx if that's what you had in mind. I'm not sure what's better, generalizing the fast-minimum-reduction code, or just having an independent implementation. Ralf

On Thu, Mar 22, 2012 at 2:24 PM, Martin Uhrin < [email protected]> wrote:

...
Dear Cctbx community,

Firstly I'd like to say thank you to Rolf, Nicholas and Paul for their expertly thought through implementation of the reduced cell algorithm. I've found it to be extremely useful for my work.

My code is all in C++ and I'd like to be able to use the Krivy-Gruber algorithm. My understanding is that only the reduced (Buerger) unit cell algorithm is implemented in C++ [1] which guarantees shortest lengths but not unique angles. From my understanding the Krivy-Gruber would also guarantee me uniqueness of unit cell angles, however this is only implemented in Python [2]. Sorry to be so verbose, I just wanted to check that I was on the right page.

Would it be possible for me to implement the Krivy-Gruber in C++ by adding in the epsilon_relative to the parameter and following the procedure found in the python version?

Many thanks, -Martin

[1] http://cctbx.sourceforge.net/current/c_plus_plus/classcctbx_1_1uctbx_1_1fast... [2] http://cctbx.sourceforge.net/current/python/cctbx.uctbx.krivy_gruber_1976.ht...

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

_______________________________________________ cctbxbb mailing list [email protected] http://phenix-online.org/mailman/listinfo/cctbxbb

-- Martin Uhrin Tel: +44 207 679 3466 Department of Physics & Astronomy Fax:+44 207 679 0595 University College London [email protected] Gower St, London, WC1E 6BT, U.K. http://www.cmmp.ucl.ac.uk

4652

Age (days ago)

4685

Last active (days ago)

List overview

Download

8 comments

3 participants

participants (3)

David Lonie
Martin Uhrin
Ralf Grosse-Kunstleve