LHC@home

Dear long suffering LHC2home and SixTrack supporter.My excuses for the long silence but I have been very very busy. After ten years of work, five years retired, 70th birthday come and gone , four years in the wilderness, it is all happening again. The BOINC servers are now virtual machines running here at CERN using the latest server_stable BOINC server. We are experiencing some problems with the WWW in particular and specifically with the Message Boards. I will try and find time soonest to reply individually to your posts or at least make sure you see this message. We are in a pre-production phase testing everything end to end from the accelerator Physicist DeskTop to the returning of the results. The ambitious aim is to start production next week with a lot of work expected in the next months as studies are made for the LHC upgrade in particular.
I myself have had to switch to the Intel ifort compiler instead of the old friend Lahey-Fujitsu lf95. Just impossible to link on Windows because of some issue re single/multi-thread libraries. No garphics for the moment, sorry. We do have Linux and Windows executables and since there is an ifort compiler for Macs, why not for them too. Coming, soon??? I have had to suspend activity on my long term goal of achieving 0 ULP difference with any Fortran compiler, at any level of optimisation, on any IEE754 compliant hardware (but I'm close). Sadly the ifort compiler gave diferent results on Windows and Linux. The problem is with formatted input (and output), FIO for short, and is well known and basically solved by David M. Gay some twenty years ago. For the short term tests I am working round it by treating all alignments, tilts and magnet errors as Single Precision rather than Double. This problem is solved in Fortran 2003 where FIO can be specified with various ROUND options on the OPEN, READ/WRITE, or FORMAT specification. Again although ifort accepts the syntax it gives the same internal representation for all rounding options. So in the meantime I am planning to implement a C routine for FIO, well for Input anyway. I also have to worry about performance improvement. (As I have said in the past this 0 ULP will allow me to try GPUs.) I shall also be running tasks several/many times on diferent hadware to ensure I have really solved the 0 ULP problem with ifort as I had done for lf95. Well that is it for now. Tired but happy and as usual I shall be giving priority to establishing a 99.9% level of service and I shall try and keep you better informed, at least of my own work.

---

....and I must acknowledge the great help being received from CERN/IT, the Ecole Polytechnique Federale in Lausanne (EPFL), and from other colleagues. We are doing our best to keep your existing credits. I shall also try and dream up a suitable award to the best? largest? most constructive supporter!

---

After ten years of work, five years retired, 70th birthday come and gone , four years in the wilderness, it is all happening again.

thank you very much!

Out of curiosity, does this involve Kahan-Babuška Summation related algorithms (see e.g. http://cage.ugent.be/~klein/papers/floating-point.pdf)? Aside from that, of course it's important to avoid catastrophic cancellation (e.g. use (x+y)*(x-y) instead of x²-y²).[/quote]

I'll look at tah tomorrow--but I indeed use for example (1-x)*(1+x) instead of 1-x**2.

---

Indeed and thanks for the comstructive feedback.

---

bigmac wrote:

Dear long suffering LHC2home and SixTrack supporter...

Thanks and as you can tell by my user id of 289, That's me, 7th anniversary with project earlier this month, hard to beleive it has been so long.

Welcome back, I guess boinc has changed somewhat since you were last around. Glad to have you back on the team.

I would like to hear where the project is going now. I know that there needs to be a lot of experiments run for either new magnets, or future upgrades. Sometimes i can't keep up with everything and i'm not sure at the moment whre this is. Will there also be other applications on this project or jsut sixtrack, I see sixtrackBNL and garfield still listed.

Thanks for the server_status link and love the new logo. Can't wait for more surprises to pop up.

I hate this message system.......just managed to
delete everything I wrote. I guess I should edit a
message and then post it.

Anway congrats for sticking with us. I think Garfield
is not being worked on, BNL might be interested in
coming aboard as I have carefully preserved their special code
for studying emittance and the electron lens. We are working on thE
web pages which will explain where we are going. We have also
had to fight our corner to avoid being displaced by the
so-called LHC@home 2.0...........The Virtual Macine concept
is great but demands more resources but I will be discussiong with
the T4T end user soon I hope. Got to get some sleep after
checking the submission of new work, should be one million
turns but of course particles get lost.

I I have your e-mail I'll send you a beautiful picture
of pahse space as a souvenir. Eric. eric.mcintosh#cern.ch

---

Do like I do as i've had many times typed a long composition taking 30 minutes, then hit send only to find the server went down, message lost in phasespace.

periodically when making long messages, or just before i hit send, i hit ctrl-a and ctrl-c, select all and copy to clipboard.

most of the time its not needed, but when the message is lost, you're glad you wasted the time to hit those combinations.

--

PM'd you my email address.

--

Yes the VM machine is a nice concept, but it does require a little more attention than a standard projects. Like gpu projects, you have to keep updated drivers, or in this case the vm. It is not for everyone.

--

New web pages expalining where you are going would be nice, I just hope you are not going in circles.

Lost particles are OK as long as they don't hit me in my head, my tin foil hat has a hole in it and is out for repair.

Wow!

You brought back some old memories there.

I used to run a searchlight board. (Without much success).

I remember Fidonet.

---