min activities in a building

[Liviu Lalescu]

You are welcome!

NSRT4 is ready. You can check it (please firstly check on the file you gave immediately above).

[Benahmed Abdelkrim]

Oh! Amazing! The test was successful. I tried it with the same file attached above, and with the same random seeds.
Time: 2min 37s

random seeds:
s10=3385507001, s11=3350186166, s12=3250444770,
s20=1795543818, s21=3245662932, s22=1143393762

Behavior has also improved significantly!

Thank you very much dear Liviu! We are very grateful; me and the arabic users for the great effort you have made to improve this NSRT version.

[Liviu Lalescu]

You are welcome! Thank you! 

I might have left some minor details unperfected, but I am too tired now.

[Benahmed Abdelkrim]

I think the warning added yesterday is no longer useful given these significant improvements in production speed. What do you think?

[Liviu Lalescu]

Sure! I'll remove it. I think soon.

[Liviu Lalescu]

I am doing it now, but there is a cosmetic thing of which I am not 100% satisfied... I might leave it this way, it works correctly in practice. I am not sure. Still thinking.

I hope it will be ready in maximum 30 minutes, so you can announce the new version.

[Benahmed Abdelkrim]

I am doing it now, but there is a cosmetic thing of which I am not 100% satisfied... I might leave it this way, it works correctly in practice. I am not sure. Still thinking.

Take your time, there's no rush...

[Liviu Lalescu]

I just put the new version, fet-7.8.0-nsrt5. Please check that I removed the warnings in building(s) constraints (two constraints, singular and plural)!

The cosmetic thing needs to wait. I cannot decide on how to make it, I need time to understand the matter. If the program works in practice, it is a negligible thing. Like the difference from correctness 99.996% and 99.997%.

Please test the new version practically also with different sets of data, I might have forgotten some things! Test with different MTR=x (0 or 1 - I saw that your latest file contained only 0s). Also, I think each activity must have a room and a building, otherwise the program might not work or even crash. We might solve these if needed.

[Benahmed Abdelkrim]

Sure! I'll test and report...

[Benahmed Abdelkrim]

I started testing, but something caught my attention: with mtr>0, I think the behavior changed; time doesn't flow normally. The seconds jump from 5 to 7 to 10, etc., randomly. Is this normal? ...

[Liviu Lalescu]

It can mean one thing: the MTR>0 implies that the generation is slower. The seconds are updated after a complete cycle. I'll have now a look at the MTR procedure to check. At the time I wrote that I was sure it is maximum optimized, but who knows?

I can tell you that with FET-official, there are many difficult files with this behavior. Some constraints take longer time.

The unoptimized fet-7.8.0-nsrt to fet-7.8.0-nsrt3 had about 5-10 seconds delay, if I remember.

Also, at the time of writing MTR, I thought it was quite inefficient, but could not improve it.

Could you tell me if the results are good, even with this "jump"?

[Benahmed Abdelkrim]

The results are generally good, the constraints are respected, no errors... I think, as you said, it's because of the MTR; perhaps it needs to be checked...

[Liviu Lalescu]

I just watched the code. Indeed, it is different for MTR=0 and MTR>0.

I hope the code is optimized as it should be, I mean without possibility of improvement, but I might investigate tomorrow some more.

The files with MTR>0 solve correctly and fast enough?

As I told you, there are some (many?) official constraints/files with this behavior.

[Liviu Lalescu]

OK, I'll have a second look now at MTR and let you know.

[Benahmed Abdelkrim]

The files with MTR>0 solve correctly and fast enough?

Yes, no major changes except this jump; the speed is good, the results are acceptable. I think we should be satisfied with these good results.