Open DannyNicholls opened 10 years ago
BorisDosen
commented
10 years ago Hey Danny,
Thank you for your points. I don't have much time today to work on this I am afraid, but I will have a better look tomorrows and then get back to you with updates for these points. I don't think it will be too much work to make it suitable for testing 3D networks. Are you using a wrapper for 3D networks, similar to one for 2D to change from pkt.dest to pkt.dest_x, pkt.dest_y and pkt.dest_z?
DannyNicholls
commented
10 years ago Cool Cool. Don't worry about getting work done for me, I doubt I will have chance to play with NEMU now, today really is well past the last day I should be testing so I am using the test bench I have. I need to be sure my networks work before obtaining synthesis results. It's more just that those would be good things for NEMU to have, but do whatever you reckon will fill the report out most!
The way I have added 3D networks is not with a wrapper as before, I have done it using the packet_t definition ...
https://github.com/DannyNicholls/NetEmulation/issues/15
I will try and get round to adding a wrapper at some point, it should be easy, then you can use the network in 3D if you want when you get round to it.
DannyNicholls
commented
10 years ago I have moved the files and folders now, it was upsetting my obsessiveness to have that in the master branch.
Boris, those graphs you have achieved on Phils network look great, cheered me up no end this morning! They look really good. If I could be as bold as to offer some advice and ask some questioons....
1) I would not title the graphs as they will be titled in reports using the standard for writing reports for whatever paper they will be published in. It might be worth looking to see if the IEEE has a standard font and size etc for axis?
2) To be more clear of the measurement units, the axis title on the y axis could be "Avg. delay (cycles)" and on the x axis "Offered traffic (fraction of capacity)". The same as in dally and towles,
3) It appears that your zero load latency is ~2e5 = 200,000 cycles?
4) The warm up estimation could have a legend?
5) Perhaps the MatlabCode folder and Scripts folder should go in NetEmulation>NEMU> folder with the src and test folders? That way each IP can have it's own scripts etc?
5) Are you working on a throughput graph?
6) Does the test work with x and y co-ordinates? Will it be much work to add a z co-ordinate? EG the last post I added to the forum? I plan on having my 3D network finished today and to have tested all my networks in simulation, so I can finish synthesis results on tuesday. It would be cool if I could use your NEMU, but don't worry though if you're not working on it today, if I can't get yours to work It won't take me long to sort mine out.
7) Is there a reason you have put NEMU_fifo.sv in the LIB folder or is it just accident? If you have not used the updated LIB_FIFO.sv, just keep the fifo you used in the NEMU folder.
Nice one!