Hey all, im currently a senior EE student, and next term will be pretty light so a friend and i plan to mess around with a pipelined MIPS-esque cpu for fun. I figured i should actually get my shit together before we start in a few weeks.
Out of curiosity, has anyone successfully done anything with out of order scheduling or superscalar in general?
I think LordDecapo has an out of order system, but it's quite old and not on the server. If you ask him about it he'll probably be able to tell you about it.
also YES more people doing pipelined stuff
I'm an EE student too btw, only in 2nd year now though
Sweet I'll have to ask him about it.
And yeah we figured just single/multicycling it would be too simple, might as well pipeline it. Hazard detection might be a struggle though.
Now if only I could get someone to look at my student server application..
If you go on to one of the servers, you'll have a higher chance of getting someone to look at your app, if you see a staff member, link him your app in-game.
EDIT: I left a message with the admins, should speed the process along
I will look up your application now.
Also, my oooexe algorithm is a modified tomasulos (my spelling sucks) to make it better work at MC speeds.
Pm me or we can talk on here if you want to know more.
Hazard detection is easy btw, just a set of xnors/xors, can be done in about 2 ticks+bussing delay.
If you have TeamSpeak, ts.openredstone.org pass: oreftw
I'm on most days between 4&11pm EST as well as most of the time on weekends.
Feel free to ask me anything you like
I'm currently working on the 16th version of a CPU that has my own IS and archiecture. Currently it is a 6tick clock with 5 stages, 5th is write back and ALL register writes can fwd, so the effective throughput is only 4 stages.
Damn nice. Do you stall on branch or flush?
If you flush do you always predict taken or have a predictor?