poke53281 | During I developed the javascript version of the or1k I wondered whether the delayed jump in the architecture has some deeper meaning or is it just an easy way to get rid of some more complex pipeline architecture? | 03:34 |
---|---|---|
poke53281 | Does someone has the answer? | 03:34 |
stekern | poke53281: "the deeper meaning" is that when the architecture was designed (around 2000) delay slots was considered "a good thing" | 05:14 |
stekern | there is work going on making that feature optional | 05:15 |
stekern | i.e. updating the architectural manual, the toolchains and the major software projects using it (e.g. Linux) | 05:16 |
stekern | Peter Gavin is the one that has been pushing this and done most of the job on it | 05:17 |
stekern | There's also a couple of implementations out there without delay slots | 05:17 |
stekern | juliusbs mor1kx prontoespresso for example | 05:18 |
poke53281 | Thanks | 05:28 |
poke53281 | I just realized that there is already a wikipedia article about this. | 05:28 |
stekern | you mean delay slots in general? | 05:37 |
juliusb | stekern: is there an accompanying change to the mor1kx RTL for that recent mor1kx-dev-env rD pull req? | 13:35 |
juliusb | I assume just asking for rD will be wrong, you need to update teh get_gpr() functions in the pipelines, right? | 13:36 |
stekern | yes, you have that in (pronto)espresso right? | 13:38 |
stekern | but I realise I haven't done that change in openrisc/cappuccino | 13:39 |
stekern | only in my pipeline rework | 13:39 |
stekern | ah, it's only in prontoespresso that you have that | 13:42 |
stekern | so, wait a bit with pulling that in then | 13:43 |
stekern | ...until I have an accompanying change in mor1kx for it | 13:44 |
juliusb | nps | 13:56 |
poke53281 | @stekern: Yes, I didn't know this concept before. | 20:18 |
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!