| --- Log opened Thu Jun 30 00:00:17 2016 | ||
| wallento | SMDhome: Yes, the stuff is still there | 01:17 |
|---|---|---|
| wallento | we need to migrate this information to the github repo for the spec | 01:17 |
| SMDhome | ZipCPU that's an extention | 01:34 |
| -!- Netsplit *.net <-> *.split quits: a0u, Empyrium | 05:08 | |
| -!- Empyrium_ is now known as Empyrium | 05:21 | |
| ZipCPU | SMDhome: Thanks! That's what I was looking for. Does that mean that a double precision hardware floating point unit exists? | 07:46 |
| ZipCPU | SMDhome: Are the 64-bit integers and vector instructions also optional extensions? | 07:48 |
| olof | ZipCPU: Yes they are | 07:50 |
| ZipCPU | Thanks! | 07:51 |
| olof | Most or1k implementations only implement ORBIS32 (the 32-bit integer parts) | 07:51 |
| ZipCPU | The ZipCPU isn't all that different--lots of configurability, although I will admit that or1k appears to be much more so. ;) | 07:51 |
| olof | :) | 08:00 |
| ZipCPU | So here's what I'm up to: I'm trying to put together a simplified comparison between a variety of soft cores. You can see how far | 08:00 |
| ZipCPU | I've gotten at opencores.org, in the ZipCPU project: trunk/doc/orconf.pdf. | 08:00 |
| ZipCPU | Feel free to let me know if I've in anyway misrepresented, or misaccounted, any platform listed there. | 08:00 |
| ZipCPU | (I'm comparing: Nios (2), uBlaze, ECO32, RISCV, and OpenRISC ...) | 08:01 |
| olof | 219 instructions? Is that for real? :) | 08:05 |
| ZipCPU | Yeah. That was my thought when I went through and counted. | 08:05 |
| ZipCPU | In the end, I counted pages in the architecture document, and noted that every page had an instruction on it. | 08:06 |
| olof | ok, looks like ORBIS32 has around 100 instructions | 08:11 |
| olof | Most of them seem to be valid for ORBIS64 as well | 08:11 |
| olof | Then there's about 30 FP instructions | 08:12 |
| ZipCPU | How should I count them? Are instructions listed in the openrisc-arch-1.1-rev0.pdf more than once? | 08:12 |
| olof | And around 90 vector instructions. No wonder that no one has implemented the vector parts yet :) | 08:15 |
| olof | I think a fair comparasion would be to count the ones in orbis32. | 08:16 |
| olof | But it's hard. Did you count the vector instructions for uBlaze? | 08:16 |
| ZipCPU | I can do that. Should I list the others in a range? For example, 100-219? | 08:17 |
| ZipCPU | Do you know how to identify microblaze vector instructions from within the rest of the set? | 08:18 |
| olof | Sorry, no. Haven't done uBlaze stuff for quite some tim | 08:18 |
| olof | e | 08:18 |
| olof | It's up to you how you want to present them, but for OpenRISC you can say that the core ISA is ~100 instructions, FP an additional 30, and vectors 90 extra | 08:20 |
| ZipCPU | I count 99 ORBIS32/ORBIS64 instructions pages, so I'll use 99 as the lower end. | 08:20 |
| olof | I guess it's the same for RISC-V | 08:20 |
| olof | Which parts of the RISC-V ISA did you use? | 08:20 |
| olof | IMD? | 08:20 |
| olof | No, sorry | 08:20 |
| olof | I mean IMA | 08:20 |
| olof | ...I think | 08:20 |
| olof | Integer, Mul/div and atomic. That's roughly the same as ORBIS32 | 08:21 |
| ZipCPU | Just the bare basic instruction set--the minimal set that needed to be implemented. | 08:21 |
| olof | aha. Then you can cut out some more for openrisc. mul, div, sra, ff1, fl1 and cust1-6 are for example not required | 08:21 |
| olof | Also not l.swa and l.lwa | 08:21 |
| olof | Actually, I think only those with "Instruction Class ORBIS32 I" at the bottom of the page are required | 08:23 |
| ZipCPU | Okay, that gets you down to 48, if I've counted right. (I missed where that instruction class was noted, I looked all over the top of the page for it ...) | 08:25 |
| olof | I think I counted ~48 class one instructions | 08:25 |
| olof | Yeah, it was a bit hard to find. I just searched for "class", and hoped to find something :) | 08:26 |
| ZipCPU | And here I didn't think the process had any "class" (sorry, bad joke ...). | 08:27 |
| olof | ...moving on... | 08:29 |
| ZipCPU | Okay, so the number is 48 (minimal), 99+30+vector (full) | 08:29 |
| ZipCPU | And the floating point and vector are extensions, as must be the other 51 basic instructions. | 08:30 |
| olof | Sounds about right | 08:31 |
| olof | Any of you globetrotters who know if I will be able to use a phone bought in the US when I'm back in Sweden? | 08:35 |
| ZipCPU | Don't look at me--I'm not much of a globetrotter, so I wouldn't know. | 08:37 |
| ZipCPU | Hmm ... 48,99+30+91 doesn't make sense if I don't similarly count the RISC-V optional instruction sets. | 08:39 |
| ZipCPU | I might just label RISCV as 50(min), and OpenRISC then as 48(min). | 08:40 |
| olof | I guess that minimum is the slightly more important number in this comparasion | 08:43 |
| ZipCPU | Yes. Especially for what I am trying to accomplish. | 08:44 |
| ZipCPU | I mean, consider, the OR1K spec outlines over 500+ special purpose registers (gasp!). Again, I tried to pick the minimum there as well. | 08:46 |
| wallento | olof: yes, mostly ;) | 08:46 |
| ZipCPU | ;) | 08:46 |
| wallento | it should be quad band and not be bought at an operator | 08:47 |
| wallento | than it should work | 08:47 |
| olof | I was thinking of retiring my Nokia 105 and by some Android phone. Thought it might be cheaper in the US | 08:49 |
| olof | s/by/buy | 08:50 |
| olof | Oh... I never realized that the OpenOCD server at port 4444 is a full TCL interpreter | 08:50 |
| olof | That could had saved me from a LOT of typing over the years | 08:53 |
| wallento | olof: I think its not much cheaper | 09:06 |
| wallento | especially now with the weakened euro | 09:07 |
| olof | wallento: We're not a euro country | 09:13 |
| olof | But still. The dollar is stronger than in a while | 09:14 |
| olof | oh.. I'm not buying stuff in dollars right now. It was worse than I thought | 09:15 |
| wallento | ah, right, forgot | 09:15 |
| wallento | but technology is evenly priced in us, eu and japan now I think | 09:16 |
| olof | Yeah, I guess you need to go to Switzerland is you want to get really high prices | 09:17 |
| olof | Into the darkness again. | 10:43 |
| olof | Really wish they would fix my home internet connection soon | 10:43 |
| -!- Netsplit *.net <-> *.split quits: fotis2, wallento, nurelin | 17:45 | |
| --- Log closed Fri Jul 01 00:00:18 2016 | ||
Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!