Open-Source MicroBlaze IP-Core working in FPGA :)

I see that you have a benchmark to download. Is there any clock speed info on your design? I would be interested in how fast it can run in an ACEX EP1K part. How many LEs is it? Does it include the 16 bit version?

--
Rick "rickman" Collins

rick.collins@XYarius.com
 Click to see the full signature
Reply to
rickman
Loading thread data ...

I know the NIOS-1 was designed to run in the ACEX parts as well as the Cyclone and others. Altera did not designed NIOS-II for the ACEX parts, I also don't think it comes in a 16 bit version. I have a need for a small, fast MCU. I have been designing my own to use with the Forth language. But the software side is a bit of work. The Altera site mentions some pretty high speeds for NIOS-1, but when you research it the speed doesn't even reach 40 MHz in the ACEX parts. I wondered if the alternate version runs any faster.

I think all of these soft CPUs are a bit larger than what I would like to see. So for now, I'll stick with what I have.

--
Rick "rickman" Collins

rick.collins@XYarius.com
 Click to see the full signature
Reply to
rickman

[snip]

Hi John, here you go!

NIOS-1 GPL2 licensing

formatting link

NIOS-II verilog also exists, but it will not be GPL ;)

Antti

Reply to
Antti Lukats

Hi Rick,

the NIOS-GPL includes CPU C-benchmarks, not FPGA benchmarking or statistics there the NIOS-1 is seems to be a university project - I simply just found it - haha just enter "NIOS verilog source" into google and that the 3rd hit!

the design clamis to be 1:1 drop in replacement for Altera NIOS and it looks that is not 100% clean room implementation, so I guess the GPL license is actual invalid because of that.

The NIOS-1 GPL uses altera memprims, and I havent yet checked it with any synthesis tools.

The NIOS-II, thats different, its completly clean-room design, uses no vendor primitives at all. some statistic are in my latest post about the NIOX core ;) no FPGA tests yet with NIOX (NI*S II compatible) cpu yet. with no optimization the NIOX looks like 60MHz+ in slow Virtex devices.

Antti

Reply to
Antti Lukats

No, but they do show 3 versions, claiming site mentions some pretty high speeds for NIOS-1, but when you research

Part of this is 'creeping featurism', as well as a general industry trend to skip 16 bits, plus the design is easier if the data size matches the PC size. 24 bits is supported in some DSPs, tho probably does not map into the Blocks seen in most FPGAs.

Maybe an 18 bit Opcode and PC would be a better FPGA medium-core ? Would top out at ~4MBit code store, and a soft CPU optimised to run from a serial memory would be an interesting direction.

-jg

Reply to
Jim Granville

I tested the NiosI to 140MHz in my custom 4-layer Cyclone board. (w/ SDRAM to 120MHz) I'm running my NiosII at 75MHz right now, but Quartus says I'm good to at least 100MHz.

The gcc/gdb/Insight tool chain, particulary for NiosII in Legacy Mode, are quite easy and good. I haven't been able to get much out of the new Eclipse based IDE, but a new release is here, and another more major one in Nov/Dec.

NiosII can be had for

Reply to
Kenneth Land

The 700 LE number is in a Cyclone, not an ACEX part. I don't think Altera gives out NIOS-II code, so I don't think you can put it in an ACEX part. I'm also not sure what the NIOX core is, but it appears to be non-public still. The info from Altera says the NIOS-I only runs

Reply to
rickman

Is the NIOX core non-public from UT? Or is this another core?

--
Rick "rickman" Collins

rick.collins@XYarius.com
 Click to see the full signature
Reply to
rickman

with the

I have done a redesign to change some features and have not synthesized it since. Before that it was around 700-800 LEs, IIRC, without optimizing. I expect this is up a bit, but I plan to optimize some parts of the design that don't synthesize well, the muxes mainly. They can map to the cascade chain for best speed and size. I am shooting for

80 MHz if I can improve the muxes and the decode.
600 LEs is a very small design, but I don't know if it will be that small in an ACEX since the LE structure is a bit different from newer parts. The ACEX parts are basically the same as the even older EP10K family.

You mean 18 bit word size, not opcode, right?

--
Rick "rickman" Collins

rick.collins@XYarius.com
 Click to see the full signature
Reply to
rickman

Yes, I guess if it was halfway decent in an older device, Altera would have released that - you can be sure they compiled it, to see :)

Both, as you tend to naturally make the opcode the same size as the word size. 18 bits allows you to do more with the opcodes, but still fit into the FPGA memory tiles, and some external memory is also x18. It also extends the memory reach, so moves the code/data ceiling further up.

-jg

Reply to
Jim Granville

Except that to get so small and fast, these designs have to match the hardware pretty closely. Since the ACEX parts are a different design and I expect much of the NIOS-II uses instantiated logic it would be more work to even fit the design to the part regardless of the resulting speed or size. Even if it didn't run well, if it were not much work, I expect they would release it.

My guess is the main difference in the NIOS-I and -II is that the -II takes advantage of the newer features in the newer chips.

--
Rick "rickman" Collins

rick.collins@XYarius.com
 Click to see the full signature
Reply to
rickman

[snip]

with

NIOX is done by me (3 evnings workload) - its completly clean version written from scratch. It has no pipeline but if running from brams can run about 1 clock instruction except load instructions what is way faster than NIOSII/e

I am testing NIOS-II/uCLinux platform simulator at the moment if that is done then I know how to verify NIOX as well, next step would be NIOX/uCLinux SoC

Antti

Reply to
Antti Lukats

Have you done any place and routes to get an idea of the speed and LE count? Or was that already posted and I missed it? I believe Altera has a version of the NIOS-II that is only 600 LEs. That would make me very happy if I could put it in an ACEX as well as a Spartan.

--
Rick "rickman" Collins

rick.collins@XYarius.com
 Click to see the full signature
Reply to
rickman

run

than

NIOX/uCLinux

I dont have and probably will not have 600LE version. current version is

Reply to
Antti Lukats

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.