[libre-riscv-dev] Regarding the patch

[Daniel Benusovich]

The change to the macro SV_ELWIDTH_TEST is complete and pushed.
An issue I see is in files such as sv_ld_elwidth.S the instruction are
formatted like so:

> inst   x28, 0(x12);
>

which can pose an issue when trying to standardize the testing macro.
Perhaps adding a second parameter that selects the format of the
instruction to be tested? The parameter field is getting a bit out of hand
already it is at 14! Is adding anymore a good idea even? I imagine a
seperate macro would be written for each format of the instructions:

> code   x28, x15, x12;
>
 inst   x28, 0(x12);
>
And then have the writer pass the macro selection. I wanted to bounce it
off you first before I went into that.

convert the files here one-by-one to litex
>  https://github.com/programmerjake/rv32
>

I saw in the mainling list a reference to migen/litex do you mean this guy
here that is python based?
https://github.com/m-labs/migen

or this guy who is also python based?
https://github.com/timvideos/litex-buildenv

Forgive me as this as the first time I have heard of it was quite recently.
Looks very convenient and promising though!

And as far as converting is concerned keep the same names/structure and
take the logic into the selected language(migen or litex) correct? Is there
anywhere in particular the results should go as of yet?

Hope you guys are having a good day,

Daniel B.

On Sat, Nov 17, 2018 at 2:58 PM Luke Kenneth Casson Leighton <lkcl at lkcl.net>
wrote:

> oh, one other thing: do you want to help evaluate migen / litex for
> use in the soc?  convert the files here one-by-one to litex
>  https://github.com/programmerjake/rv32
>
> just as a proof-of-concept.
>
> l.
>