Changes from NRAO

[mschiller-nrao]

1. Implemented Twiddle generation in VHDL
2. Adding files for Sigasi library mappings
3. Fixed some compile issues in Questa
4. Added gitlab CI configs (mostly for NRAO internal use)
5. Added testbenches for Twiddle Generation
6. A python module for FFT verification
7. Rounding updates to implement round 2 even.

Sorry about the churn on this pull request... I wasn't happy with what I managed to do in GIT I had 1000's of changes as compared to your branch. Which I have no idea how happened.

[mschiller-nrao]

r2sdf Bit accurate model implemented in new r2sdf testbench!

[mschiller-nrao]

This is my last checkin for a bit, have to get back to document writing. But the ngVLA configurations are working with versal and agileX support.

Features implemented:

1. Versal/AgileX complex multiplier (the new dsps support special modes for complex multiplies)
2. Inferred Twiddles
3. Bit accurate model is implemented

Features still planned for "someday" on ngVLA:

1. DIF/DIT support (eg ability to take bit-revorder inputs)
2. Fix reorder support for wb_factor=16, 8192 points, which appears to not work currently
3. IFFT support (pretty trivial)
4. Possibly elimination of XPM/AlteraMF memories for inferred versions.

But those will have to wait for a later time (and a later pull-request)

[talonmyburgh]

Created origin/pre-nrao-merge branch to retain state pre merge from NRAO. Will address bugs if/when they crop up.

[talonmyburgh]

@mschiller-nrao having issue with Vivado support for:

use ieee.fixed_float_types.all;
use ieee.fixed_pkg.all;

Inside of common_pkg.vhd.

Appears that this functionality is not supported in Vivado's 2008 feature set. I've tried this solution but had no luck.

Unfortunately due to how integrated CASPER is with Vivado, if Vivado won't support it, we need to implement a manual workaround. Any ideas? Otherwise I can give it a whirl.

[talonmyburgh]

Could try something like:

function round_even (x : real) return integer is
begin
    if x >= 0.0 then
        return integer(x + 0.5);
    else
        return integer(x - 0.5);
    end if;
end function round_even;

and have an additional overload-type function that accepts slv input?

[mschiller-nrao]

What version of vivado are you using? Because that works in vivado 2022.2 (I’ve compiled this through synthesis for the Versal premium), so I’m assuming you are using an older version? Is the file loaded as “vhdl-2008”? That’s critical to making fixed pkg work as intended. From: Talon ***@***.***> Sent: Monday, May 8, 2023 10:12 AM To: talonmyburgh/casper_dspdevel ***@***.***> Cc: Matthew Schiller ***@***.***>; Mention ***@***.***> Subject: Re: [talonmyburgh/casper_dspdevel] Changes from NRAO (PR #18) {External} @mschiller-nrao<https://github.com/mschiller-nrao> having issue with Vivado support for: use ieee.fixed_float_types.all; use ieee.fixed_pkg.all; Inside of common_pkg.vhd. Appears that this functionality is not supported in Vivado's 2008 feature set. I've tried this <https://support.xilinx.com/s/question/0D52E00006hpbL3SAI/cannot-find-fixedpkg-in-ieee-or-ieeeproposed?language=en_US> solution but had no luck. Unfortunately due to how integrated CASPER is with Vivado, if Vivado won't support it, we need to implement a manual workaround. Any ideas? Otherwise I can give it a whirl. — Reply to this email directly, view it on GitHub<#18 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A6RX4EFOFXALDXKFTLPVZJLXFD5MFANCNFSM6AAAAAAWHZ62LM>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>> [ { ***@***.***": "http://schema.org", ***@***.***": "EmailMessage", "potentialAction": { ***@***.***": "ViewAction", "target": "#18 (comment)", "url": "#18 (comment)", "name": "View Pull Request" }, "description": "View this Pull Request on GitHub", "publisher": { ***@***.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[talonmyburgh]

Ah... been using Vivado 2020.1 since that has good compatability with MATLAB 2019a and the casper toolflow.

I have been loading the file as VHDL-2008. Must just be the Vivado version :/

[mschiller-nrao]

I hate vendors. And I hate tools That feature has been in the vhdl standard since… 2008… I’m assuming it’s saying the package isn’t even available? So you should be able to download the fixed point libraries from: https://opensource.ieee.org/vasg/Packages/-/tree/a03b96d46e0fa28df73b7e259861a5ff596c23a5/ieee and compile them in. I’m pretty sure I’ve made that work before. I don’t have Vivado 2020.1 installed but maybe I’ll start downloading it to help figure this out. From: Talon ***@***.***> Sent: Monday, May 8, 2023 10:24 AM To: talonmyburgh/casper_dspdevel ***@***.***> Cc: Matthew Schiller ***@***.***>; Mention ***@***.***> Subject: Re: [talonmyburgh/casper_dspdevel] Changes from NRAO (PR #18) {External} Ah... been using Vivado 2020.1 since that has good compatability with MATLAB 2019a and the casper toolflow. I have been loading the file as VHDL-2008. Must just be the Vivado version :/ — Reply to this email directly, view it on GitHub<#18 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A6RX4EFGI6KZKFH4EPYB2MTXFD6XFANCNFSM6AAAAAAWHZ62LM>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>> [ { ***@***.***": "http://schema.org", ***@***.***": "EmailMessage", "potentialAction": { ***@***.***": "ViewAction", "target": "#18 (comment)", "url": "#18 (comment)", "name": "View Pull Request" }, "description": "View this Pull Request on GitHub", "publisher": { ***@***.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[talonmyburgh]

Preaching to the crowd here :') If I could place and route with nextpnr I'd do away with Vivado all together.

Thanks. I'll add you as a contributor so you can push directly to this repository.

[talonmyburgh]

Some actions also failed to successfully run. Will address those soon too!

[mschiller-nrao]

I'm not surprised, I never got github actions (something about github secrets that I didn't understand and didn't take the time to figure out) to work on my forked repo. So I couldn't test them (there's some weirdness on Actions and forked repos on github).

But those should be fixable, since internally the code works on gitlab.

The only thing I never got to work was the casper_delay tests, but those were broken in my test environment when I forked... So I assumed they were broken in github actions too.... But if not... Ugh more tool problems.

[talonmyburgh]

Nah the casper_delay had some known issues - not your stuff :)

[talonmyburgh]

@mschiller-nrao can I leave it for you to add the necessary files from here to this repository to make the fixed point stuff work?

Best method may just be to submodule it into the repository - which I'm happy to do if you agree? We had to do some submoduling of XPM libraries for GHDL.

[mschiller-nrao]

Yeah I'll have to download 2020.1 (which is slowly downloading...) to be sure I get it right... But if you want to just hold off, I should be able to address this in the next 24 hours or so.

[talonmyburgh]

No rush. Thank you!

[mschiller-nrao]

Ugh I hate tool problem...

Good news: My Test design works
Bad News: I didn't change anything...... (other than selecting a random ultrascale+ part and using vivado 2020.1)

I'm hoping the problem is I have it in "versal" mode using the new tech_select_package files I added. Hopefully it'll break in "xpm" mode which is what I assume you are using....... Let me try that.

(I have c_test_select_default = 3, in this test build, and I'm going to change it back to 1)

[mschiller-nrao]

Ugh so in my VHDL only test design it seems to always work.... So that's fun...
project_1_test_casper.xpr.zip

A Vivado archive of the project is attached if you want to try buidling it on your machine to rule out something there.

I guess the next step is trying to figure out how your fancy matlab stuff works.. But I don't think I have the right matlab version =c(
I've

[mschiller-nrao]

And I don't have a copy of simulink......... =c( I might be stuck @talonmyburgh I don't seem to be able to replicate the problem.

Assuming I'm at the same version of Vivado (see screencapture above) I think we're looking at a project file settings... is all your casper VHDL files set to 2008? or just a few of them?

[talonmyburgh]

@mschiller-nrao Just a few of them we made VHDL2008. @mschiller-nrao could you try with the committed Vivado projects:

• for wbfft
• for wb_pfb
• for r2sdf_fft
• for filterbank

And try synthesize those? I had issue synthesizing the wbfft project (as is) then changed common_pkg.vhd to be VHDL2008 and still had issue.

Sorry I have not documented the locations of these project files - probably would've made your life easier

[mschiller-nrao]

AHHH So that's where the top levels are! So I did start to get failures with those top level projects.. So I'm working on fixing WBfft currently and then will move on to the others.. Sorry I didn't see those checked in projects before.. Matthew Schiller

…

________________________________ From: Talon ***@***.***> Sent: Tuesday, May 9, 2023 11:59:17 AM To: talonmyburgh/casper_dspdevel Cc: Matthew Schiller; Mention Subject: Re: [talonmyburgh/casper_dspdevel] Changes from NRAO (PR #18) {External} Just a few of them we made VHDL2008. @mschiller-nrao<https://github.com/mschiller-nrao> could you try with the committed Vivado projects: * for wbfft<https://github.com/talonmyburgh/casper_dspdevel/tree/master/casper_wb_fft/casper_wb_fft_vivproj> * for wb_pfb<https://github.com/talonmyburgh/casper_dspdevel/tree/master/casper_wbpfb/wbpfb_vivproj> * for r2sdf_fft<https://github.com/talonmyburgh/casper_dspdevel/tree/master/r2sdf_fft/r2SDF_FFT_proj> * for filterbank<https://github.com/talonmyburgh/casper_dspdevel/tree/master/casper_filter/casper_filter_proj> And try synthesize those? I had issue synthesizing the wbfft project (as is) then changed common_pkg.vhd to be VHDL2008 and still had issue. Sorry I have not documented the locations of these project files - probably would've made your life easier — Reply to this email directly, view it on GitHub<#18 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A6RX4EHKZAQSKDKBVRHOEBLXFJSVLANCNFSM6AAAAAAWHZ62LM>. You are receiving this because you were mentioned.Message ID: ***@***.***> [ { ***@***.***": "http://schema.org", ***@***.***": "EmailMessage", "potentialAction": { ***@***.***": "ViewAction", "target": "#18 (comment)", "url": "#18 (comment)", "name": "View Pull Request" }, "description": "View this Pull Request on GitHub", "publisher": { ***@***.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[mschiller-nrao]

Alright I fixed one:

Those files were the ones that needed to be 2008, but there was also some type non-sense because I made one of the types more generic with a record of unconstrained std_logic_vector.

But with unconstrained std_logic_vector the top level (ish) needs to declare the size in the signal declaration. That should be fixed too.. so your top level declares the proper sizes i wb_fft.

I'll look at fixing the other ones and then check this all in.

[talonmyburgh]

Thanks a tonne @mschiller-nrao !

[mschiller-nrao]

So yesterday, I fell off my bicycle and broke both my arms. I'm going to check in the progress I have made, but some of the test benches are still failing because changes I made to the code haven't been. Implemented in the test benches. You should be able to find examples in. The actual place and routable top levels on how to fix the code in the testbenches. I will try to get to it when I'm feeling a little bit better. But right now I'm typing via voice recognition. Voice recognition and coding don't really get along too well.

…

________________________________ From: Talon ***@***.***> Sent: Friday, May 12, 2023 6:15:03 AM To: talonmyburgh/casper_dspdevel Cc: Matthew Schiller; Mention Subject: Re: [talonmyburgh/casper_dspdevel] Changes from NRAO (PR #18) {External} Thanks a tonne @mschiller-nrao<https://github.com/mschiller-nrao> ! — Reply to this email directly, view it on GitHub<#18 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A6RX4EGWORJYHO2G7SZNPR3XFYESPANCNFSM6AAAAAAWHZ62LM>. You are receiving this because you were mentioned.Message ID: ***@***.***> [ { ***@***.***": "http://schema.org", ***@***.***": "EmailMessage", "potentialAction": { ***@***.***": "ViewAction", "target": "#18 (comment)", "url": "#18 (comment)", "name": "View Pull Request" }, "description": "View this Pull Request on GitHub", "publisher": { ***@***.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[talonmyburgh]

Oh yikes! Don't worry I made a branch pre the changes so I'm working off of that. I can attempt to make some fixes to the master branch but am constrained for time with my own project.
Hope you get better soon!

[mschiller-nrao]

there testbench works again on on my machine

[talonmyburgh]

Thanks @mschiller-nrao I'm surprised/pleased you're already back at it. I hope you are healing well.

Please shout if there is anything you need from me to get things working. I understand Github Actions are a pain and occasionally with run for a while due to scheduling. I'd maybe recommend branching, testing locally and then occasionally adding PR's so we don't use up our action minutes - though I don't imagine we will.

[mschiller-nrao]

I gotta make my local have the same version of vunit as github…. It clearly isn’t because the issues are related to: VUnit/vunit#777 For now I’m not going to work on it more….. to avoid using all the minutes.. But I fixed the polyphase filter bank with updates to run.py Matthew Schiller From: Talon ***@***.***> Sent: Friday, May 19, 2023 10:01 AM To: talonmyburgh/casper_dspdevel ***@***.***> Cc: Matthew Schiller ***@***.***>; Mention ***@***.***> Subject: Re: [talonmyburgh/casper_dspdevel] Changes from NRAO (PR #18) {External} Thanks @mschiller-nrao<https://github.com/mschiller-nrao> I'm surprised/pleased you're already back at it. I hope you are healing well. Please shout if there is anything you need from me to get things working. I understand Github Actions are a pain and occasionally with run for a while due to scheduling. I'd maybe recommend branching, testing locally and then occasionally adding PR's so we don't use up our action minutes - though I don't imagine we will. — Reply to this email directly, view it on GitHub<#18 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A6RX4ECPIUMGP7CRIAATPNTXG54JDANCNFSM6AAAAAAWHZ62LM>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>> [ { ***@***.***": "http://schema.org", ***@***.***": "EmailMessage", "potentialAction": { ***@***.***": "ViewAction", "target": "#18 (comment)", "url": "#18 (comment)", "name": "View Pull Request" }, "description": "View this Pull Request on GitHub", "publisher": { ***@***.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[mschiller-nrao]

The other testbenches aren’t running right now (skipped). But it wouldn’t surprise me if they are broken as well for the same issue Matthew Schiller From: Matthew Schiller Sent: Friday, May 19, 2023 10:05 AM To: 'talonmyburgh/casper_dspdevel' ***@***.***>; talonmyburgh/casper_dspdevel ***@***.***> Cc: Mention ***@***.***> Subject: RE: [talonmyburgh/casper_dspdevel] Changes from NRAO (PR #18) {External} I gotta make my local have the same version of vunit as github…. It clearly isn’t because the issues are related to: VUnit/vunit#777 For now I’m not going to work on it more….. to avoid using all the minutes.. But I fixed the polyphase filter bank with updates to run.py Matthew Schiller From: Talon ***@***.******@***.***>> Sent: Friday, May 19, 2023 10:01 AM To: talonmyburgh/casper_dspdevel ***@***.******@***.***>> Cc: Matthew Schiller ***@***.******@***.***>>; Mention ***@***.******@***.***>> Subject: Re: [talonmyburgh/casper_dspdevel] Changes from NRAO (PR #18) {External} Thanks @mschiller-nrao<https://github.com/mschiller-nrao> I'm surprised/pleased you're already back at it. I hope you are healing well. Please shout if there is anything you need from me to get things working. I understand Github Actions are a pain and occasionally with run for a while due to scheduling. I'd maybe recommend branching, testing locally and then occasionally adding PR's so we don't use up our action minutes - though I don't imagine we will. — Reply to this email directly, view it on GitHub<#18 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A6RX4ECPIUMGP7CRIAATPNTXG54JDANCNFSM6AAAAAAWHZ62LM>. You are receiving this because you were mentioned.Message ID: ***@***.******@***.***>> [ { ***@***.***": "http://schema.org", ***@***.***": "EmailMessage", "potentialAction": { ***@***.***": "ViewAction", "target": "#18 (comment)", "url": "#18 (comment)", "name": "View Pull Request" }, "description": "View this Pull Request on GitHub", "publisher": { ***@***.***": "Organization", "name": "GitHub", "url": "https://github.com" } } ]

[talonmyburgh]

Excellent thanks!

Also feel free to approve your own PR's when you see fit. Arguably more important than VUnit support is the Vivado 2020.1 project support that we have... so please try maintain that.