WIP: AXI_XBAR parameters

[WRoenninger]

This is a first WIP to update axi_xbar according to #153.

I also added and adapted the documentation that was in doc/axi_xbar.md for inline for auto generation. The doc file therefore should probably be removed when this is merged to prevent documentation duplication.

• Renamed ports and paramters of axi_xbar, added inline documentation.
• Added parameter sweep to the xbar portion of scripts/run_vsim.sh
• Renamed TB parameters / added appropriate macros for assign
  • Changed wrong transfer warnings in tb_axi_xbar_pkg to errors

The axi_pkg::xbar_cfg_t struct is still in as it is currently used in axi_lite_xbar also and can be removed when its ports are updated.

Edit: Also when looking at https://pulp-platform.github.io/axi/xbar_parameters/module.axi_xbar.html the link to doc/axi_xbar.png is currently broken. How do I link properly tho the image in the inline documentation?

[andreaskurth]

Very nice, thanks already!

Edit: Also when looking at https://pulp-platform.github.io/axi/xbar_parameters/module.axi_xbar.html the link to doc/axi_xbar.png is currently broken. How do I link properly tho the image in the inline documentation?

You need to move axi_xbar.png into docs/ folder (does not yet exist on master). You can then specify a path relative to that folder.

[WRoenninger]

Added the docs/ folder and copied the axi_xbar.svg/.png to it. Are the now redundant files for the axi_xabr documentation save to delete from the doc/ folder with this PR?

[andreaskurth]

Are the now redundant files for the axi_xabr documentation save to delete from the doc/ folder with this PR?

Yes

[WRoenninger]

Seems however that the link to the .png in line 28 to docs/axi_xbar.png is still broken. How do I specify this properly, or does it only show up on master?

[WRoenninger]

Is there a possibility to check the pipeline status of a failing gitlab-ci? Just to check if the ci failed on a stall due to a functional failing module, or that the timeout time was not long enough and the ci was still simulating.

[andreaskurth]

Seems however that the link to the .png in line 28 to docs/axi_xbar.png is still broken. How do I specify this properly, or does it only show up on master?

You need to remove the docs/ prefix; the relative URL is simply axi_xbar.png.

Is there a possibility to check the pipeline status of a failing gitlab-ci? Just to check if the ci failed on a stall due to a functional failing module, or that the timeout time was not long enough and the ci was still simulating.

Only if you have access to the mirrored repository on which CI runs. For this branch, the timeout was too low, because the crossbar simulated for more than 3 hours. -- Why is that?

[WRoenninger]

Why is that?

The simulation script currently has 5 different parameters which each get checked for two values in the script. This equates to 32 simulation runs. In the tb each instantiated rand_master completes 100 read and 100 write bursts. However the wider tb configurations run quite slow as there is a lot going on (e.g. 6 slv_ports, 9 mst_ports). Adding new the EnableAtops parameter doubled the runtime.

I could tune down the numbers though, e.g. to less slv_ports. I suspect they tend to slow down the simulation quite a bit.

Would it be reasonable to split the script for axi_xbar? So that there is not only one axi_xbar thread in the ci, but multiple. However I think the simulate script is currently not set up for this in a scalable / non-hacky manner. Also adding the internal pipeline from #116 will again double the number of simulations to be run by the ci.

Also similar issue will be in #33, there are at least 5 parameters which should be checked in range by the ci.

[WRoenninger]

Just seen: As is scripts/axi_intercon_gen.py will be broken after implementing these changes.