[top_darjeeling] Instantiate AC Range checks #25862
Conversation
Razer6
requested review from
hcallahan-lowrisc,
moidx,
rswarbrick,
andreaskurth and
vogelpi
and removed request for
a team,
cfrantz,
hcallahan-lowrisc and
moidx
|
I don't think splitting up makes sense. The one big commit instantiates the ac ranges on the HJSON (5 lines or so) and this triggers ipgen to copy over the IP. If we split it up, there would be inconsistent IPs |
Razer6
force-pushed
the
ac-ranges-instance
branch
from
4066f22 to
112e519
Compare
|
@rswarbrick @vogelpi Since #25894 is merged, this now correctly maps the AC ranges. Can you take another look? |
Razer6
force-pushed
the
ac-ranges-instance
branch
3 times, most recently
from
17c5cb1 to
c9e2994
Compare
There was a problem hiding this comment.
A thought for the future: Perhaps we'll want to formally introduce a TL-UL to TL-UL bus bridge with base address translation in hw/ip/tlul at some point. Then we can extend topgen / tlgen and do traversals to understand reachability.
hw/top_darjeeling/ip_autogen/ac_range_check/data/ac_range_check.hjson
Outdated
Show resolved
Hide resolved
|
Oh and CI flags a failing whitespace check: |
Razer6
force-pushed
the
ac-ranges-instance
branch
from
c9e2994 to
59a9d7a
Compare
Razer6
force-pushed
the
ac-ranges-instance
branch
2 times, most recently
from
ff8dbc9 to
4f905a6
Compare
Razer6
force-pushed
the
ac-ranges-instance
branch
2 times, most recently
from
5838567 to
af82526
Compare
vogelpi
force-pushed
the
ac-ranges-instance
branch
5 times, most recently
from
5ddc911 to
616c034
Compare
Razer6
force-pushed
the
ac-ranges-instance
branch
2 times, most recently
from
3259101 to
b7b8e54
Compare
Signed-off-by: Robert Schilling <[email protected]>
… name Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Signed-off-by: Robert Schilling <[email protected]> Co-authored-by: Pirmin Vogel <[email protected]>
Signed-off-by: Robert Schilling <[email protected]> Co-authored-by: Pirmin Vogel <[email protected]>
This adds a base address translation (BAT) on the muxed host TLUL port before exiting to the AC ranges. This BAT removes the 1-2GB offset and downlifts the address to 0-1GB when exiting. Co-authored-by: Robert Schilling <[email protected]> Signed-off-by: Pirmin Vogel <[email protected]>
Signed-off-by: Pirmin Vogel <[email protected]>
Signed-off-by: Robert Schilling <[email protected]>
Razer6
force-pushed
the
ac-ranges-instance
branch
from
b7b8e54 to
55f0ae0
Compare
|
Thanks for cleaning up ac_range_check.sv @Razer6 ! |
|
|
||
| ipgen_render("ac_range_check", topname, params, out_path) | ||
| ipgen_render('ac_range_check', topname, params, out_path) |
There was a problem hiding this comment.
Can we not have quote vs double quotes discrepancies? The overwhelming number of strings are double quoted, and we have the policy of preserving the prevailing convention. You can see topcfg/validate.py is mostly single quotes.
| @@ -2,6 +2,7 @@ | |||
| // Licensed under the Apache License, Version 2.0, see LICENSE for details. | |||
| // SPDX-License-Identifier: Apache-2.0 | |||
| ${gencmd} | |||
| <% import textwrap %>\ | |||
| <% racl_role_vec_len = 2 ** racl_config['nr_role_bits'] %>\ | |||
|
|
|||
| package top_racl_pkg; | |||
There was a problem hiding this comment.
Weird this doesn't get some uniquifying name. The name top_racl_pkg can become confusing in multi-top.
This PR instantiates the AC ranges for a single egress port. All Egress ports are muxed to a single egress port before leaving Darjeeling. An additional external miscellaneous port is added to the system that is coming from the external uncore but muxed with same mux and going through the AC ranges.