Initial U-Ray database import #1651

mkurc-ant · 2020-08-28T07:20:38Z

This PR adds a skeleton for Xilinx US/US+ support in SymbiFlow. There were some minor changes to xc/common/cmake files to accomodate X-Ray and U-Ray specifics. I assumed that all db import scripts will be named like prjxray_<something>.py and prjuray_<something>.py. Most of the required files that are not critical to early stages of db import are just templates/placeholders for now.

Apart from all necessary directories and CMakes that allows make env to succeed it also adds the channel formation script. prjuray_form_channels.py. The script has been adjusted to US/US+ specifics (eg. different const network connections) so that it succeeds and the channels.db can be successfully created.

With this PR two submodules were added: prjuray-db and prjuray-tools. The latter one is meant to provide U-Ray database access code after SymbiFlow/prjuray-tools#3 is successfully landed (without it the db import fails).

litghost · 2020-08-28T23:34:07Z

xc/common/cmake/arch_define.cmake

@@ -106,7 +106,7 @@ function(ADD_XC_ARCH_DEFINE)
    CELLS_SIM ${YOSYS_CELLS_SIM} ${VPR_CELLS_SIM}
    VPR_ARCH_ARGS ${VPR_ARCH_ARGS}
    RR_PATCH_TOOL
-      ${symbiflow-arch-defs_SOURCE_DIR}/xc/common/utils/prjxray_routing_import.py
+      ${symbiflow-arch-defs_SOURCE_DIR}/xc/common/utils/${PRJRAY_NAME}_routing_import.py


I'd prefer if don't hard fork these scripts. They aren't small pieces of code to begin with, and I'm not excited about needing to maintain two sets. Can we factor out the xc7 vs xcu specific stuff?

Actually I agree. I took that approach with assign_tile_pin_direction.py.

So the idea would be to have the common code in eg. rounting_import.py plus two wrapper scripts prjxray_routing_import.py and prjuray_routing_import.py that would contain the specific code. These specific scripts would then be used in CMake. Does that make sense?

Sounds good to me.

mkurc-ant · 2020-09-17T14:34:08Z

@litghost I created a doc with brief description of my ideas for the const network: https://docs.google.com/document/d/1Lhkqsa9YTY03ZpmeB6Q2POEYa8Tgm1yIrSBuV5bNADQ

Google Docs - create and edit documents online, for free.
Create a new document and edit with others at the same time -- from your computer, phone or tablet. Get stuff done with or without an internet connection. Use Docs to edit Word files. Free from Google.

litghost · 2020-09-30T20:28:25Z

xc/common/utils/prjuray_timing.py

@@ -0,0 +1,530 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-


Why does this need a full copy/paste?

I'm not entirely happy with that but I didn't wanted any U-Ray scripts depend on X-Ray stuff.

The content of this file is copied from third_party/prjxray/prjxray/timing.py. This code is not present in prjuray-tools so I couldn't import it from there. Since it is not directly related to X-Ray/U-Ray database but rather to how it is handled in arch-defs I decided to put it in arch-defs, at least for U-Ray import for now.

It may be worth to rename this to eg. timing.py and make X-Ray import use it as well since it is common to both archs. Any thoughts on that?

We don't have a uray routing timing model just yet, so can we stub this out, rather than assuming the 7-series and US+ route timing model will be the same?

xc/xcu/CMakeLists.txt

litghost · 2020-09-30T20:30:56Z

xc/common/utils/prjuray_form_channels.py

+# =============================================================================
+
+
+def yield_downstream_nodes(conn, node_pkey):


Why is this specialized for prjuray? Should be the same between prjxray and prjuray?

The function is only used in classify_const_nodes which is prjuray specific so it is not needed in prjxray. Even though that it would work for xray data in the same way.

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

…m happy Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

litghost · 2020-10-06T18:16:00Z

xc/common/utils/prjuray_form_channels.py

@@ -0,0 +1,516 @@
+#!/usr/bin/env python3
+""" Classify 7-series nodes and generate channels for required nodes.


File level comment still needs to be updated? At a minimum 7-series -> UltraScale here?

litghost · 2020-10-06T18:17:17Z

xc/common/utils/prjuray_form_channels.py

+
+def yield_downstream_nodes(conn, node_pkey):
+    """
+    For the given node pkey yields pkeys of all its immediate downstream nodes.


This comment does mention that pseudo pips are avoided, etc. This function comment doesn't provide enough context of what and why this function is doing.

litghost · 2020-10-06T18:18:15Z

xc/common/utils/prjuray_form_channels.py

+
+def identify_const_nodes(conn):
+    """
+    Creates a new table for const nodes named "const _nodes". Identifies them


"const _nodes" -> "const_nodes"?

Rather than add a table, maybe should add a node property on the nodes table?

litghost

Comments in

probot-autolabeler bot added arch-artix7 lang-python Issue uses (or requires) Python language. lang-xml Issues uses (or requires) XML language. third-party type-utils Issues is related to the scripts inside the repo. type-vpr labels Aug 28, 2020

mkurc-ant requested review from acomodi, litghost and tmichalak August 28, 2020 07:22

tmichalak mentioned this pull request Aug 28, 2020

[DNM] WIP: Architecture definition for UltraScale+ #1634

Closed

litghost reviewed Aug 28, 2020

View reviewed changes

mkurc-ant force-pushed the prjuray_channels branch from d606a60 to 572e9d3 Compare September 11, 2020 14:44

mkurc-ant added the kokoro:force-run label Sep 15, 2020

symbiflow-kokoro removed the kokoro:force-run label Sep 15, 2020

mkurc-ant force-pushed the prjuray_channels branch 3 times, most recently from 7854e57 to f627401 Compare September 15, 2020 10:34

mkurc-ant requested a review from litghost September 17, 2020 14:34

mkurc-ant force-pushed the prjuray_channels branch 4 times, most recently from f1eb8c3 to af6d762 Compare September 18, 2020 13:17

mkurc-ant mentioned this pull request Sep 22, 2020

Kokoro disk is full #1670

Closed

mkurc-ant changed the title ~~[WIP] Initial U-Ray database import~~ Initial U-Ray database import Sep 29, 2020

mkurc-ant force-pushed the prjuray_channels branch from 61d84dd to 3c17d4d Compare September 29, 2020 12:48

mkurc-ant mentioned this pull request Sep 29, 2020

[WIP] U-Ray arch import #1683

Closed

litghost reviewed Sep 30, 2020

View reviewed changes

xc/xcu/CMakeLists.txt Outdated Show resolved Hide resolved

litghost reviewed Sep 30, 2020

View reviewed changes

mkurc-ant force-pushed the prjuray_channels branch 2 times, most recently from fed8c13 to 9a80f2f Compare October 5, 2020 15:05

mkurc-ant added 20 commits October 6, 2020 12:33

Added prjuray-db submodule

329f442

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Added prjuray-tools as submodule

7e6857c

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Created prjuray_form_channels.py basing on its X-Ray version

827fa98

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Some fixes to the CMake build system to accomodate U-Ray architecture

f0100c7

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Added a skeleton arch for Zynq UltraScale+

ee8604b

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Isolated common code out of prjxray_form_channels.py

01a4055

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Derived prjuray_form_channels.py

f2b2022

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Added grid range limiting capability to form_channels.py

9259412

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Added placeholder files for prjuray_* scripts to make the build syste…

44b8f39

…m happy Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Bumped prjuray-tools

a54619c

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Integrated grid limit for channel formation into CMake

a5539ef

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Fixed classification of const nodes

597837d

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Code formatting

9d1af50

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Changed const node classification

8443cdc

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Fixed incorrect classification of some const source nodes

c4a3e8a

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Added more placeholder Python scripts

ea50917

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Code formatting

96145f2

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Added skipping of installation of US/US+ devices

8752541

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Bumped prjuray-tools and made it installed by requirements.txt

e297178

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

Updated docstrings in prjuray_form_channels.py

acd94ea

Signed-off-by: Maciej Kurc <mkurc@antmicro.com>

mkurc-ant force-pushed the prjuray_channels branch from 9a80f2f to acd94ea Compare October 6, 2020 11:12

litghost reviewed Oct 6, 2020

View reviewed changes

umarcor closed this Mar 22, 2022

umarcor deleted the prjuray_channels branch March 22, 2022 21:44

mithro mentioned this pull request Apr 29, 2022

Supporting Ultrascale+ is in plan? #2620

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Initial U-Ray database import #1651

Initial U-Ray database import #1651

mkurc-ant commented Aug 28, 2020

litghost Aug 28, 2020

mkurc-ant Sep 1, 2020

litghost Sep 1, 2020

mkurc-ant commented Sep 17, 2020 •

edited by unfurl-links bot

litghost Sep 30, 2020

mkurc-ant Oct 1, 2020

litghost Oct 6, 2020

litghost Sep 30, 2020

mkurc-ant Oct 1, 2020

litghost Oct 6, 2020 •

edited

litghost Oct 6, 2020

litghost Oct 6, 2020

litghost Oct 6, 2020

litghost left a comment

		@@ -0,0 +1,530 @@
		#!/usr/bin/env python3
		# -- coding: utf-8 --

		# =============================================================================


		def yield_downstream_nodes(conn, node_pkey):

		@@ -0,0 +1,516 @@
		#!/usr/bin/env python3
		""" Classify 7-series nodes and generate channels for required nodes.

Initial U-Ray database import #1651

Initial U-Ray database import #1651

Conversation

mkurc-ant commented Aug 28, 2020

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mkurc-ant commented Sep 17, 2020 • edited by unfurl-links bot

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

litghost Oct 6, 2020 • edited

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

litghost left a comment

Choose a reason for hiding this comment

mkurc-ant commented Sep 17, 2020 •

edited by unfurl-links bot

litghost Oct 6, 2020 •

edited