OpenROAD is a tool to build a chip from a synthesized netlist to a physical design for manufacturing.
The unifying principle behind the design of OpenROAD is for all of the tools to reside in one tool, with one process, and one database. All tools in the flow should use Tcl commands exclusively to control them instead of external “configuration files”. File based communication between tools and forking processes is strongly discouraged. This architecture streamlines the construction of a flexible tool flow and minimizes the overhead of invoking each tool in the flow.
Tool File Organization¶
Every tool follows the following file structure.
CMakelists.txt - add_subdirectory's src/CMakelists.txt src/ - sources and private headers src/CMakelists.txt include/<toolname>/ - exported headers test/ test/regression
CMakeLists.txt - top level cmake file src/Main.cc src/OpenROAD.cc - OpenROAD class functions src/OpenROAD.i - top level swig, %includes tool swig files src/OpenROAD.tcl - basic read/write lef/def/db commands src/OpenROAD.hh - OpenROAD top level class, has instances of tools
Submodule repos in /src (note these are NOT in src/module)
OpenDB OpenSTA replace ioPlacer FastRoute TritonMacroPlace OpenRCX flute3 eigen
Submodules that are shared by multiple tools are owned by OpenROAD so that there are not redundant source trees and compiles.
Each tool submodule cmake file builds a library that is linked by the
OpenROAD application. The tools should not define a
If the tool is tcl only and has no c++ code it does not need to have
a cmake file.
None of the tools have commands to read or write LEF, DEF, Verilog or database files. These functions are all provided by the OpenROAD framework for consistency.
Tools should package all state in a single class. An instance of each tool class resides in the top level OpenROAD object. This allows multiple tools to exist at the same time. If any tool keeps state in global variables (even static) only one tool can exist at a time. Many of the tools being integrated were not built with this goal in mind and will only work on one design at a time. Eventually all of the tools should be upgraded to remove this deficiency as they are re-written to work in the OpenROAD framework.
Each tool should use a unique namespace for all of its code. The same namespace should be used for any Tcl commands. Internal Tcl commands stay inside the namespace, user visible Tcl commands will be exported to the global namespace. User commands should be simple Tcl commands such as ‘global_place_design’ that do not create tool instances that must be based to the commands. Defining Tcl commands for a tool class is fine for internals, but not for user visible commands. Commands have an implicit argument of the current OpenROAD class object. Functions to get individual tools from the OpenROAD object can be defined.
Initialization (c++ tools only)¶
The OpenRoad class only has pointers to each tools with functions to get each tool. Each tool has (at a minimum) a function to make an instance of the tool class, and an initialization function that is called after all of the tools have been made, and a funtion to delete the tool. This small header does NOT include the class definition for the tool so that the OpenRoad framework does not have to know anything about the tool internals or include a gigantic header file.
MakeTool.hh defines the following:
Tool *makeTool(); void initTool(OpenRoad *openroad); void deleteTool(Tool *tool);
The OpenRoad::init() function calls all of the makeTool functions and then all of the initTool() functions. The init functions are called from the bottom of the tool dependences. Each init function grabs the state it needs out of the OpenRoad instance.
Tools should provide Tcl commands to control them. Tcl object based tool interfaces are not user friendly. Define Tcl procedures that take keyword arguments that reference the OpenRoad object to get tool state. OpenSTA has Tcl utilities to parse keyword arguements (sta::parse_keyword_args). See OpenSTA/tcl/*.tcl for examples. Use swig to define internal functions to C++ functionality.p
Tcl files can be included by encoding them in cmake into a string that is evaluated at run time (See Resizer::init()).
Each “tool” has a /test directory containing a script nameed “regression” to run “unit” tests. With no arguments it should run default unit tests.
No database files should be in tests. Read LEF/DEF/Verilog to make a database.
The regression script should not depend on the current working directory. It should be able to be run from any directory. Use filenames relative to the script name rather the the current working directory.
Regression scripts should print a consise summary of test failures. The regression script should return an exit code of zero if there are no errors and 1 if there are errors. The script should not print thousands of lines of internal tool info.
Checking out the OpenROAD repo with –recursive installs all of the OpenRoad tools and their submodules.
git clone --recusive https://github.com/The-OpenROAD-Project/OpenROAD.git cd OpenROAD mkdir build cd build cmake .. make
All tools build using cmake and must have a CMakeLists.txt file in their tool directory.
This builds the ‘openroad’ executable in /build.
Note that removing submodules from a repo when moving it into OpenROAD is less than obvious. Here are the steps:
git submodule deinit <path_to_submodule> git rm <path_to_submodule> git commit-m "Removed submodule " rm -rf .git/modules/<path_to_submodule>
Tool Work Flow¶
To work on one of the tools inside OpenROAD when it is a submodule requires updating the OpenROAD repo to integrate your changes. Submodules point to a specific version (hash) of the submodule repo and do not automatically track changes to the submodule repo.
Work on OpenROAD should be done in the
To make changes to a submodule, first check out a branch of the submodule (git clone –recursive does not check out a branch, just a specific commit).
cd src/<tool> git checkout <branch>
<branch> is the branch used for development of the tool when it is inside
OpenROAD. The convention is for
After making changes inside the tool source tree, stage and commit them to the tool repo and push them to the remote repo.
git add ... git commit -m "massive improvement" git push
If instead you have done development in a different branch or source tree, merge those changes into the branch used for OpenROAD.
Once the changes are in the OpenROAD submodule source tree it will show them as a diff in the hash for the directory.
cd openroad git stage <tool_submodule_dir> git commit -m "merge tool massive improvement" git push
Example of Adding a Tool to OpenRoad¶
The branch “add_tool” illustrates how to add a tool to OpenRoad. Use
git checkout add_tool to checkout the branch. To see the changes
between OpenRoad with and without Tool use
git diff master.
This adds a directory OpenRoad/src/tool that illustrates a tool named “Tool” that uses the file structure described and defines a command to run the tool with keyword and flag arguments as illustrated below:
% toolize foo Helping 23/6 Gotta pos_arg1 foo Gotta param1 0.000000 Gotta flag1 false % toolize -flag1 -key1 2.0 bar Helping 23/6 Gotta pos_arg1 bar Gotta param1 2.000000 Gotta flag1 true % help toolize toolize [-key1 key1] [-flag1] pos_arg1
Tool commands should be documented in the top level OpenROAD README.md file. Detailed documentation should be the tool/README.md file.
Verilog to DB (dbSTA)
Init Floorplan (OpenROAD)
I/O placement (ioPlacer)
PDN generation (pdngen
Tapcell and Welltie insertion (tapcell with LEF/DEF)
I/O placement (ioPlacer)
Global placement (RePlAce)
Gate Resizing and buffering (Resizer)
Detailed placement (OpenDP)
Clock Tree Synthesis (TritonCTS)
Repair Hold Violations (Resizer)
Global route (FastRoute)
Detailed route (TritonRoute)n
Final timing/power report (OpenSTA)
OpenROAD submodules reference tool
CMakeLists.txt does not use glob. https://gist.github.com/mbinna/c61dbb39bca0e4fb7d1f73b0d66a4fd1
No main.cpp or main procedure.
No compiler warnings for gcc, clang with optimization enabled.
Does not call flute::readLUT (called once by OpenRoad).
Tcl command(s) documented in top level README.md in flow order.
Command line tool documentation in tool README.
Conforms to Tcl command naming standards (no camel case).
Does not read configuration files.
Use command arguments or support commands.
.clang-formatat tool root directory to aid foreign programmers.
No jenkins/, Jenkinsfile, Dockerfile in tool directory.
regression script named “test/regression” with default argument that runs tests. Not tests/regression-tcl.sh, not test/run_tests.py etc.
Regression runs independent of current directory.
Regression only prints test results or summary, does not belch 1000s of lines of output.
Test scripts use OpenROAD tcl commands (not itcl, not internal accessors).
Regressions report no memory errors with valgrind.
Regressions report no memory leaks with valgrind (difficult).
James Cherry, Dec 2019