Merge branch 'master' of https://github.com/verilog-to-routing/vtr-verilog-to-routing into fix_3d_delay

Author: amin1377

.github/scripts/install_dependencies.sh

@@ -29,6 +29,7 @@ sudo apt install -y \
   libncurses5-dev \
   libx11-dev \
   libxft-dev \
+  libxml2-utils \
   libxml++2.6-dev \
   libreadline-dev \
   tcllib \

.github/workflows/test.yml

@@ -197,6 +197,11 @@ jobs:
             params: '-DVTR_ASSERT_LEVEL=3 -DWITH_BLIFEXPLORER=on -DVPR_USE_EZGL=off',
             suite: 'vtr_reg_basic'
           },
+          {
+            name: 'Basic with CAPNPROTO disabled',
+            params: '-DVTR_ASSERT_LEVEL=3 -DWITH_BLIFEXPLORER=on -DVTR_ENABLE_CAPNPROTO=off',
+            suite: 'vtr_reg_basic'
+          },
           {
             name: 'Basic with VTR_ENABLE_DEBUG_LOGGING',
             params: '-DVTR_ASSERT_LEVEL=3 -DWITH_BLIFEXPLORER=on -DVTR_ENABLE_DEBUG_LOGGING=on',
@@ -297,7 +302,13 @@ jobs:
         CMAKE_PARAMS: ${{ matrix.params }}
         BUILD_TYPE: debug
         LSAN_OPTIONS: 'exitcode=42' #Use a non-standard exit code to ensure LSAN errors are detected
+        # In Ubuntu 20240310.1.0, the entropy of ASLR has increased (28 -> 32). LLVM 14 in this
+        # image is not compatible with this increased ASLR entropy. Apparently, memory sanitizer
+        # depends on LLVM and all CI tests where VTR_ENABLE_SANITIZE is enabled fail. For a temporary
+        # fix, we manually reduce the entropy. This quick fix should be removed in the future
+        # when github deploys a more stable Ubuntu image.
       run: |
+        sudo sysctl -w vm.mmap_rnd_bits=28
         export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
         ./.github/scripts/build.sh
         # We skip QoR since we are only checking for errors in sanitizer runs
@@ -349,6 +360,7 @@ jobs:
         CMAKE_PARAMS: '-DVTR_ASSERT_LEVEL=3 -DVTR_ENABLE_SANITIZE=on -DVTR_IPO_BUILD=off -DWITH_BLIFEXPLORER=on -DWITH_PARMYS=OFF -DWITH_ODIN=on'
         BUILD_TYPE: debug
       run: |
+        sudo sysctl -w vm.mmap_rnd_bits=28
         export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
         ./.github/scripts/build.sh
         ./run_reg_test.py odin_reg_basic -show_failures -j2

CMakeLists.txt

@@ -23,7 +23,7 @@ set(VTR_IPO_BUILD "auto" CACHE STRING "Should VTR be compiled with interprocedur
 set_property(CACHE VTR_IPO_BUILD PROPERTY STRINGS auto on off)
 
 #Allow the user to configure how much assertion checking should occur
-set(VTR_ASSERT_LEVEL "2" CACHE STRING "VTR assertion checking level. 0: no assertions, 1: fast assertions, 2: regular assertions, 3: additional assertions with noticable run-time overhead, 4: all assertions (including those with significant run-time cost)")
+set(VTR_ASSERT_LEVEL "2" CACHE STRING "VTR assertion checking level. 0: no assertions, 1: fast assertions, 2: regular assertions, 3: additional assertions with noticeable run-time overhead, 4: all assertions (including those with significant run-time cost)")
 set_property(CACHE VTR_ASSERT_LEVEL PROPERTY STRINGS 0 1 2 3 4)
 
 option(VTR_ENABLE_STRICT_COMPILE "Specifies whether compiler warnings should be treated as errors (e.g. -Werror)" OFF)
@@ -418,11 +418,13 @@ if(${WITH_PARMYS}) # define cmake params to compile Yosys
         set(MAKE_PROGRAM "make")
     endif()
 
-    if(NOT DEFINED "${CMAKE_BUILD_PARALLEL_LEVEL}")
-        set(CUSTOM_BUILD_PARALLEL_LEVEL 16)
-    else()
-        set(CUSTOM_BUILD_PARALLEL_LEVEL "${CMAKE_BUILD_PARALLEL_LEVEL}")
-    endif()
+    # Commented out since a make file should not call another make command with
+    # threads. It should pass this information from the parent automatically.
+    # if(NOT DEFINED "${CMAKE_BUILD_PARALLEL_LEVEL}")
+    #     set(CUSTOM_BUILD_PARALLEL_LEVEL 16)
+    # else()
+    #     set(CUSTOM_BUILD_PARALLEL_LEVEL "${CMAKE_BUILD_PARALLEL_LEVEL}")
+    # endif()
     add_subdirectory(yosys)
 endif()
 

blifexplorer/CMakeLists.txt

@@ -2,7 +2,7 @@ cmake_minimum_required(VERSION 3.16)
 
 project("blifexplorer")
 
-set(CMAKE_CXX_STANDARD 14)
+set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 

doc/src/quickstart/index.rst

@@ -48,6 +48,12 @@ On most unix-like systems you can run:
 
     > make
 
+The default front-end for VTR is :ref:`Parmys<parmys>`, but you can build with ODIN II instead using the command below. This is required to run :ref:`Synthesizing with ODIN II<synthesizing_with_odin_ii>`.
+
+.. code-block:: bash
+
+    > make CMAKE_PARAMS="-DWITH_ODIN=on"
+
 from the VTR root directory (hereafter referred to as :term:`$VTR_ROOT`) to build VTR.
 
 .. note:: 
@@ -63,6 +69,8 @@ from the VTR root directory (hereafter referred to as :term:`$VTR_ROOT`) to buil
     * define VTR_ROOT as a variable in your shell (e.g. if ``~/trees/vtr`` is the path to the VTR source tree on your machine, run the equivalent of ``VTR_ROOT=~/trees/vtr`` in BASH) which will allow you to run the commands as written in this guide, or
     * manually replace `$VTR_ROOT` in the example commands below with your path to the VTR source tree.
 
+
+
 For more details on building VTR on various operating systems/platforms see :doc:`Building VTR</BUILDING>`.
 
 
@@ -235,6 +243,7 @@ Next we need to run the three main sets of tools:
 * :ref:`ABC` performs 'logic optimization' which simplifies the circuit logic, and 'technology mapping' which converts logic equations into the Look-Up-Tables (LUTs) available on an FPGA, and
 * :ref:`VPR` which performs packing, placement and routing of the circuit to implement it on the targetted FPGA architecture.
 
+.. _synthesizing_with_odin_ii:
 Synthesizing with ODIN II
 ~~~~~~~~~~~~~~~~~~~~~~~~~
 

doc/src/tutorials/flow/basic_flow.rst

@@ -17,16 +17,9 @@ The following steps show you to run the VTR design flow to map a sample circuit
 
         $VTR_ROOT/vtr_flow/scripts/run_vtr_task.py basic_no_timing 
         
-    The subdirectory ``regression_tests/vtr_reg_basic`` contains tests that are to be run before each commit. They check for basic functionallity to make sure nothing was           extremely out of order. This command runs the VTR flow on a set of circuits and a single architecture. 
+    The subdirectory ``regression_tests/vtr_reg_basic`` contains tests that are to be run before each commit. They check for basic functionality to make sure nothing was extremely out of order. This command runs the VTR flow on a set of circuits and a single architecture. 
     The files generated from the run are stored in ``basic_no_timing/run[#]`` where ``[#]`` is the number of runs you have done.
-    If this is your first time running the flow, the results will be stored in basic_no_timing/run001.
-    When the script completes, enter the following command:
-
-    .. code-block:: shell
-
-         ../../../scripts/python_libs/vtr/parse_vtr_task.py basic_no_timing/
-
-    This parses out the information of the VTR run and outputs the results in a text file called ``run[#]/parse_results.txt``.
+    If this is your first time running the flow, the results will be stored in basic_no_timing/run001. The command parses out the information of the VTR run and outputs the results in a text file called ``run[#]/parse_results.txt``.
 
     More info on how to run the flow on multiple circuits and architectures along with different options later.
     Before that, we need to ensure that the run that you have done works.

doc/src/vpr/basic_flow.rst

@@ -4,7 +4,7 @@ Basic flow
 The Place and Route process in VPR consists of several steps:
 
 - Packing (combinines primitives into complex blocks)
-- Placment (places complex blocks within the FPGA grid)
+- Placement (places complex blocks within the FPGA grid)
 - Routing (determines interconnections between blocks)
 - Analysis (analyzes the implementation)
 

doc/src/vpr/command_line_usage.rst

@@ -1298,6 +1298,14 @@ The following options are only valid when the router is in timing-driven mode (t
 
     **Default:** ``1.2``
 
+.. option:: --router_profiler_astar_fac <float>
+    
+    Controls the directedness of the timing-driven router's exploration when doing router delay profiling of an architecture.
+    The router delay profiling step is currently used to calculate the place delay matrix lookup.
+    Values between 1 and 2 are resonable; higher values trade some quality for reduced run-time.
+
+    **Default:** ``1.2``
+
 .. option:: --max_criticality <float>
 
     Sets the maximum fraction of routing cost that can come from delay (vs. coming from routability) for any net.

doc/src/vtr/benchmarks.rst

@@ -102,7 +102,7 @@ These designs use many precisions including binary, different fixed point types
     proxy               Proxy/synthetic benchmarks
     =================   ======================================
 
-The VTR benchmarks are provided as Verilog (enabling full flexibility to modify and change how the designs are implemented) under: ::
+The Koios benchmarks are provided as Verilog (enabling full flexibility to modify and change how the designs are implemented) under: ::
 
     $VTR_ROOT/vtr_flow/benchmarks/verilog/koios
 
@@ -207,4 +207,4 @@ real application domains. On the other hand, MLP benchmarks include modules that
 and move data. Pre-synthesized netlists for the synthetic benchmarks are added to VTR project, but MLP netlists should
 be downloaded separately.
 
-.. note:: The NoC MLP benchmarks are not included with the VTR release (due to their size). However they can be downloaded and extracted by running ``make get_noc_mlp_benchmarks`` from the root of the VTR tree.  They can also be `downloaded manually <https://www.eecg.utoronto.ca/~vaughn/titan/>`_.
+.. note:: The NoC MLP benchmarks are not included with the VTR release (due to their size). However they can be downloaded and extracted by running ``make get_noc_mlp_benchmarks`` from the root of the VTR tree.  They can also be `downloaded manually <https://www.eecg.utoronto.ca/~vaughn/titan/>`_.

libs/EXTERNAL/libargparse/README.md

@@ -2,7 +2,7 @@ libargparse
 ===========
 This is (yet another) simple command-line parser for C++ applications, inspired by Python's agparse module.
 
-It requires only a C++11 compiler, and has no external dependancies.
+It requires only a C++11 compiler, and has no external dependencies.
 
 One of the advantages of libargparse is that all conversions from command-line strings to program types (bool, int etc.) are performed when the command line is parsed (and not when the options are accessed).
 This avoids command-line related errors from showing up deep in the program execution, which can be problematic for long-running programs.

libs/EXTERNAL/libargparse/src/argparse.cpp

@@ -1,10 +1,10 @@
 #include <algorithm>
-#include <array>
 #include <list>
 #include <cassert>
 #include <string>
 #include <set>
 #include <limits>
+#include <utility>
 
 #include "argparse.hpp"
 #include "argparse_util.hpp"
@@ -16,16 +16,16 @@ namespace argparse {
      * ArgumentParser
      */
 
-    ArgumentParser::ArgumentParser(std::string prog_name, std::string description_str, std::ostream& os)
-        : description_(description_str)
+    ArgumentParser::ArgumentParser(const std::string& prog_name, std::string description_str, std::ostream& os)
+        : description_(std::move(description_str))
         , formatter_(new DefaultFormatter())
         , os_(os)
         {
         prog(prog_name);
         argument_groups_.push_back(ArgumentGroup("arguments"));
     }
 
-    ArgumentParser& ArgumentParser::prog(std::string prog_name, bool basename_only) {
+    ArgumentParser& ArgumentParser::prog(const std::string& prog_name, bool basename_only) {
         if (basename_only) {
             prog_ = basename(prog_name);
         } else {
@@ -35,17 +35,17 @@ namespace argparse {
     }
 
     ArgumentParser& ArgumentParser::version(std::string version_str) {
-        version_ = version_str;
+        version_ = std::move(version_str);
         return *this;
     }
 
     ArgumentParser& ArgumentParser::epilog(std::string epilog_str) {
-        epilog_ = epilog_str;
+        epilog_ = std::move(epilog_str);
         return *this;
     }
 
     ArgumentGroup& ArgumentParser::add_argument_group(std::string description_str) {
-        argument_groups_.push_back(ArgumentGroup(description_str));
+        argument_groups_.push_back(ArgumentGroup(std::move(description_str)));
         return argument_groups_[argument_groups_.size() - 1];
     }
 
@@ -72,7 +72,7 @@ namespace argparse {
     void ArgumentParser::parse_args_throw(int argc, const char* const* argv) {
         std::vector<std::string> arg_strs;
         for (int i = 1; i < argc; ++i) {
-            arg_strs.push_back(argv[i]);
+            arg_strs.emplace_back(argv[i]);
         }
 
         parse_args_throw(arg_strs);
@@ -241,7 +241,7 @@ namespace argparse {
                     } else if (arg->nargs() == '+' || arg->nargs() == '*') {
                         if (arg->nargs() == '+') {
                             assert(nargs_read >= 1);
-                            assert(values.size() >= 1);
+                            assert(!values.empty());
                         }
 
                         for (const auto& value : values) {
@@ -410,11 +410,11 @@ namespace argparse {
      * ArgumentGroup
      */
     ArgumentGroup::ArgumentGroup(std::string name_str)
-        : name_(name_str)
+        : name_(std::move(name_str))
         {}
 
     ArgumentGroup& ArgumentGroup::epilog(std::string str) {
-        epilog_ = str;
+        epilog_ = std::move(str);
         return *this;
     }
     std::string ArgumentGroup::name() const { return name_; }
@@ -425,10 +425,10 @@ namespace argparse {
      * Argument
      */
     Argument::Argument(std::string long_opt, std::string short_opt)
-        : long_opt_(long_opt)
-        , short_opt_(short_opt) {
+        : long_opt_(std::move(long_opt))
+        , short_opt_(std::move(short_opt)) {
 
-        if (long_opt_.size() < 1) {
+        if (long_opt_.empty()) {
             throw ArgParseError("Argument must be at least one character long");
         }
 
@@ -445,7 +445,7 @@ namespace argparse {
     }
 
     Argument& Argument::help(std::string help_str) {
-        help_ = help_str;
+        help_ = std::move(help_str);
         return *this;
     }
 
@@ -476,12 +476,12 @@ namespace argparse {
     }
 
     Argument& Argument::metavar(std::string metavar_str) {
-        metavar_ = metavar_str;
+        metavar_ = std::move(metavar_str);
         return *this;
     }
 
     Argument& Argument::choices(std::vector<std::string> choice_values) {
-        choices_ = choice_values;
+        choices_ = std::move(choice_values);
         return *this;
     }
 
@@ -536,7 +536,7 @@ namespace argparse {
     }
 
     Argument& Argument::group_name(std::string grp) {
-        group_name_ = grp;
+        group_name_ = std::move(grp);
         return *this;
     }
 

libs/EXTERNAL/libargparse/src/argparse.hpp

@@ -34,10 +34,10 @@ namespace argparse {
     class ArgumentParser {
         public:
             //Initializes an argument parser
-            ArgumentParser(std::string prog_name, std::string description_str=std::string(), std::ostream& os=std::cout);
+            ArgumentParser(const std::string& prog_name, std::string description_str=std::string(), std::ostream& os=std::cout);
 
             //Overrides the program name
-            ArgumentParser& prog(std::string prog, bool basename_only=true);
+            ArgumentParser& prog(const std::string& prog, bool basename_only=true);
 
             //Sets the program version
             ArgumentParser& version(std::string version);

libs/EXTERNAL/libcatch2

@@ -17,14 +17,38 @@ target_include_directories(libarchfpga PUBLIC ${LIB_INCLUDE_DIRS})
 
 set_target_properties(libarchfpga PROPERTIES PREFIX "") #Avoid extra 'lib' prefix
 
-#Specify link-time dependancies
+#Specify link-time dependencies
 target_link_libraries(libarchfpga
                         libvtrutil
                         libpugixml
                         libpugiutil
-                        libvtrcapnproto
 )
 
+if(${VTR_ENABLE_CAPNPROTO})
+    target_link_libraries(libarchfpga libvtrcapnproto)
+    target_compile_definitions(libarchfpga PRIVATE VTR_ENABLE_CAPNPROTO)
+endif()
+
+# Using filesystem library requires additional compiler/linker options for GNU implementation prior to 9.1
+# and LLVM implementation prior to LLVM 9.0;
+if(CMAKE_C_COMPILER_ID STREQUAL "GNU")
+    # Get the compiler version string
+    execute_process(COMMAND ${CMAKE_C_COMPILER} --version OUTPUT_VARIABLE _gcc_output)
+    string(REGEX MATCH "[0-9]+\\.[0-9]+(\\.[0-9]+)?" _gcc_version ${_gcc_output})
+
+    if(_gcc_version VERSION_LESS "9.1")
+        target_link_libraries(libarchfpga stdc++fs)
+    endif()
+elseif(CMAKE_C_COMPILER_ID STREQUAL "Clang")
+    # Get the compiler version string
+    execute_process(COMMAND ${CMAKE_C_COMPILER} --version OUTPUT_VARIABLE _clang_output)
+    string(REGEX MATCH "[0-9]+\\.[0-9]+(\\.[0-9]+)?" _clang_version ${_clang_output})
+
+    if(_clang_version VERSION_LESS "9.0")
+        target_link_libraries(libarchfpga c++fs)
+    endif()
+endif()
+
 target_compile_definitions(libarchfpga PUBLIC ${INTERCHANGE_SCHEMA_HEADERS})
 
 #Create the test executable

libs/libarchfpga/CMakeLists.txt

@@ -1,14 +1,16 @@
 #include "device_grid.h"
 
+#include <utility>
+
 DeviceGrid::DeviceGrid(std::string grid_name, vtr::NdMatrix<t_grid_tile, 3> grid)
-    : name_(grid_name)
-    , grid_(grid) {
+    : name_(std::move(grid_name))
+    , grid_(std::move(grid)) {
     count_instances();
 }
 
 DeviceGrid::DeviceGrid(std::string grid_name, vtr::NdMatrix<t_grid_tile, 3> grid, std::vector<t_logical_block_type_ptr> limiting_res)
-    : DeviceGrid(grid_name, grid) {
-    limiting_resources_ = limiting_res;
+    : DeviceGrid(std::move(grid_name), std::move(grid)) {
+    limiting_resources_ = std::move(limiting_res);
 }
 
 size_t DeviceGrid::num_instances(t_physical_tile_type_ptr type, int layer_num) const {