Merge branch 'openfpga' into patch-1

Author: Wang-Yuanqi-source

.github/workflows/nightly_test.yml

@@ -4,20 +4,23 @@ on:
   # We want to run the CI when anything is pushed to master.
   # Since master is a protected branch this only happens when a PR is merged.
   # This is a double check in case the PR was stale and had some issues.
-  push:
-    branches:
-      - master
-    paths-ignore: # Prevents from running if only docs are updated
-      - 'doc/**'
-      - '**/*README*'
-      - '**.md'
-      - '**.rst'
-  pull_request:
-    paths-ignore: # Prevents from running if only docs are updated
-      - 'doc/**'
-      - '**/*README*'
-      - '**.md'
-      - '**.rst'
+  # NOTE: This was turned off in late October 2024 since the Nightly Tests were
+  #       no longer working on the self-hosted runners. Will turn this back on
+  #       once the issue is resolved.
+  # push:
+  #   branches:
+  #     - master
+  #   paths-ignore: # Prevents from running if only docs are updated
+  #     - 'doc/**'
+  #     - '**/*README*'
+  #     - '**.md'
+  #     - '**.rst'
+  # pull_request:
+  #   paths-ignore: # Prevents from running if only docs are updated
+  #     - 'doc/**'
+  #     - '**/*README*'
+  #     - '**.md'
+  #     - '**.rst'
   workflow_dispatch:
   schedule:
   - cron: '0 0 * * *' # daily

doc/src/arch/reference.rst

@@ -2337,8 +2337,8 @@ The ``<direct>`` tag and its contents are described below.
     :req_param y_offset: The y location of the receiving CLB relative to the driving CLB.
     :req_param z_offset: The z location of the receiving CLB relative to the driving CLB.
     :opt_param switch_name: [Optional, defaults to delay-less switch if not specified] The name of the ``<switch>`` from ``<switchlist>`` to be used for this direct connection.
-    :opt_param from_side: The associated from_pin's block size (must be one of ``left``, ``right``, ``top``, ``bottom`` or left unspecified)
-    :opt_param to_side: The associated to_pin's block size (must be one of ``left``, ``right``, ``top``, ``bottom`` or left unspecified)
+    :opt_param from_side: The associated from_pin's block side (must be one of ``left``, ``right``, ``top``, ``bottom`` or left unspecified)
+    :opt_param to_side: The associated to_pin's block side (must be one of ``left``, ``right``, ``top``, ``bottom`` or left unspecified)
 
     Describes a dedicated connection between two complex block pins that skips general interconnect.
     This is useful for describing structures such as carry chains as well as adjacent neighbour connections.

doc/src/vpr/command_line_usage.rst

@@ -1074,12 +1074,16 @@ The following options are only used when FPGA device and netlist contain a NoC r
 
     .. note:: noc_flows_file are required to specify if NoC optimization is turned on (--noc on).
 
-.. option:: --noc_routing_algorithm {xy_routing | bfs_routing}
+.. option:: --noc_routing_algorithm {xy_routing | bfs_routing | west_first_routing | north_last_routing | negative_first_routing | odd_even_routing}
 
     Controls the algorithm used by the NoC to route packets.
 
     * ``xy_routing`` Uses the direction oriented routing algorithm. This is recommended to be used with mesh NoC topologies.
-    * ``bfs_routing`` Uses the breadth first search algorithm. The objective is to find a route that uses a minimum number of links. This can be used with any NoC topology.
+    * ``bfs_routing`` Uses the breadth first search algorithm. The objective is to find a route that uses a minimum number of links. This algorithm is not guaranteed to generate deadlock-free traffic flow routes, but can be used with any NoC topology.
+    * ``west_first_routing`` Uses the west-first routing algorithm. This is recommended to be used with mesh NoC topologies.
+    * ``north_last_routing`` Uses the north-last routing algorithm. This is recommended to be used with mesh NoC topologies.
+    * ``negative_first_routing`` Uses the negative-first routing algorithm. This is recommended to be used with mesh NoC topologies.
+    * ``odd_even_routing`` Uses the odd-even routing algorithm. This is recommended to be used with mesh NoC topologies.
 
     **Default:** ``bfs_routing``
 
@@ -1091,28 +1095,45 @@ The following options are only used when FPGA device and netlist contain a NoC r
     * ``noc_placement_weighting = 1`` means noc placement is considered equal to timing and wirelength.
     * ``noc_placement_weighting > 1`` means the placement is increasingly dominated by NoC parameters.
 
-    **Default:** ``0.6``
+    **Default:** ``5.0``
+
+.. option:: --noc_aggregate_bandwidth_weighting <float>
+
+    Controls the importance of minimizing the NoC aggregate bandwidth. This value can be >=0, where 0 would mean the aggregate bandwidth has no relevance to placement.
+    Other positive numbers specify the importance of minimizing the NoC aggregate bandwidth compared to other NoC-related cost terms.
+    Weighting factors for NoC-related cost terms are normalized internally. Therefore, their absolute values are not important, and
+    only their relative ratios determine the importance of each cost term.
+
+    **Default:** ``0.38``
 
 .. option:: --noc_latency_constraints_weighting <float>
 
-    Controls the importance of meeting all the NoC traffic flow latency constraints.
+    Controls the importance of meeting all the NoC traffic flow latency constraints. This value can be >=0, where 0 would mean latency constraints have no relevance to placement.
+    Other positive numbers specify the importance of meeting latency constraints compared to other NoC-related cost terms.
+    Weighting factors for NoC-related cost terms are normalized internally. Therefore, their absolute values are not important, and
+    only their relative ratios determine the importance of each cost term.
 
-    * ``latency_constraints = 0`` means the latency constraints have no relevance to placement.
-    * ``0 < latency_constraints < 1`` means the latency constraints are weighted equally to the sum of other placement cost components. 
-    * ``latency_constraints > 1`` means the placement is increasingly dominated by reducing the latency constraints of the traffic flows.
-    
-    **Default:** ``1``
+    **Default:** ``0.6``
 
 .. option:: --noc_latency_weighting <float>
 
     Controls the importance of reducing the latencies of the NoC traffic flows.
-    This value can be >=0, 
+    This value can be >=0, where 0 would mean the latencies have no relevance to placement
+    Other positive numbers specify the importance of minimizing aggregate latency compared to other NoC-related cost terms.
+    Weighting factors for NoC-related cost terms are normalized internally. Therefore, their absolute values are not important, and
+    only their relative ratios determine the importance of each cost term.
 
-    * ``latency = 0`` means the latencies have no relevance to placement.
-    * ``0 < latency < 1`` means the latencies are weighted equally to the sum of other placement cost components. 
-    * ``latency > 1`` means the placement is increasingly dominated by reducing the latencies of the traffic flows.
-    
-    **Default:** ``0.05``
+    **Default:** ``0.02``
+
+.. option:: --noc_congestion_weighting <float>
+
+    Controls the importance of reducing the congestion of the NoC links.
+    This value can be >=0, where 0 would mean the congestion has no relevance to placement.
+    Other positive numbers specify the importance of minimizing congestion compared to other NoC-related cost terms.
+    Weighting factors for NoC-related cost terms are normalized internally. Therefore, their absolute values are not important, and
+    only their relative ratios determine the importance of each cost term.
+
+    **Default:** ``0.25``
 
 .. option:: --noc_swap_percentage <float>
 
@@ -1122,7 +1143,7 @@ The following options are only used when FPGA device and netlist contain a NoC r
     * ``0`` means NoC blocks will be moved at the same rate as other blocks. 
     * ``100`` means all swaps attempted by the placer are NoC router blocks.
 
-    **Default:** ``40``    
+    **Default:** ``0``
 
 .. option:: --noc_placement_file_name <file>
 

libs/libarchfpga/src/arch_check.cpp

@@ -101,7 +101,7 @@ void check_port_direct_mappings(t_physical_tile_type_ptr physical_tile, t_sub_ti
     if (pb_type->num_pins > (sub_tile->num_phy_pins / sub_tile->capacity.total())) {
         archfpga_throw(__FILE__, __LINE__,
                        "Logical Block (%s) has more pins than the Sub Tile (%s).\n",
-                       logical_block->name, sub_tile->name);
+                       logical_block->name.c_str(), sub_tile->name.c_str());
     }
 
     auto& pin_direct_maps = physical_tile->tile_block_pin_directs_map.at(logical_block->index);
@@ -110,7 +110,7 @@ void check_port_direct_mappings(t_physical_tile_type_ptr physical_tile, t_sub_ti
     if (pb_type->num_pins != (int)pin_direct_map.size()) {
         archfpga_throw(__FILE__, __LINE__,
                        "Logical block (%s) and Sub tile (%s) have a different number of ports.\n",
-                       logical_block->name, physical_tile->name);
+                       logical_block->name.c_str(), physical_tile->name.c_str());
     }
 
     for (auto pin_map : pin_direct_map) {
@@ -126,7 +126,7 @@ void check_port_direct_mappings(t_physical_tile_type_ptr physical_tile, t_sub_ti
             || sub_tile_port->equivalent != block_port->equivalent) {
             archfpga_throw(__FILE__, __LINE__,
                            "Logical block (%s) and Physical tile (%s) do not have equivalent port specifications. Sub tile port %s, logical block port %s\n",
-                           logical_block->name, sub_tile->name, sub_tile_port->name, block_port->name);
+                           logical_block->name.c_str(), sub_tile->name.c_str(), sub_tile_port->name, block_port->name);
         }
     }
 }