Merge pull request #2348 from verilog-to-routing/3d_link_cost_in_placement

Enhanced 3D Placement Cost and Router Lookahead with Cross-Die Delay

Author: vaughnbetz

doc/src/vpr/command_line_usage.rst

@@ -381,6 +381,9 @@ Use the options below to override this default naming behaviour.
 .. option:: --write_placement_delay_lookup <file>
 
     Writes the placement delay lookup to the specified file.
+.. option:: --write_initial_place_file <file>
+
+    Writes out the the placement chosen by the initial placement algorithm to the specified file
 
 .. option:: --outfile_prefix <string>
 
@@ -769,6 +772,19 @@ If any of init_t, exit_t or alpha_t is specified, the user schedule, with a fixe
 
     **Default:**  ``criticality_timing``
 
+.. option:: --place_bounding_box_mode {auto_bb | cube_bb | per_layer_bb}
+
+    Specifies the type of the wirelength estimator used during placement. For single layer architectures, cube_bb (a 3D bounding box) is always used (and is the same as per_layer_bb).
+    For 3D architectures, cube_bb is appropriate if you can cross between layers at switch blocks, while if you can only cross between layers at output pins per_layer_bb (one bouding box per layer) is more accurate and appropriate.
+
+    ``auto_bb``: The bounding box type is determined automatically based on the cross-layer connections.
+
+    ``cube_bb``: ``cube_bb`` bounding box is used to estimate the wirelength.
+
+    ``per_layer_bb``: ``per_layer_bb`` bounding box is used to estimate the wirelength
+
+    **Default:** ``auto_bb``
+
 .. option:: --place_chan_width <int>
 
     Tells VPR how many tracks a channel of relative width 1 is expected to need to complete routing of this circuit.

libs/librrgraph/src/base/rr_graph_utils.cpp

@@ -5,13 +5,9 @@
  ***************************************************************************/
 #include <queue>
 #include <random>
-#include <algorithm>
 
 #include "rr_graph_utils.h"
 
-#include "vtr_memory.h"
-#include "vtr_time.h"
-
 #include "vpr_error.h"
 
 #include "rr_graph_obj.h"
@@ -119,4 +115,27 @@ vtr::vector<RRNodeId, std::vector<RREdgeId>> get_fan_in_list(const RRGraphView&
         });
 
     return node_fan_in_list;
+}
+
+bool inter_layer_connections_limited_to_opin(const RRGraphView& rr_graph) {
+    bool limited_to_opin = true;
+    for (const auto& from_node : rr_graph.nodes()) {
+        for (t_edge_size edge : rr_graph.edges(from_node)) {
+            RRNodeId to_node = rr_graph.edge_sink_node(from_node, edge);
+            int from_layer = rr_graph.node_layer(from_node);
+            int to_layer = rr_graph.node_layer(to_node);
+
+            if (from_layer != to_layer) {
+                if (rr_graph.node_type(from_node) != e_rr_type::OPIN) {
+                    limited_to_opin = false;
+                    break;
+                }
+            }
+        }
+        if (!limited_to_opin) {
+            break;
+        }
+    }
+
+    return limited_to_opin;
 }

libs/librrgraph/src/base/rr_graph_utils.h

@@ -48,4 +48,12 @@ vtr::vector<RRNodeId, std::vector<RREdgeId>> get_fan_in_list(const RRGraphView&
 int seg_index_of_cblock(const RRGraphView& rr_graph, t_rr_type from_rr_type, int to_node);
 int seg_index_of_sblock(const RRGraphView& rr_graph, int from_node, int to_node);
 
+/**
+ * @brief This function checks whether all inter-die connections are form OPINs. Return "true"
+ * if that is the case. Can be used for multiple purposes. For example, to determine which type of bounding
+ * box to be used to estimate the wire-length of a net.
+ * @param rr_graph
+ * @return
+ */
+bool inter_layer_connections_limited_to_opin(const RRGraphView& rr_graph);
 #endif

libs/librrgraph/src/base/rr_graph_view.h

@@ -234,14 +234,15 @@ class RRGraphView {
     }
 
     /** @brief Get string of information about routing resource node. The string will contain the following information.
-     * type, side, x_low, x_high, y_low, y_high, length, direction, segment_name
+     * type, side, x_low, x_high, y_low, y_high, length, direction, segment_name, layer num
      * This function is inlined for runtime optimization.
      */
     inline const std::string node_coordinate_to_string(RRNodeId node) const {
         std::string start_x;                                           //start x-coordinate
         std::string start_y;                                           //start y-coordinate
         std::string end_x;                                             //end x-coordinate
         std::string end_y;                                             //end y-coordinate
+        std::string layer_num_str;                                     //layer number
         std::string arrow;                                             //direction arrow
         std::string coordinate_string = node_type_string(node);        //write the component's type as a routing resource node
         coordinate_string += ":" + std::to_string(size_t(node)) + " "; //add the index of the routing resource node
@@ -256,12 +257,14 @@ class RRGraphView {
             coordinate_string += ")"; //add the side of the routing resource node
             // For OPINs and IPINs the starting and ending coordinate are identical, so we can just arbitrarily assign the start to larger values
             // and the end to the lower coordinate
-            start_x = " (" + std::to_string(node_xhigh(node)) + ","; //start and end coordinates are the same for OPINs and IPINs
-            start_y = std::to_string(node_yhigh(node)) + ")";
+            start_x =  " (" + std::to_string(node_xhigh(node)) + ","; //start and end coordinates are the same for OPINs and IPINs
+            start_y = std::to_string(node_yhigh(node)) + ",";
+            layer_num_str = std::to_string(node_layer(node)) + ")";
         } else if (node_type(node) == SOURCE || node_type(node) == SINK) {
             // For SOURCE and SINK the starting and ending coordinate are identical, so just use start
-            start_x = "(" + std::to_string(node_xhigh(node)) + ",";
-            start_y = std::to_string(node_yhigh(node)) + ")";
+            start_x = " (" + std::to_string(node_xhigh(node)) + ",";
+            start_y = std::to_string(node_yhigh(node)) + ",";
+            layer_num_str = std::to_string(node_layer(node)) + ")";
         } else if (node_type(node) == CHANX || node_type(node) == CHANY) { //for channels, we would like to describe the component with segment specific information
             RRIndexedDataId cost_index = node_cost_index(node);
             int seg_index = rr_indexed_data_[cost_index].seg_index;
@@ -272,26 +275,29 @@ class RRGraphView {
             arrow = "->"; //we will point the coordinates from start to finish, left to right
 
             if (node_direction(node) == Direction::DEC) {                //signal travels along decreasing direction
+
                 start_x = " (" + std::to_string(node_xhigh(node)) + ","; //start coordinates have large value
-                start_y = std::to_string(node_yhigh(node)) + ")";
-                end_x = "(" + std::to_string(node_xlow(node)) + ","; //end coordinates have smaller value
-                end_y = std::to_string(node_ylow(node)) + ")";
+                start_y = std::to_string(node_yhigh(node)) + ",";
+                end_x = " (" + std::to_string(node_xlow(node)) + ","; //end coordinates have smaller value
+                end_y = std::to_string(node_ylow(node)) + ",";
+                layer_num_str = std::to_string(node_layer(node)) + ")";
             }
 
             else {                                                      // signal travels in increasing direction, stays at same point, or can travel both directions
                 start_x = " (" + std::to_string(node_xlow(node)) + ","; //start coordinates have smaller value
-                start_y = std::to_string(node_ylow(node)) + ")";
-                end_x = "(" + std::to_string(node_xhigh(node)) + ","; //end coordinates have larger value
-                end_y = std::to_string(node_yhigh(node)) + ")";
+                start_y = std::to_string(node_ylow(node)) + ",";
+                end_x = " (" + std::to_string(node_xhigh(node)) + ","; //end coordinates have larger value
+                end_y = std::to_string(node_yhigh(node)) + ",";
+                layer_num_str = std::to_string(node_layer(node)) + ")"; //layer number
                 if (node_direction(node) == Direction::BIDIR) {
                     arrow = "<->"; //indicate that signal can travel both direction
                 }
             }
         }
 
-        coordinate_string += start_x + start_y; //Write the starting coordinates
+        coordinate_string +=  start_x + start_y + layer_num_str; //Write the starting coordinates
         coordinate_string += arrow;             //Indicate the direction
-        coordinate_string += end_x + end_y;     //Write the end coordinates
+        coordinate_string += end_x + end_y + layer_num_str;     //Write the end coordinates
         return coordinate_string;
     }
 

utils/route_diag/src/main.cpp

@@ -84,6 +84,8 @@ static void do_one_route(const Netlist<>& net_list,
     bounding_box.xmax = device_ctx.grid.width() + 1;
     bounding_box.ymin = 0;
     bounding_box.ymax = device_ctx.grid.height() + 1;
+    bounding_box.layer_min = 0;
+    bounding_box.layer_max = device_ctx.grid.get_num_layers() - 1;
 
     t_conn_cost_params cost_params;
     cost_params.criticality = router_opts.max_criticality;
@@ -203,9 +205,12 @@ static void profile_source(const Netlist<>& net_list,
                     vtr::ScopedStartFinishTimer delay_timer(vtr::string_fmt(
                         "Routing Src: %d Sink: %d", source_rr_node,
                         sink_rr_node));
-                    successfully_routed = profiler.calculate_delay(RRNodeId(source_rr_node), RRNodeId(sink_rr_node),
-                                                        router_opts,
-                                                        &delays[sink_x][sink_y]);
+
+                    successfully_routed = profiler.calculate_delay(RRNodeId(source_rr_node),
+                                                                   RRNodeId(sink_rr_node),
+                                                                   router_opts,
+                                                                   &delays[sink_x][sink_y],
+                                                                   layer_num);
                 }
 
                 if (successfully_routed) {

vpr/src/base/SetupVPR.cpp

@@ -622,6 +622,8 @@ static void SetupPlacerOpts(const t_options& Options, t_placer_opts* PlacerOpts)
 
     PlacerOpts->constraints_file = Options.constraints_file;
 
+    PlacerOpts->write_initial_place_file = Options.write_initial_place_file;
+
     PlacerOpts->pad_loc_type = Options.pad_loc_type;
 
     PlacerOpts->place_chan_width = Options.PlaceChanWidth;
@@ -661,6 +663,7 @@ static void SetupPlacerOpts(const t_options& Options, t_placer_opts* PlacerOpts)
     PlacerOpts->place_static_move_prob = Options.place_static_move_prob;
     PlacerOpts->place_static_notiming_move_prob = Options.place_static_notiming_move_prob;
     PlacerOpts->place_high_fanout_net = Options.place_high_fanout_net;
+    PlacerOpts->place_bounding_box_mode = Options.place_bounding_box_mode;
     PlacerOpts->RL_agent_placement = Options.RL_agent_placement;
     PlacerOpts->place_agent_multistate = Options.place_agent_multistate;
     PlacerOpts->place_checkpointing = Options.place_checkpointing;

vpr/src/base/read_options.cpp

@@ -430,6 +430,41 @@ struct ParsePlaceAlgorithm {
     }
 };
 
+struct ParsePlaceBoundingBox {
+    ConvertedValue<e_place_bounding_box_mode> from_str(std::string str) {
+        ConvertedValue<e_place_bounding_box_mode> conv_value;
+        if (str == "auto_bb") {
+            conv_value.set_value(AUTO_BB);
+        } else if (str == "cube_bb") {
+            conv_value.set_value(CUBE_BB);
+        } else if (str == "per_layer_bb") {
+            conv_value.set_value(PER_LAYER_BB);
+        } else {
+            std::stringstream msg;
+            msg << "Invalid conversion from '" << str << "' to e_place_algorithm (expected one of: " << argparse::join(default_choices(), ", ") << ")";
+            conv_value.set_error(msg.str());
+        }
+        return conv_value;
+    }
+
+    ConvertedValue<std::string> to_str(e_place_bounding_box_mode val) {
+        ConvertedValue<std::string> conv_value;
+        if (val == AUTO_BB) {
+            conv_value.set_value("auto_bb");
+        } else if (val == CUBE_BB) {
+            conv_value.set_value("cube_bb");
+        } else {
+            VTR_ASSERT(val == PER_LAYER_BB);
+            conv_value.set_value("per_layer_bb");
+        }
+        return conv_value;
+    }
+
+    std::vector<std::string> default_choices() {
+        return {"auto_bb", "cube_bb", "per_layer_bb"};
+    }
+};
+
 struct ParsePlaceAgentAlgorithm {
     ConvertedValue<e_agent_algorithm> from_str(std::string str) {
         ConvertedValue<e_agent_algorithm> conv_value;
@@ -1569,6 +1604,11 @@ argparse::ArgumentParser create_arg_parser(std::string prog_name, t_options& arg
         .metavar("RR_GRAPH_FILE")
         .show_in(argparse::ShowIn::HELP_ONLY);
 
+    file_grp.add_argument(args.write_initial_place_file, "--write_initial_place_file")
+        .help("Writes out the the placement chosen by the initial placement algorithm to the specified file")
+        .metavar("INITIAL_PLACE_FILE")
+        .show_in(argparse::ShowIn::HELP_ONLY);
+
     file_grp.add_argument(args.read_vpr_constraints_file, "--read_vpr_constraints")
         .help("Reads the floorplanning constraints that packing and placement must respect from the specified XML file.")
         .show_in(argparse::ShowIn::HELP_ONLY);
@@ -2007,6 +2047,20 @@ argparse::ArgumentParser create_arg_parser(std::string prog_name, t_options& arg
         .default_value("10")
         .show_in(argparse::ShowIn::HELP_ONLY);
 
+    place_grp.add_argument<e_place_bounding_box_mode, ParsePlaceBoundingBox>(args.place_bounding_box_mode, "--place_bounding_box_mode")
+        .help(
+            "Specifies the type of bounding box to be used in 3D architectures.\n"
+            "\n"
+            "MODE options:\n"
+            "  auto_bb     : Automatically determine the appropriate bounding box based on the connections between layers.\n"
+            "  cube_bb            : Use 3D bounding boxes.\n"
+            "  per_layer_bb     : Use per-layer bounding boxes.\n"
+            "\n"
+            "Choose one of the available modes to define the behavior of bounding boxes in your 3D architecture. The default mode is 'automatic'.")
+        .default_value("auto_bb")
+        .choices({"auto_bb", "cube_bb", "per_layer_bb"})
+        .show_in(argparse::ShowIn::HELP_ONLY);
+
     place_grp.add_argument<bool, ParseOnOff>(args.RL_agent_placement, "--RL_agent_placement")
         .help(
             "Uses a Reinforcement Learning (RL) agent in choosing the appropiate move type in placement."

vpr/src/base/read_options.h

@@ -27,6 +27,7 @@ struct t_options {
     argparse::ArgValue<std::string> constraints_file;
     argparse::ArgValue<std::string> write_rr_graph_file;
     argparse::ArgValue<std::string> read_rr_graph_file;
+    argparse::ArgValue<std::string> write_initial_place_file;
     argparse::ArgValue<std::string> read_vpr_constraints_file;
     argparse::ArgValue<std::string> write_vpr_constraints_file;
 
@@ -127,6 +128,7 @@ struct t_options {
     argparse::ArgValue<std::vector<float>> place_static_move_prob;
     argparse::ArgValue<std::vector<float>> place_static_notiming_move_prob;
     argparse::ArgValue<int> place_high_fanout_net;
+    argparse::ArgValue<e_place_bounding_box_mode> place_bounding_box_mode;
 
     argparse::ArgValue<bool> RL_agent_placement;
     argparse::ArgValue<bool> place_agent_multistate;

vpr/src/base/vpr_api.cpp

@@ -452,6 +452,8 @@ void vpr_create_device_grid(const t_vpr_setup& vpr_setup, const t_arch& Arch) {
     float target_device_utilization = vpr_setup.PackerOpts.target_device_utilization;
     device_ctx.grid = create_device_grid(vpr_setup.device_layout, Arch.grid_layouts, num_type_instances, target_device_utilization);
 
+    VTR_ASSERT_MSG(device_ctx.grid.get_num_layers() <= MAX_NUM_LAYERS, "Number of layers should be less than MAX_NUM_LAYERS. If you need more layers, please increase the value of MAX_NUM_LAYERS in vpr_types.h");
+
     /*
      *Report on the device
      */

vpr/src/base/vpr_context.h

@@ -402,6 +402,12 @@ struct PlacementContext : public Context {
      * placer_debug_net or placer_debug_block parameters in the command line.
      */
     bool f_placer_debug = false;
+
+    /**
+     * Set this variable to ture if the type of the bounding box used in placement is of the type cube. If it is false,
+     * it would mean that per-layer bounding box is used. For the 2D architecture, the cube bounding box would be used.
+     */
+    bool cube_bb = false;
 };
 
 /**