Change place flow in init place

vpr/src/base/echo_files.cpp

@@ -122,6 +122,7 @@ void alloc_and_load_echo_file_info() {
     setEchoFileName(E_ECHO_ENDPOINT_TIMING, "endpoint_timing.echo.json");
     setEchoFileName(E_ECHO_LOOKAHEAD_MAP, "lookahead_map.echo");
     setEchoFileName(E_ECHO_RR_GRAPH_INDEXED_DATA, "rr_indexed_data.echo");
+    setEchoFileName(E_ECHO_COMPRESSED_GRIDS, "compressed_grids.echo");
 }
 
 void free_echo_file_info() {

vpr/src/base/echo_files.h

@@ -52,6 +52,7 @@ enum e_echo_files {
     E_ECHO_ENDPOINT_TIMING,
     E_ECHO_LOOKAHEAD_MAP,
     E_ECHO_RR_GRAPH_INDEXED_DATA,
+    E_ECHO_COMPRESSED_GRIDS,
 
     //Timing Graphs
     E_ECHO_PRE_PACKING_TIMING_GRAPH,

vpr/src/place/compressed_grid.cpp

@@ -142,3 +142,40 @@ int grid_to_compressed_approx(const std::vector<int>& coords, int point) {
         return std::distance(coords.begin(), itr - 1);
     return std::distance(coords.begin(), itr);
 }
+
+/*Print the contents of each cluster to an echo file*/
+void echo_compressed_grids(char* filename, const std::vector<t_compressed_block_grid>& comp_grids) {
+    FILE* fp;
+    fp = vtr::fopen(filename, "w");
+
+    auto& device_ctx = g_vpr_ctx.device();
+
+    fprintf(fp, "--------------------------------------------------------------\n");
+    fprintf(fp, "Compressed Grids: \n");
+    fprintf(fp, "--------------------------------------------------------------\n");
+    fprintf(fp, "\n");
+
+    for (int i = 0; i < (int)comp_grids.size(); i++) {
+        fprintf(fp, "\n\nGrid type: %s \n", device_ctx.logical_block_types[i].name);
+
+        fprintf(fp, "X coordinates: \n");
+        for (int j = 0; j < (int)comp_grids[i].compressed_to_grid_x.size(); j++) {
+            fprintf(fp, "%d ", comp_grids[i].compressed_to_grid_x[j]);
+        }
+        fprintf(fp, "\n");
+
+        fprintf(fp, "Y coordinates: \n");
+        for (int k = 0; k < (int)comp_grids[i].compressed_to_grid_y.size(); k++) {
+            fprintf(fp, "%d ", comp_grids[i].compressed_to_grid_y[k]);
+        }
+        fprintf(fp, "\n");
+
+        fprintf(fp, "Subtiles: \n");
+        for (int s = 0; s < (int)comp_grids[i].compatible_sub_tiles_for_tile.size(); s++) {
+            fprintf(fp, "%d ", comp_grids[i].compressed_to_grid_y[s]);
+        }
+        fprintf(fp, "\n");
+    }
+
+    fclose(fp);
+}

vpr/src/place/compressed_grid.h

@@ -64,4 +64,6 @@ int grid_to_compressed(const std::vector<int>& coords, int point);
  */
 int grid_to_compressed_approx(const std::vector<int>& coords, int point);
 
+void echo_compressed_grids(char* filename, const std::vector<t_compressed_block_grid>& comp_grids);
+
 #endif

vpr/src/place/place.cpp

@@ -510,8 +510,7 @@ void try_place(const t_placer_opts& placer_opts,
 
     vtr::ScopedStartFinishTimer timer("Placement");
 
-    initial_placement(placer_opts.pad_loc_type,
-                      placer_opts.constraints_file.c_str());
+    initial_placement(placer_opts.pad_loc_type, placer_opts.constraints_file.c_str());
 
 #ifdef ENABLE_ANALYTIC_PLACE
     /*
@@ -1088,10 +1087,10 @@ static void placement_inner_loop(const t_annealing_state* state,
         }
 #ifdef VERBOSE
         VTR_LOG("t = %g  cost = %g   bb_cost = %g timing_cost = %g move = %d\n",
-                t, costs->cost, costs->bb_cost, costs->timing_cost, inner_iter);
+                state->t, costs->cost, costs->bb_cost, costs->timing_cost, inner_iter);
         if (fabs((costs->bb_cost) - comp_bb_cost(CHECK)) > (costs->bb_cost) * ERROR_TOL)
-            VPR_ERROR(VPR_ERROR_PLACE,
-                      "fabs((*bb_cost) - comp_bb_cost(CHECK)) > (*bb_cost) * ERROR_TOL");
+            VPR_ERROR(VPR_ERROR_PLACE, "bb_cost is %g, comp_bb_cost is %g\n", costs->bb_cost, comp_bb_cost(CHECK));
+            //"fabs((*bb_cost) - comp_bb_cost(CHECK)) > (*bb_cost) * ERROR_TOL");
 #endif
 
         /* Lines below prevent too much round-off error from accumulating
@@ -1101,8 +1100,10 @@ static void placement_inner_loop(const t_annealing_state* state,
          */
         ++(*moves_since_cost_recompute);
         if (*moves_since_cost_recompute > MAX_MOVES_BEFORE_RECOMPUTE) {
+            //VTR_LOG("recomputing costs from scratch, old bb_cost is %g\n", costs->bb_cost);
             recompute_costs_from_scratch(placer_opts, delay_model,
                                          criticalities, costs);
+            //VTR_LOG("new_bb_cost is %g\n", costs->bb_cost);
             *moves_since_cost_recompute = 0;
         }
 
@@ -1551,7 +1552,7 @@ static e_move_result try_swap(const t_annealing_state* state,
     //VTR_ASSERT(check_macro_placement_consistency() == 0);
 #if 0
     //Check that each accepted swap yields a valid placement
-    check_place(*costs, delay_model, place_algorithm);
+    check_place(*costs, delay_model, criticalities, place_algorithm);
 #endif
 
     return move_outcome;
@@ -2005,9 +2006,9 @@ static double comp_bb_cost(e_cost_methods method) {
     }
 
     if (method == CHECK) {
-        VTR_LOG("\n");
-        VTR_LOG("BB estimate of min-dist (placement) wire length: %.0f\n",
-                expected_wirelength);
+        /*VTR_LOG("\n");
+         * VTR_LOG("BB estimate of min-dist (placement) wire length: %.0f\n",
+         * expected_wirelength);*/
     }
     return cost;
 }
@@ -2313,6 +2314,15 @@ static void get_non_updateable_bb(ClusterNetId net_id, t_bb* bb_coord_new) {
 
     ClusterBlockId bnum = cluster_ctx.clb_nlist.net_driver_block(net_id);
     pnum = net_pin_to_tile_pin_index(net_id, 0);
+
+    if (unsigned(pnum) > (physical_tile_type(bnum)->pin_width_offset.size() - 1)) {
+        //Need to sync physical pins if pnum not found in physical pins
+        place_sync_external_block_connections(bnum);
+
+        //Call pnum again after resyncing
+        pnum = net_pin_to_tile_pin_index(net_id, 0);
+    }
+
     x = place_ctx.block_locs[bnum].loc.x
         + physical_tile_type(bnum)->pin_width_offset[pnum];
     y = place_ctx.block_locs[bnum].loc.y

vtr_flow/tasks/regression_tests/vtr_reg_nightly_test1/arithmetic_tasks/figure_8/config/config.txt

@@ -128,4 +128,4 @@ qor_parse_file=qor_standard.txt
 # Pass requirements
 pass_requirements_file=pass_requirements_chain.txt
 
-script_params=-lut_size 6 -routing_failure_predictor off
+script_params=-lut_size 6 -routing_failure_predictor off -seed 2

vtr_flow/tasks/regression_tests/vtr_reg_nightly_test2/titan_quick_qor/config/config.txt

@@ -67,4 +67,4 @@ pass_requirements_file=pass_requirements_vpr_titan.txt
 #A large number of routing iterations is set to ensure the router doesn't give up to easily on the larger benchmarks
 #To be more run-time comparable to commercial tools like Quartus, we run with higher placer effort (inner_num=2) and lower astar_fac (1.0)
 #Set a 24hr timeout so they don't run forever
-script_params=-starting_stage vpr --route_chan_width 300 --max_router_iterations 400 --router_lookahead map -timeout 86400 --initial_pres_fac 1.0 --router_profiler_astar_fac 1.5
+script_params=-starting_stage vpr --route_chan_width 300 --max_router_iterations 400 --router_lookahead map -timeout 86400 --initial_pres_fac 1.0 --router_profiler_astar_fac 1.5 --seed 3

vtr_flow/tasks/regression_tests/vtr_reg_strong/strong_post_routing_sync/config/config.txt

@@ -44,4 +44,4 @@ qor_parse_file=qor_standard.txt
 pass_requirements_file=pass_requirements.txt
 
 #Script parameters
-script_params=-track_memory_usage -check_equivalent -starting_stage abc --gen_post_synthesis_netlist on --sweep_dangling_primary_ios off --sweep_constant_primary_outputs off --seed 2
+script_params=-track_memory_usage -check_equivalent -starting_stage abc --gen_post_synthesis_netlist on --sweep_dangling_primary_ios off --sweep_constant_primary_outputs off --seed 3