fix num_bin in custom design

src/database.py

@@ -5,17 +5,8 @@ import math
 from utils import *
 
 
-def load_dataset(args, logger, placement=None):
+def load_dataset(args, logger, params):
     rawdb, gpdb = None, None
-    if args.custom_path != "":
-        params = get_custom_design_params(args)
-    elif args.custom_json != "":
-        logger.info("Detect json mode. Please make sure that tech_lef are included first.")
-        params = get_custom_json_params(args)
-    else:
-        params = get_single_design_params(
-            args.dataset_root, args.dataset, args.design_name, placement
-        )
     parser = IOParser()
     if args.load_from_raw:
         logger.info("loading from original benchmark...")
@@ -116,14 +107,13 @@ class PlaceData(object):
         self.region_boxes = region_boxes
         self.region_boxes_end = region_boxes_end
 
-        # TODO: more cases?
+        # TODO: more cases, hardcode?
         self.node_special_type = torch.zeros(len(node_id2celltype_name), dtype=torch.int32)
-        if False:
+        for node_id, celltype_name in enumerate(node_id2celltype_name):
-            for node_id, celltype_name in enumerate(node_id2celltype_name):
+            if celltype_name.startswith("CORE/BUF"):
-                if celltype_name.startswith("CORE/BUF"):
+                self.node_special_type[node_id] = 1
-                    self.node_special_type[node_id] = 1
+            if celltype_name.startswith("CORE/DFF"):
-                if celltype_name.startswith("CORE/DFF"):
+                self.node_special_type[node_id] = 2
-                    self.node_special_type[node_id] = 2
 
         dataset_format = ""
         if "aux" in dataset_path.keys():
@@ -452,6 +442,7 @@ class PlaceData(object):
         :obj:`*keys`.
         If :obj:`*keys` is not given, the conversion is applied to all present
         attributes."""
+        self.device = device
         return self.apply(lambda x: x.to(device, **kwargs), *keys)
 
     def cpu(self, *keys):
@@ -539,14 +530,15 @@ class PlaceData(object):
 
     def prescale_by_site_width(self):
         # inplace scaling
-        self.die_info /= self.site_width
+        scalar_at = torch.tensor([self.site_width], dtype=torch.float32, device=self.die_info.device)
-        self.region_boxes /= self.site_width
+        self.die_info /= scalar_at
-        self.node_pos /= self.site_width
+        self.region_boxes /= scalar_at
-        self.node_lpos /= self.site_width
+        self.node_pos /= scalar_at
-        self.node_size /= self.site_width
+        self.node_lpos /= scalar_at
-        self.pin_rel_cpos /= self.site_width
+        self.node_size /= scalar_at
-        self.pin_rel_lpos /= self.site_width
+        self.pin_rel_cpos /= scalar_at
-        self.pin_size /= self.site_width
+        self.pin_rel_lpos /= scalar_at
+        self.pin_size /= scalar_at
         self.__die_scale__ *= self.site_width
         return self
 

src/run_placement_nesterov.py

@@ -12,8 +12,8 @@ def get_trunc_node_pos_fn(mov_node_size, data):
 
 def run_placement_main_nesterov(args, logger):
     total_start = time.time()
-    setup_dataset_args(args)
+    params = find_design_params(args, logger)
-    data, rawdb, gpdb = load_dataset(args, logger)
+    data, rawdb, gpdb = load_dataset(args, logger, params)
     device = torch.device(
         "cuda:{}".format(args.gpu) if torch.cuda.is_available() else "cpu"
     )

utils/__init__.py

@@ -3,4 +3,4 @@ from .logger import setup_logger
 from .visualization import *
 from .tools import *
 from .get_design_params import get_single_design_params, get_multiple_design_params, get_custom_design_params, get_custom_json_params
-from .setup_dataset import setup_dataset_args
+from .setup_dataset import find_design_params

utils/setup_dataset.py

@@ -1,4 +1,41 @@
-def setup_dataset_args(args):
+from .get_design_params import get_single_design_params, get_custom_design_params, get_custom_json_params
+
+
+def find_design_params(args, logger, placement=None):
+    if args.custom_path != "":
+        params = get_custom_design_params(args)
+    elif args.custom_json != "":
+        logger.info("Detect json mode. Please make sure that tech_lef are included first.")
+        params = get_custom_json_params(args)
+    else:
+        params = get_single_design_params(
+            args.dataset_root, args.dataset, args.design_name, placement
+        )
+    log_design_params(logger, params)
+    setup_design_args(args)
+    return params
+
+
+def log_design_params(logger, params: dict):
+    content = "Design Info:\n"
+    num_items = 0
+    if "benchmark" in params.keys():
+        content += f"benchmark: {params['benchmark']}\n"
+        num_items += 1
+    if "design_name" in params.keys():
+        content += f"design_name: {params['design_name']}\n"
+        num_items += 1
+    for key, value in params.items():
+        if key == "design_name" or key == "benchmark":
+            continue
+        content += f"{key}: {value}"
+        if num_items < len(params) - 1:
+            content += "\n"
+        num_items += 1
+    logger.info(content)
+
+
+def setup_design_args(args):
     if args.design_name in ["adaptec1", "bigblue1"]:
         args.num_bin_x = args.num_bin_y = 512
         args.target_density = 1.0