Update simulator

doc/source/dataprep.md

@@ -102,10 +102,10 @@ The data generator is a python script that generates random stream set and netwo
 - **--num_ins:** Number of problem instances to generate, default is 1.
 - **--num_stream:** Number of flows in each problem instance, default is 8.
 - **--num_sw:** Number of network bridge in each problem instance, default is 8.
-- **--period:** Period pattern of stream set.
-- **--size:** Size pattern of stream set.
-- **--deadline:** Deadline pattern of stream set.
-- **--topo:** Topology pattern of network.
+- **--period:** Period pattern of stream set. [ns]
+- **--size:** Size pattern of stream set. [bytes]
+- **--deadline:** Deadline pattern of stream set. [ns]
+- **--topo:** Topology pattern of network. [0:line, 1:ring, 2:tree, 3:mesh, 4:mesh_2d]
 - **--output:** Output directory for generated files.
 
 To generate a single dataset:

doc/source/simulation.md

@@ -27,12 +27,12 @@ python setup.py build_ext --inplace
 python3 -m tsnkit.simulation.tas [TASK PATH] [CONFIG PATH]
 ```
 
-- Task path: The stream set file as described in [previous section](dataprep.md).
-- Config path: The folder containing the generated configuration files. The detailed format can be also found in [previous section](dataprep.md). *Please note this path should be the folder path that containing the configuration files, such as `./data/output/`*
-- Iter: The number of network cycles to run the simulation. Default is `1` (use `--iter N` to change).
-- Verbose: If set to `True` (`--verbose`), the simulator prints detailed logs; otherwise it prints a summary. Default is `False`.
-- No-draw: Disable plotting by passing `--no-draw` (useful for benchmarking).
-
+- `task`: The stream set file as described in [previous section](dataprep.md).
+- `config`: The folder containing the generated configuration files. The detailed format can be also found in [previous section](dataprep.md). *Note this should be a folder path that containing the configuration files, such as `./data/output/`*
+- `--iter`: The number of network cycles to run the simulation. Default is `1` (use `--iter N` to change).
+- `--verbose`: If set to `True` (`--verbose`), the simulator prints detailed logs; otherwise it prints a summary. Default is `False`.
+- `--no-draw`: Disable plotting by passing `--no-draw`.
+- `--output`: Save the simulation logs into a `.csv` file (by default current folder). 
 
 The simulator will automatically infer the network settings from the configuration files, thus a separate network path is not required.
 
@@ -50,8 +50,6 @@ During the runtime, the script outputs logs as following to show the forwarding
     [Listener 8]:       Flow 0 - Receive at 8023800
     ```
 
-
-
 The final log indicates any potential errors and the send/receive times for each flow:
 
 

tsnkit/core/_common.py

@@ -7,6 +7,7 @@
 
 from typing import Any, Sequence
 import argparse
+import os
 from .. import core
 
 
@@ -16,19 +17,42 @@ def parse_command_line_args():
         description="Process the stream and network paths."
     )
 
-    # Add the positional arguments
-
-    parser.add_argument("task", type=str, help="The file path to the stream CSV file.")
-    parser.add_argument("net", type=str, help="The file path to the network CSV file.")
+    # Switch to optional flags for all parameters
+    parser.add_argument(
+        "task",
+        type=str,
+        help="The file path to the stream CSV file.",
+    )
     parser.add_argument(
-        "output", type=str, nargs="?", help="The output folder path.", default="./"
+        "net",
+        type=str,
+        help="The file path to the network CSV file.",
     )
     parser.add_argument(
-        "workers", type=int, nargs="?", help="The number of workers.", default=1
+        "--output",
+        type=str,
+        default="./",
+        nargs="?",
+        help="The output folder path.",
     )
     parser.add_argument(
-        "name", type=str, nargs="?", help="The name of the experiment.", default="-"
+        "--workers",
+        type=int,
+        default=1,
+        nargs="?",
+        help="The number of workers.",
     )
+    parser.add_argument(
+        "--name",
+        type=str,
+        default="-",
+        nargs="?",
+        help="The name of the experiment.",
+    )
+
+    parsed = parser.parse_args()
+    ## TODO: Put me somewhere else.
+    os.makedirs(parsed.output, exist_ok=True)
 
     args = parse_command_line_constants(parser)
 
@@ -62,7 +86,7 @@ def benchmark(stream_path, network_path):
 
 if __name__ == "__main__":
     args = parse_command_line_args()
-    benchmark(args.STREAM_PATH, args.NETWORK_PATH)
+    benchmark(args.task, args.net)
 
 
 def _interface(name: str) -> Any:

tsnkit/core/_config.py

@@ -224,7 +224,7 @@ def is_valid_route_format(init_list: List[List]) -> bool:
 
     @staticmethod
     def is_valid_route_logic(init_list: List[List]) -> bool:
-        ## [NOTE] May casue some problem for un-continuous link_id
+        ## [NOTE] May cause some problem for un-continuous link_id
         routes: List[List[Union[Tuple[int, int], Link]]] = [
             [] for i in range(max([int(x[0]) for x in init_list]) + 1)
         ]

tsnkit/data/dataset_spec.py

@@ -85,19 +85,7 @@ def line(num_sw, num_queue, data_rate, header):
         net[i + num_sw, i] = 1
         net[i, i + num_sw] = 1
 
-    result = []
-    for i in range(num_node):
-        for j in range(num_node):
-            if net[i][j]:
-                link = []
-                link.append((i, j))
-                link.append(num_queue)
-                link.append(data_rate)
-                link.append(ERROR)
-                link.append(0)
-                result.append(link)
-
-    result = pd.DataFrame(result, columns=["link", "q_num", "rate", "t_proc", "t_prop"])
+    result = _convert_2darray_to_csv(net, num_node, num_queue, data_rate) 
     result.to_csv(header + ".csv", index=False)
     return net
 
@@ -119,24 +107,13 @@ def ring(num_sw, num_queue, data_rate, header):
     net[0, num_sw - 1] = 1
     net[num_sw - 1, 0] = 1
 
-    result = []
-    for i in range(num_node):
-        for j in range(num_node):
-            if net[i][j]:
-                link = []
-                link.append((i, j))
-                link.append(num_queue)
-                link.append(data_rate)
-                link.append(ERROR)
-                link.append(0)
-                result.append(link)
-
-    result = pd.DataFrame(result, columns=["link", "q_num", "rate", "t_proc", "t_prop"])
+    result = _convert_2darray_to_csv(net, num_node, num_queue, data_rate) 
     result.to_csv(header + ".csv", index=False)
     return net
 
 
 def tree(num_sw, num_queue, data_rate, header):
+    # Aka. STAR
     num_node = num_sw * 2 + 1
     net = np.zeros(shape=(num_node, num_node))
 
@@ -145,19 +122,8 @@ def tree(num_sw, num_queue, data_rate, header):
         net[i * 2 + 1, i] = 1
         net[i, i * 2 + 2] = 1
         net[i * 2 + 2, i] = 1
-    result = []
-    for i in range(num_node):
-        for j in range(num_node):
-            if net[i][j]:
-                link = []
-                link.append((i, j))
-                link.append(num_queue)
-                link.append(data_rate)
-                link.append(ERROR)
-                link.append(0)
-                result.append(link)
 
-    result = pd.DataFrame(result, columns=["link", "q_num", "rate", "t_proc", "t_prop"])
+    result = _convert_2darray_to_csv(net, num_node, num_queue, data_rate) 
     result.to_csv(header + ".csv", index=False)
     return net
 
@@ -170,12 +136,13 @@ def mesh(num_sw, num_queue, data_rate, header):
     for i in range(0, num_sw - 1):
         net[i, i + 1] = 1
         net[i + 1, i] = 1
+
     ## Connect the switch and the end-station
     for i in range(num_sw):
         net[i + num_sw, i] = 1
         net[i, i + num_sw] = 1
 
-    ## Connect the mesh
+    ## Connect the ring
     net[0, num_sw - 1] = 1
     net[num_sw - 1, 0] = 1
 
@@ -184,24 +151,70 @@ def mesh(num_sw, num_queue, data_rate, header):
         net[i, num_sw - i - 1] = 1
         net[num_sw - i - 1, i] = 1
 
-    result = []
-    for i in range(num_node):
-        for j in range(num_node):
-            if net[i][j]:
-                link = []
-                link.append((i, j))
-                link.append(num_queue)
-                link.append(data_rate)
-                link.append(ERROR)
-                link.append(0)
-                result.append(link)
+    result = _convert_2darray_to_csv(net, num_node, num_queue, data_rate) 
+    result.to_csv(header + ".csv", index=False)
+    return net
 
-    result = pd.DataFrame(result, columns=["link", "q_num", "rate", "t_proc", "t_prop"])
+def mesh_2d(num_sw, num_queue, data_rate, header):
+
+    n = int(np.sqrt(num_sw))
+    num_node = num_sw + 4 * n - 4
+    net = np.zeros(shape=(num_node, num_node))
+
+    if n ** 2 != num_sw:
+        raise ValueError("Wrong num_sw for mesh_2d, col_len != row_len")
+
+    row_len = col_len = int(np.sqrt(num_sw))
+
+    # Save to mat
+    mat = []
+    count = 0
+    for i in range(row_len):
+        if i % 2 == 0:
+            row = list(range(count, count + col_len))
+        else:
+            row = list(range(count, count + col_len))[::-1]
+        mat.append(row)
+        count += col_len
+
+    # Fill net
+    searched = set()
+    es_id = num_sw
+
+    def _dfs(x, y):
+        nonlocal es_id, searched
+        if x < 0 or y < 0 or x >= row_len or y >= col_len:
+            return
+        if (x,y) in searched:
+            return
+        searched.add((x,y))
+
+        # Add es on border
+        if x == 0 or y == 0 or x == row_len - 1 or y == col_len - 1:
+            net[mat[x][y]][es_id] = 1
+            net[es_id][mat[x][y]] = 1
+            es_id += 1
+
+        # Connect sw neighbors
+        def _search_nxt(x, y, nx, ny):
+            if 0 <= nx < row_len and 0 <= ny < col_len: 
+                net[mat[x][y]][mat[nx][ny]] = 1
+                net[mat[nx][ny]][mat[x][y]] = 1
+                _dfs(nx, ny)
+
+        _search_nxt(x, y, x-1, y)
+        _search_nxt(x, y, x+1, y)
+        _search_nxt(x, y, x, y-1)
+        _search_nxt(x, y, x, y+1)
+
+    _dfs(0, 0)
+
+    result = _convert_2darray_to_csv(net, num_node, num_queue, data_rate) 
     result.to_csv(header + ".csv", index=False)
     return net
 
 
-TOPO_FUNC = [line, ring, tree, mesh]
+TOPO_FUNC = [line, ring, tree, mesh, mesh_2d]
 
 
 def generate_flowset(
@@ -227,6 +240,7 @@ def generate_flowset(
             result.to_csv(header + ".csv", index=False)
             return
 
+        # NOTE: Prioritize uti(ES) <= 75% for traffic generation
         availble_es = np.argwhere(uti_ports <= 0.75).reshape(-1)
         if availble_es.size == 0:
             availble_es = np.array([x for x in range(num_es)])
@@ -251,3 +265,24 @@ def generate_flowset(
             i += 1
         else:
             continue
+
+
+def _convert_2darray_to_csv(net, num_node, num_queue, data_rate):
+    result = []
+    for i in range(num_node):
+        for j in range(num_node):
+            if net[i][j]:
+                link = []
+                link.append((i, j))
+                link.append(num_queue)
+                link.append(data_rate)
+                link.append(ERROR)
+                link.append(0)
+                result.append(link)
+
+    result = pd.DataFrame(result, columns=["link", "q_num", "rate", "t_proc", "t_prop"])
+    return result
+
+
+if __name__ == "__main__":
+    mesh_2d(9,1,1,"test")

tsnkit/data/generator.py

@@ -9,7 +9,9 @@
 import traceback
 import itertools
 import pandas as pd
+import numpy as np
 from tqdm import tqdm
+import os
 import networkx as nx
 from .dataset_spec import generate_flowset
 from .dataset_spec import TOPO_FUNC
@@ -27,6 +29,7 @@ def __init__(self, num_ins, num_stream, num_sw, period, size, deadline, topo):
         self.topo = [topo] if isinstance(topo, int) else topo
 
     def run(self, path):
+        os.makedirs(path, exist_ok=True)
         param_combinations = list(itertools.product(
                                 self.num_stream, 
                                 self.num_sw,
@@ -54,14 +57,14 @@ def run(self, path):
                             data_rate=1,
                             header=path + header + "_topo",
                         )
-                        _flowset = generate_flowset(
+                        _ = generate_flowset(
                             nx.DiGraph(net),
                             size,
                             period,
                             deadline,
                             num_stream,
                             num_sw,
-                            num_sw,
+                            num_sw if topo != 4 else int(np.sqrt(num_sw)-1) * 4,
                             path + header + "_task",
                         )
                         exp_info = [
@@ -126,7 +129,7 @@ def int_or_int_list(value):
         "--topo",
         type=int_or_int_list,
         default=0,
-        help="Topology type: 0-Line, 1-Ring, 2-Tree, 3-Mesh",
+        help="Topology type: 0-Line, 1-Ring, 2-Tree, 3-Mesh, 4-2dMesh",
     )
     parser.add_argument(
         "--output",