Displaying measured and theoretical model side to side.

pipeline.py

@@ -7,7 +7,7 @@ Execute all the stuff:
 - comparing graphs
 
 Dependencies:
-pip3 install tqdm numpy matlpotlib adjusttext 
+pip3 install tqdm numpy matlpotlib adjusttext
 """
 
 import argparse
@@ -22,6 +22,8 @@ import csv
 import matplotlib.pyplot as plt
 import numpy as np
 from adjustText import adjust_text
+from operator import itemgetter
+from matplotlib.ticker import MultipleLocator
 
 
 parser = argparse.ArgumentParser(description="Generate graphs, execute them, and then evaluate their quality", \
@@ -55,7 +57,7 @@ def get_costs(csvname):
     """Get csv file with execution times and compute average execution time
         for a cycle as well as total execution time"""
     with open(csvname, "r") as csvfile:
-        csvfile.readline() # To remove first line where there are number of ndoes and edges
+        csvfile.readline() # To remove first line where there are number of nodes and edges
         csvreader = csv.DictReader(csvfile, delimiter='\t')
         total=0.
         nb_rows=0
@@ -64,8 +66,6 @@ def get_costs(csvname):
             nb_rows += 1
         return total / float(nb_rows), total
 
-
-
 def process_graph(graph):
     tqdm.write("Processing " + graph, end=" ")
     basename,ext = os.path.splitext(graph)
@@ -83,7 +83,7 @@ def process_graph(graph):
     tqdm.write("Executing graphs")
     for graph in tqdm(glob.iglob("*.ag")):
         tqdm.write(graph)
-        subprocess.run([graph_exec, "-m", "-b", graph], check=True)
+        subprocess.run([graph_exec, "-m", "-b", "-c", "60000", graph], check=True)
         # Get execution times for reports (-m option)
         basename,_ = os.path.splitext(os.path.basename(graph))
         reports = glob.glob("*"+basename + "*.csv")
@@ -130,29 +130,79 @@ def results_to_csv(graphname, qualities, costs):
             result["Total"] = total
             writer.writerow(result)
 
-def plot(qualities, costs):
+def load_theo(filename):
+    """Load quality and cost from the theoretical model"""
+    with open(filename, "r", newline='') as csvfile:
+        csvreader = csv.DictReader(csvfile, delimiter='\t')
+        qualities=[]
+        costs=[]
+        for row in csvreader:
+            qualities.append(row["Quality"])
+            costs.append(row["Cost"])
+        return qualities, costs
+
+def sort_by_quality(qualities, costs):
+    return [[*x] for x in zip(*sorted(zip(qualities, costs), key=itemgetter(0)))]
+
+def plot(qualities, costs, qualities_th, costs_th):
     q = []
     c_cycle = []
     c_total = []
     texts= []
+
+    fig, axes = plt.subplots(2,1)
+
+    ax1= axes[0]
+    ax2 = axes[1]
+
     for k in sorted(qualities.keys()):
         q.append(qualities[k])
         cycle, total = costs[k]
         c_cycle.append(cycle)
         c_total.append(total)
-        texts.append(plt.text(qualities[k], cycle, k, ha='center', va='center'))
+        name = k.split("-")[-1]
+        texts.append(ax1.text(qualities[k], cycle, name, ha='center', va='center'))
     q.append(1.)
     name = list(sorted(costs.keys()))[0]
     cost, total = costs[name]
     c_cycle.append(cost)
     c_total.append(total)
-    texts.append(plt.text(1.0, cost, name, ha='center', va='center'))
-    plt.plot(q, c_cycle, 'ro')
-    #plt.plot(q, c_total)
-    plt.ylabel("cost per cycle (ms)")
-    plt.xlabel("quality")
-    adjust_text(texts)
+    texts.append(ax1.text(1.0, cost, "0", ha='center', va='center'))
+
+    q,c_cycle = sort_by_quality(q, c_cycle)
+
+    #print("Measured: ", q, c_cycle)
+
+    color='tab:red'
+
+    ax1.set_ylabel("cost per cycle (ms)")
+    ax1.set_xlabel("quality")
+    ax1.scatter(q, c_cycle,  label="Measured", color=color)
+    #ax1.tick_params(axis='y', labelcolor=color)
+
+
+    texts_th = []
+    for (i, (quality, cost)) in enumerate(zip(qualities_th, costs_th)):
+        texts_th.append(ax2.text(quality, cost, str(i), ha='center', va='center'))
+
+    color = 'tab:blue'
+    ax2.set_ylabel("cost")
+    ax2.set_xlabel("quality")
+
+    qualities_th,costs_th = sort_by_quality(qualities_th, costs_th)
+
+    ax2.scatter(qualities_th, costs_th,  label="Model", color=color)
+    #ax2.tick_params(axis='y', color=color)
+
+
+    adjust_text(texts, ax=ax1)
+    adjust_text(texts_th, ax=ax2)
+
+
+    fig.tight_layout()
+    fig.legend()
     plt.show()
+    #print("limits are: x1=[", ax1.get_xlim(), "], x2=[", ax2.get_xlim(), "]")
 
 if args.graph:
     for graph in tqdm(args.graph):
@@ -162,8 +212,9 @@ if args.graph:
         tqdm.write("Costs are " + str(costs))
         basename,_ = os.path.splitext(os.path.basename(graph))# We stay in the directory created in process_graph
         results_to_csv(basename + "-exec-report", qualities, costs)
+        q_th, c_th = load_theo(basename + "-theo.csv")
         # Display in a graph
-        plot(qualities, costs)
+        plot(qualities, costs, q_th, c_th)
         os.chdir("..")
 elif args.nodes:
     print("Processing all graphs ", end="")

src/audiograph.rs

@@ -351,7 +351,7 @@ impl AudioGraph {
                 else {//We are not after a resampler so we propagate the min buffer size of the previous edges.
                     (buf_size, samplerate)
                 };
-                println!("Buffer size={}; samplerate={}", new_buf_size, new_samplerate);
+                //println!("Buffer size={}; samplerate={}", new_buf_size, new_samplerate);
                 //Modify all outcoming buffer sizes
                 let mut output_edges = self.outputs_mut(node);
                 while let Some(edge) = output_edges.next_edge(&self.graph) {

src/audiograph_exec.rs

@@ -29,7 +29,7 @@ use audio_adaptive::sndfile;
 
 const CHANNELS: i32 = 1;
 const SAMPLE_RATE: u32 = 44_100;
-const NB_CYCLES : u32 = 6000;
+const NB_CYCLES : u32 = 12000;
 const FRAMES_PER_BUFFER : usize = 64;
 
 #[derive(Clone, Copy, Debug)]