Update show_models.py

cache_utils/show_models.py

@@ -11,17 +11,104 @@ Using the following naming convention:
     ------
 """
 import matplotlib.pyplot as plt
+import seaborn as sns
+import pandas as pd
 import numpy as np
+import itertools
+import os
 
 nb_cores = 8
 nb_slices = 8
+num_core = nb_cores
 
 cores = list(range(nb_cores))
 slices = list(range(nb_slices))
 
+img_dir = os.getenv("PWD")+"/"
+def plot(filename, g=None):
+    if g is not None:
+        g.savefig(img_dir + filename)
+    else:
+        plt.savefig(img_dir + filename)
+    # tikzplotlib.save(
+    #   img_dir+filename+".tex",
+    #   axis_width=r'0.175\textwidth',
+    #   axis_height=r'0.25\textwidth'
+    # )
+    print(img_dir + filename, "saved")
+    plt.close()
+
+def ring_distance(x0, x1):
+    """
+    return (a, b) where `a` is the core distance and `b` the larger "ring step"
+    it is possible that going from 0->7 costs one more than 3->4
+    """
+    dist = abs(x0-x1)
+    if x0 // (num_core/2) != x1 // (num_core/2):
+        # côté du coeur différent
+        return min((num_core-1-dist, 2), (dist-1, 1))
+    else:
+        return dist, 0
+
+def slice_msg_distance(source, dest):
+    """
+    Si l'expéditeur est à l'extrémité d'une des lignes, il envoie toujours dans le même sens
+    (vers toute sa ligne d'abord), sinon, il prend le chemin le plus court
+    le bonus correspond au fait que 0->7 puisse coûter 1 de plus que 3->4
+    """
+    dist = abs(source-dest)
+    if source // (num_core/2) == dest // (num_core/2):
+        return (dist, 0)
+
+    # Pour aller de l'autre côté
+    up, down = (num_core-1-dist, 2), (dist-1, 1)
+    if source in [0, 7]:
+        return down
+    if source in [3, 4] or source in [2, 5]:
+        return up
+    if source in [1, 6]:
+        return min(up, down)
+
+    raise IndexError
+
+def ha_dist(core, is_QPI):
+    """
+    distance to Home Agent
+    """
+    if is_QPI:
+        if core < 4:
+            return core, 0
+        return 7-core, 1 # +1 for PCI
+
+    if core < 4:
+        return 3-core, 0
+    return core-4, 0
+
+def cclockwise_dist(source, dest):
+    base = (dest+8-source)%8
+    side_jump = 0
+    if source < 4 and dest >= 4:
+        side_jump = 1
+    elif source >= 4 and dest < 4:
+        side_jump = 2
+    return base, side_jump
+
+def cclockwise_ha_dist(core, is_QPI):
+    """
+    counter-clockwise distance to Home Agent
+    """
+    if is_QPI:
+        return cclockwise_dist(core, 7)
+    return cclockwise_dist(core, 3)
+
+def no_QPI_dist(source, dest):
+    """
+    Path not using QPI hop
+    """
+    return abs(source-dest), 1 if source // 4 != dest //4 else 0
 
 
-def hit_jump_ring():
+def miss():
     """
     - ini : initial cost
     - core_step : cost to go from one core to the following (eg 0 to 1)
@@ -29,26 +116,74 @@ def hit_jump_ring():
 
     Issue: on a same socket, we observe always the first 4 or last 4 patterns, but not mixed
     """
-    def hit(helper, slice, i, c, l):
+    def miss_topo(main_core, slice, i, c, l, k):
-        if helper // (nb_cores/2) != slice // (nb_cores/2):
+        x, y, z = slice_msg_distance(main_core, slice)
-            mini, maxi = min(slice, helper), max(slice, helper)
+        return i+c*x+l*y+k*z
-            return i+l+c*(min(mini+nb_cores-1-maxi, maxi-mini-1))
-        else:
-            return i+c*abs(slice-helper)
 
-    ini, core_step, ring_step = 4, 1, 5
+    ini, core_step, ring_step, other_ring_step = 4, 1, 5, 2
     fig, axs = plt.subplots(nrows=1, ncols=8, figsize=(15, 5))
 
-    for i, helper in enumerate(range(8)):
+    for i, slice in enumerate(range(8)):
-        axs[i].plot(cores, [hit(helper, slice, ini, core_step, ring_step) for slice in slices], "ro")
+        axs[i].plot(cores, [miss_topo(main_core, slice, ini, core_step, ring_step, other_ring_step) for main_core in cores], "ro")
-        axs[i].set_title(f"helper = {helper}")
+        axs[i].set_title(f"slice_group = {slice}")
-        axs[i].set_ylabel("clflush_hit_time")
+        axs[i].set_ylabel("clflush_miss_n")
-        axs[i].set_xlabel("slice_group")
+        axs[i].set_xlabel("main_core_fixed")
-        axs[i].set_ylim([0, 20])
+        axs[i].set_ylim([0, 25])
 
     plt.tight_layout()
     plt.show()
 
 
+def hit():
+    """
+    - ini : initial cost
+    - core_step : cost to go from one core to the following (eg 0 to 1)
+    - ring_step : cost to go from one line to the other (eg 0 to 7)
 
-hit_jump_ring()
+    Issue: on a same socket, we observe always the first 4 or last 4 patterns, but not mixed
+    """
+    def hit_topo(main, helper, slice_g, i, c, l):
+        helper = helper%8
+
+        main_slice_local = slice_msg_distance(slice_g, main)
+        slice_QPI = cclockwise_dist(0, slice_g) # clockwise
+        QPI_slice_r = cclockwise_dist(0, slice_g)
+        slice_r_helper = slice_msg_distance(slice_g, helper)
+
+        costs = (main_slice_local[0]+slice_QPI[0]+QPI_slice_r[0]+slice_r_helper[0], main_slice_local[1]+slice_QPI[1]+QPI_slice_r[1]+slice_r_helper[1])
+        return ini+costs[0]*c+costs[1]*l
+
+
+    ini, core_step, ring_step = 12, 1, 1.5
+    fig, axs = plt.subplots(nrows=1, ncols=8, figsize=(15, 5))
+
+    # Define the ranges for x, y, z
+    main = range(8)
+    helper = range(8, 16)
+    slice_g = range(8)
+
+    # Create a DataFrame with all combinations of x, y, and z
+    data = pd.DataFrame([
+            (x, y, z) for z in slice_g
+            for x, y in itertools.product(main, helper)
+        ],
+        columns=['main', 'helper', 'slice_group']
+    )
+
+    # Define the function
+    def my_function(x):
+        return hit_topo(x["main"], x["helper"], x["slice_group"], ini, core_step, ring_step)
+
+    # Apply the function to create a new column
+    data['predicted_hit'] = data.apply(my_function, axis=1)
+
+    fig = sns.FacetGrid(data, col="main", row="helper")
+    fig.map(sns.scatterplot, "slice_group", "predicted_hit", color="r", marker="+")
+    fig.map(sns.scatterplot, "slice_group", "predicted_hit", color="r", marker="x")
+    fig.set_titles(col_template="$A$ = {col_name}", row_template="$V$ = {row_name}")
+
+    plot("model_hit.png", g=fig)
+
+
+
+hit()