Reduce thread block size

=> ~780 MLUPS

2 files changed, 26 insertions, 28 deletions

diff --git a/codegen_lbm.py b/codegen_lbm.py
index 29c2dad..235a548 100644
--- a/codegen_lbm.py
+++ b/codegen_lbm.py
@@ -173,10 +173,10 @@ class D2Q9_BGK_Lattice:
     def evolve(self):
         if self.tick:
             self.tick = False
-            self.program.collide_and_stream(self.queue, (self.nX,self.nY), (64,1), self.cl_pop_a, self.cl_pop_b, self.cl_moments, self.cl_material)
+            self.program.collide_and_stream(self.queue, (self.nX,self.nY), (32,1), self.cl_pop_a, self.cl_pop_b, self.cl_moments, self.cl_material)
         else:
             self.tick = True
-            self.program.collide_and_stream(self.queue, (self.nX,self.nY), (64,1), self.cl_pop_b, self.cl_pop_a, self.cl_moments, self.cl_material)
+            self.program.collide_and_stream(self.queue, (self.nX,self.nY), (32,1), self.cl_pop_b, self.cl_pop_a, self.cl_moments, self.cl_material)
 
     def sync(self):
         self.queue.finish()
diff --git a/lbm_codegen.ipynb b/lbm_codegen.ipynb
index 34e9e6b..0b85e19 100644
--- a/lbm_codegen.ipynb
+++ b/lbm_codegen.ipynb
@@ -7,8 +7,6 @@
    "outputs": [],
    "source": [
     "from sympy import *\n",
-    "import matplotlib.pyplot as plt\n",
-    "\n",
     "init_printing()"
    ]
   },
@@ -132,7 +130,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 22,
+   "execution_count": 8,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -141,7 +139,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 23,
+   "execution_count": 9,
    "metadata": {},
    "outputs": [
     {
@@ -151,7 +149,7 @@
        "       f_next_6, f_next_7, f_next_8], dtype=object)"
       ]
      },
-     "execution_count": 23,
+     "execution_count": 9,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -163,7 +161,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 24,
+   "execution_count": 10,
    "metadata": {},
    "outputs": [
     {
@@ -173,7 +171,7 @@
        "       f_curr_6, f_curr_7, f_curr_8], dtype=object)"
       ]
      },
-     "execution_count": 24,
+     "execution_count": 10,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -185,7 +183,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 25,
+   "execution_count": 11,
    "metadata": {},
    "outputs": [
     {
@@ -199,7 +197,7 @@
        "_curr_7 + f_curr_8"
       ]
      },
-     "execution_count": 25,
+     "execution_count": 11,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -211,7 +209,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 55,
+   "execution_count": 12,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -221,7 +219,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 56,
+   "execution_count": 13,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -231,7 +229,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 26,
+   "execution_count": 14,
    "metadata": {},
    "outputs": [
     {
@@ -245,7 +243,7 @@
        "⎣u_y⎦"
       ]
      },
-     "execution_count": 26,
+     "execution_count": 14,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -257,7 +255,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 27,
+   "execution_count": 15,
    "metadata": {},
    "outputs": [
     {
@@ -353,7 +351,7 @@
        "36         36         36   ⎠⎦"
       ]
      },
-     "execution_count": 27,
+     "execution_count": 15,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -373,7 +371,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 28,
+   "execution_count": 16,
    "metadata": {},
    "outputs": [
     {
@@ -524,7 +522,7 @@
        " ⎦"
       ]
      },
-     "execution_count": 28,
+     "execution_count": 16,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -536,7 +534,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 29,
+   "execution_count": 17,
    "metadata": {},
    "outputs": [
     {
@@ -612,7 +610,7 @@
        "                      ⎝                ⎝                 2      ⎠⎠⎦⎠"
       ]
      },
-     "execution_count": 29,
+     "execution_count": 17,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -623,7 +621,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 30,
+   "execution_count": 18,
    "metadata": {},
    "outputs": [
     {
@@ -678,7 +676,7 @@
        "                                               ⎦⎠"
       ]
      },
-     "execution_count": 30,
+     "execution_count": 18,
      "metadata": {},
      "output_type": "execute_result"
     }
@@ -690,7 +688,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 31,
+   "execution_count": 19,
    "metadata": {},
    "outputs": [
     {
@@ -726,7 +724,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 32,
+   "execution_count": 20,
    "metadata": {},
    "outputs": [
     {
@@ -752,7 +750,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 49,
+   "execution_count": 21,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -764,7 +762,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 53,
+   "execution_count": 22,
    "metadata": {},
    "outputs": [
     {
@@ -785,7 +783,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 54,
+   "execution_count": 23,
    "metadata": {},
    "outputs": [
     {