Add/float64 benchmarks

Makefile

@@ -12,18 +12,24 @@ benchmark-mojo:
 	mojo mojo_impl/naive.mojo 10000
 	mojo mojo_impl/naive.mojo 100000
 	mojo mojo_impl/naive.mojo 1000000
-	mojo mojo_impl/naive.mojo 10000000
+
+# segfault :(
+# mojo mojo_impl/naive.mojo 10000000
 
 	mojo mojo_impl/optimized_a.mojo 100
 	mojo mojo_impl/optimized_a.mojo 1000
 	mojo mojo_impl/optimized_a.mojo 10000
 	mojo mojo_impl/optimized_a.mojo 100000
 	mojo mojo_impl/optimized_a.mojo 1000000
-	mojo mojo_impl/optimized_a.mojo 10000000
+
+# segfault :(
+# mojo mojo_impl/optimized_a.mojo 10000000
 
 	mojo mojo_impl/optimized_b.mojo 100
 	mojo mojo_impl/optimized_b.mojo 1000
 	mojo mojo_impl/optimized_b.mojo 10000
 	mojo mojo_impl/optimized_b.mojo 100000
 	mojo mojo_impl/optimized_b.mojo 1000000
-	mojo mojo_impl/optimized_b.mojo 10000000
+
+# segfault :(
+#	mojo mojo_impl/optimized_b.mojo 10000000

README.md

@@ -61,8 +61,8 @@ sizes.
 
 3. In addition to being performance winners, the Mojo variants are
 parameterized by the number of dimensions (`dims`) and by data type (`dtype`).
-In other words, the same generic code can run, for example, `float16`,
-`float64` or with 3, 4 or more dimensions. In GIS systems the number of
+In other words, the same generic code can run, for example, `int16`, `float16`,
+`float64`, and with 3, 4 or more dimensions. In GIS systems the number of
 dimensions is sometimes referred to as XY, XYZ, or XYZM, where Z is "height",
 and M is "measure".
 

mojo_impl/naive.mojo

@@ -46,7 +46,7 @@ fn envelope[dtype: DType, dims: Int](tensor: Tensor[dtype]) -> SIMD[dtype, 2 * d
 
     return result
 
-alias dtype = DType.float32
+alias dtype = DType.float64
 alias dims = 2
 
 fn main() raises:
@@ -76,3 +76,4 @@ fn main() raises:
     print("microsecs:", secs * 10 ** 6)
     print("ms:", ms)
     print("s:", secs)
+    print()

mojo_impl/optimized_a.mojo

@@ -57,7 +57,7 @@ fn envelope[dtype: DType, dims: Int](tensor: Tensor[dtype]) -> SIMD[dtype, 2 * d
     return result
 
 
-alias dtype = DType.float32
+alias dtype = DType.float64
 alias dims = 2
 
 
@@ -87,3 +87,4 @@ fn main() raises:
     print("microsecs:", secs * 10 ** 6)
     print("ms:", ms)
     print("s:", secs)
+    print()

mojo_impl/optimized_b.mojo

@@ -59,7 +59,7 @@ fn envelope[dtype: DType, dims: Int](tensor: Tensor[dtype]) -> SIMD[dtype, 2 * d
 
     return result.simd_load[2 * dims]()
 
-alias dtype = DType.float32
+alias dtype = DType.float64
 alias dims = 2
 
 fn main() raises:

py_impl/naive_benchmark.ipynb

@@ -16,12 +16,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "multipoint_10_2 = np.array(np.random.rand(2, 10 ** 2), dtype=np.float32)\n",
-    "multipoint_10_3 = np.array(np.random.rand(2, 10 ** 3), dtype=np.float32)\n",
-    "multipoint_10_4 = np.array(np.random.rand(2, 10 ** 4), dtype=np.float32)\n",
-    "multipoint_10_5= np.array(np.random.rand(2, 10 ** 5), dtype=np.float32)\n",
-    "multipoint_10_6 = np.array(np.random.rand(2, 10 ** 6), dtype=np.float32)\n",
-    "multipoint_10_7 = np.array(np.random.rand(2, 10 ** 7), dtype=np.float32)"
+    "dtype=np.float64\n",
+    "\n",
+    "multipoint_10_2 = np.array(np.random.rand(2, 10 ** 2), dtype)\n",
+    "multipoint_10_3 = np.array(np.random.rand(2, 10 ** 3), dtype)\n",
+    "multipoint_10_4 = np.array(np.random.rand(2, 10 ** 4), dtype)\n",
+    "multipoint_10_5= np.array(np.random.rand(2, 10 ** 5), dtype)\n",
+    "multipoint_10_6 = np.array(np.random.rand(2, 10 ** 6), dtype)\n",
+    "multipoint_10_7 = np.array(np.random.rand(2, 10 ** 7), dtype)"
    ]
   },
   {
@@ -33,7 +35,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "13.5 µs ± 183 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
+      "11.1 µs ± 16.4 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
      ]
     }
    ],
@@ -52,7 +54,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "103 µs ± 95 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)\n"
+      "101 µs ± 99.5 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)\n"
      ]
     }
    ],
@@ -71,7 +73,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "1 ms ± 6.2 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)\n"
+      "1.01 ms ± 2.05 µs per loop (mean ± std. dev. of 7 runs, 1,000 loops each)\n"
      ]
     }
    ],
@@ -90,7 +92,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "10.5 ms ± 132 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
+      "10.4 ms ± 10.8 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
      ]
     }
    ],
@@ -109,7 +111,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "109 ms ± 399 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
+      "109 ms ± 532 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)\n"
      ]
     }
    ],
@@ -128,15 +130,21 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "1.09 s ± 2.14 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
+      "1.09 s ± 1.84 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)\n"
      ]
     }
    ],
    "source": [
     "%%timeit\n",
-    "\n",
     "envelope(x_coords=list(multipoint_10_7[0]), y_coords=list(multipoint_10_7[1]))"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {

py_impl/optimized_benchmark.ipynb

@@ -16,12 +16,14 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "multipoint_10_2 = np.array(np.random.rand(2, 10 ** 2), dtype=np.float32)\n",
-    "multipoint_10_3 = np.array(np.random.rand(2, 10 ** 3), dtype=np.float32)\n",
-    "multipoint_10_4 = np.array(np.random.rand(2, 10 ** 4), dtype=np.float32)\n",
-    "multipoint_10_5= np.array(np.random.rand(2, 10 ** 5), dtype=np.float32)\n",
-    "multipoint_10_6 = np.array(np.random.rand(2, 10 ** 6), dtype=np.float32)\n",
-    "multipoint_10_7 = np.array(np.random.rand(2, 10 ** 7), dtype=np.float32)"
+    "dtype = np.float64\n",
+    "\n",
+    "multipoint_10_2 = np.array(np.random.rand(2, 10 ** 2), dtype)\n",
+    "multipoint_10_3 = np.array(np.random.rand(2, 10 ** 3), dtype)\n",
+    "multipoint_10_4 = np.array(np.random.rand(2, 10 ** 4), dtype)\n",
+    "multipoint_10_5= np.array(np.random.rand(2, 10 ** 5), dtype)\n",
+    "multipoint_10_6 = np.array(np.random.rand(2, 10 ** 6), dtype)\n",
+    "multipoint_10_7 = np.array(np.random.rand(2, 10 ** 7), dtype)"
    ]
   },
   {
@@ -33,7 +35,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "2.84 µs ± 21.6 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
+      "2.75 µs ± 9.27 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
      ]
     }
    ],
@@ -52,7 +54,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "3.09 µs ± 57.9 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
+      "3.07 µs ± 12.3 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
      ]
     }
    ],
@@ -71,7 +73,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "4.14 µs ± 10.2 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
+      "6.08 µs ± 21.2 ns per loop (mean ± std. dev. of 7 runs, 100,000 loops each)\n"
      ]
     }
    ],
@@ -90,7 +92,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "22.5 µs ± 62.3 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)\n"
+      "40.1 µs ± 137 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)\n"
      ]
     }
    ],
@@ -109,7 +111,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "195 µs ± 229 ns per loop (mean ± std. dev. of 7 runs, 10,000 loops each)\n"
+      "392 µs ± 643 ns per loop (mean ± std. dev. of 7 runs, 1,000 loops each)\n"
      ]
     }
    ],
@@ -128,7 +130,7 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "2.64 ms ± 16.7 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
+      "5.2 ms ± 17.5 µs per loop (mean ± std. dev. of 7 runs, 100 loops each)\n"
      ]
     }
    ],