[GR-33233] Resize meso benchmarks and add tool for simple analysis of benchmarks.

PullRequest: graalpython/1934

Author: steve-s

graalpython/com.oracle.graal.python.benchmarks/python/meso/backprop_rodinia-sized2.py

@@ -0,0 +1,213 @@
+# Copyright (c) 2020, 2021, Oracle and/or its affiliates. All rights reserved.
+# DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
+#
+# The Universal Permissive License (UPL), Version 1.0
+#
+# Subject to the condition set forth below, permission is hereby granted to any
+# person obtaining a copy of this software, associated documentation and/or
+# data (collectively the "Software"), free of charge and under any and all
+# copyright rights in the Software, and any and all patent rights owned or
+# freely licensable by each licensor hereunder covering either (i) the
+# unmodified Software as contributed to or provided by such licensor, or (ii)
+# the Larger Works (as defined below), to deal in both
+#
+# (a) the Software, and
+#
+# (b) any piece of software and/or hardware listed in the lrgrwrks.txt file if
+# one is included with the Software each a "Larger Work" to which the Software
+# is contributed by such licensors),
+#
+# without restriction, including without limitation the rights to copy, create
+# derivative works of, display, perform, and distribute the Software and make,
+# use, sell, offer for sale, import, export, have made, and have sold the
+# Software and the Larger Work(s), and to sublicense the foregoing rights on
+# either these or other terms.
+#
+# This license is subject to the following condition:
+#
+# The above copyright notice and either this complete permission notice or at a
+# minimum a reference to the UPL must be included in all copies or substantial
+# portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+# SOFTWARE.
+
+"""
+ ******************************************************************
+ * HISTORY
+ * 15-Oct-94  Jeff Shufelt (js), Carnegie Mellon University
+ *	Prepared for 15-681, Fall 1994.
+ * Modified by Shuai Che
+ ******************************************************************
+"""
+
+import random
+import math
+
+# eta value
+ETA = 0.3
+
+# momentum value
+MOMENTUM = 0.3
+
+# (16, 1 can not be changed)
+n_hidden = 16
+n_out = 1
+
+
+def bpnn_layerforward(l1, l2, conn, n1, n2):
+    ## Set up thresholding unit ##
+    l1[0] = 1.0
+    ## For each unit in second layer ##
+    r1 = 1 + n2
+    r2 = 1 + n1
+    for j in range(1, r1):
+        ## Compute weighted Sum of its inputs ##
+        Sum = 0.0
+        for k in range(r2):
+            Sum += conn[k*r1 + j] * l1[k]
+
+        l2[j] = (1.0 / (1.0 + math.exp(-Sum)))
+
+
+def bpnn_output_error(delta, target, output, nj):
+    errSum = 0.0
+    for j in range(1, 1 + nj):
+        o = output[j]
+        t = target[j]
+        v = o * (1.0 - o) * (t - o)
+        delta[j] = v
+        errSum += abs(v)
+
+    return errSum
+
+
+def bpnn_hidden_error(delta_h, nh, delta_o, no, who, hidden):
+    errSum = 0.0
+    for j in range(1, 1 + nh):
+        Sum = 0.0
+        for k in range(1, 1 + no):
+            Sum += delta_o[k] * who[j * (1 + no) + k]
+        h = hidden[j]
+        delta_h[j] = h * (1.0 - h) * Sum
+        errSum += abs(delta_h[j])
+
+    return errSum
+
+
+def bpnn_adjust_weights(delta, ndelta, ly, nly, w, oldw):
+    ly[0] = 1.0
+    for j in range(1, (1 + ndelta)):
+        for k in range(nly + 1):
+            val = ETA * delta[j] * ly[k]
+            val += MOMENTUM * oldw[k*(1 + ndelta) + j]
+            oldw[k*(1 + ndelta) + j] = val
+            w[k*(1 + ndelta) + j] += val
+
+
+def bpnn_train_kernel(_iu_list, _hu_list, _iw_list, _ou_list, _hw_list, _od_list, _t_list, _hd_list, _hw_prev_list, _iw_prev_list, layer_size):
+    bpnn_layerforward(_iu_list, _hu_list, _iw_list, layer_size, n_hidden)
+    bpnn_layerforward(_hu_list, _ou_list, _hw_list, n_hidden, n_out)
+    out_err = bpnn_output_error(_od_list, _t_list, _ou_list, n_out)
+    hid_err = bpnn_hidden_error(
+        _hd_list, n_hidden, _od_list, n_out, _hw_list, _hu_list)
+    bpnn_adjust_weights(_od_list, n_out, _hu_list,
+                        n_hidden, _hw_list, _hw_prev_list)
+    bpnn_adjust_weights(_hd_list, n_hidden, _iu_list,
+                        layer_size, _iw_list, _iw_prev_list)
+
+    return (out_err, hid_err)
+
+
+class Data:
+    def __init__(self):
+        self._iu_list = None
+        self._hw_list = None
+        self._hu_list = None
+        self._ou_list = None
+        self._hd_list = None
+        self._od_list = None
+        self._iw_list = None
+        self._iw_prev_list = None
+        self._hw_prev_list = None
+        self._t_list = None
+
+        self.zeros_list = None
+        self.random_list = None
+
+
+data = Data()
+
+default_size = 2 ** 16
+
+
+def measure(layer_size=default_size):
+    print("Starting training kernel")
+    bpnn_train_kernel(data._iu_list, data._hu_list, data._iw_list, data._ou_list, data._hw_list,
+                      data._od_list, data._t_list, data._hd_list, data._hw_prev_list, data._iw_prev_list, layer_size)
+
+
+def __benchmark__(layer_size=default_size):
+    measure(layer_size)
+
+
+def __setup__(layer_size=default_size):
+    random.seed(7)
+    print("Input layer size : %d" % layer_size)
+    # Creates a new fully-connected network from scratch,
+    # with the given numbers of input, hidden, and output units.
+    # Threshold units are automatically included.  All weights are
+    # randomly initialized.
+
+    # Space is also allocated for temporary storage (momentum weights,
+    # error computations, etc).
+
+    ## the input units ##
+    data._iu_list = [random.random() for i in range(layer_size + 1)]
+
+    ## weights from hidden to output layer ##
+    data._hw_list = [random.random()
+                     for i in range((n_out + 1) * (n_hidden + 1))]
+
+    ## the hidden units ##
+    data._hu_list = [0. for i in range(n_hidden + 1)]
+    ## the output units ##
+    data._ou_list = [0. for i in range(n_out + 1)]
+
+    ## storage for hidden unit error ##
+    data._hd_list = [0. for i in range(n_hidden + 1)]
+    ## storage for output unit error ##
+    data._od_list = [0. for i in range(n_out + 1)]
+
+    ## weights from input to hidden layer ##
+    data._iw_list = [0. for i in range((n_hidden + 1) * (layer_size + 1))]
+
+    ## The next two are for momentum ##
+    ## previous change on input to hidden wgt ##
+    data._iw_prev_list = [0. for i in range((n_hidden + 1) * (layer_size + 1))]
+    ## previous change on hidden to output wgt ##
+    data._hw_prev_list = [0. for i in range((n_out + 1) * (n_hidden + 1))]
+
+    ## storage for target vector ##
+    data._t_list = [0.1 for i in range(n_out + 1)]
+
+    data.zeros_list = [data._hu_list, data._ou_list, data._hd_list,
+                       data._od_list, data._iw_list, data._iw_prev_list, data._hw_prev_list]
+    data.random_list = [data._iu_list, data._hw_list]
+
+
+def __cleanup__(layer_size=default_size):
+    # clean up written data
+    for l in data.zeros_list:
+        for i in range(len(l)):
+            l[i] = 0.
+    for l in data.random_list:
+        for i in range(len(l)):
+            l[i] = random.random()
+    for i in range(len(data._t_list)):
+        data._t_list[i] = 0.1