Add two files to generate and test options of the Genetic algorithm

generate_genetic_options will save a file containg the
best found options by the genetic algorithm.
eval_genetic_options would benchmark these options from the file.

Author: Jules Pondard

python/experimental/options_search/eval_genetic_options.py

@@ -0,0 +1,21 @@
+import tensor_comprehensions as tc
+import tensor_comprehensions.tclib as tclib
+#import ipdb
+cache = tc.MappingOptionsCache("genetic_savedopt_conv_default.txt")
+import utils
+tc_code, tc_name, inp, init_input_sz = utils.get_convolution_example()
+print("divs : " +str(utils.getAllDivs(inp)))
+tup = cache.load(tc_code, tc_name, inp, 1)
+if(tup == []):
+    exit()
+best_options, = tup
+best_options = best_options.getDict()
+optsVect = utils.getRawVectorFromTcOpt(best_options)
+#optsVect = utils.catVec_to_optVec(optsVect)
+opts = utils.optionsFromVector(optsVect)
+print(opts)
+
+tc_prog = tc.compile(tc_code, tc_name, opts, *inp)
+time = my_utils.evalTime(opts, estimator="median")
+print(time)
+

python/experimental/options_search/generate_genetic_options.py

@@ -0,0 +1,16 @@
+import numpy as np
+#import ipdb
+import torch
+import tensor_comprehensions as tc
+
+import utils
+
+tc_code, tc_name, inp, init_input_sz = utils.get_convolution_example()
+#config = tc.autotuner_settings
+#config["pop_size"]=50
+#config["generations"]=1
+
+opts = tc.MappingOptions("naive")
+print(opts)
+
+tc.autotune(tc_code, tc_name, *inp, starting_options=opts, cache_filename="genetic_savedopt_conv_default.txt", store_to_cache=True)