convert extracted trends to schedcat overhead profiles

1 files changed, 86 insertions, 0 deletions

diff --git a/distill_schedcat_profiles.py b/distill_schedcat_profiles.py
new file mode 100755
index 0000000..fa3c482
--- /dev/null
+++ b/distill_schedcat_profiles.py
@@ -0,0 +1,86 @@
+#!/usr/bin/env python
+
+
+from plot import decode
+from util import load_csv_file, write_csv_file
+
+
+from os.path  import splitext, basename
+from optparse import make_option as o, OptionParser
+from collections import defaultdict
+
+options = None
+opts = [
+    # output options
+    o('-p', '--prefix', action='store', dest='prefix'),
+    o('-t', '--tag', action='store', dest='tag'),
+]
+
+defaults = {
+    # output options
+    'prefix' : '',
+    'tag'    : '',
+}
+
+
+class DataFile(object):
+    def __init__(self, fname):
+        bname     = basename(fname)
+        name, ext = splitext(bname)
+        self.conf = decode(name)
+        self.sched = self.conf['scheduler']
+        self.overhead = self.conf['overhead']
+        self.host  = self.conf['host']
+        self.name = name
+        self.data = load_csv_file(fname)
+
+
+def group_by_scheduler(files):
+    g = defaultdict(list)
+    for f in files:
+        g[f.sched].append(f)
+    return g
+
+NUM_INDEX = 2
+WC_INDEX  = 5
+AVG_INDEX = 6
+
+def write_profiles(files):
+    host   = files[0].host
+    sched  = files[0].sched
+    num    = len(files[0].data)
+
+    def row(i, idx):
+        return [int(files[0].data[i][NUM_INDEX])] + [f.data[i][idx] for f in files]
+
+    avg  = [row(i, AVG_INDEX) for i in xrange(num)]
+    wc   = [row(i, WC_INDEX) for i in xrange(num)]
+    header = ['NUM_TASKS'] + [f.overhead.replace('-', '_') for f in files]
+
+    avg_name = '%soh_host=%s_scheduler=%s_stat=avg%s.csv' % \
+        (options.prefix, host, sched, options.tag)
+    wc_name  = '%soh_host=%s_scheduler=%s_stat=wc%s.csv' % \
+        (options.prefix, host, sched, options.tag)
+
+    print 'Generating %s.' % (avg_name)
+    write_csv_file(avg_name, [header] + avg, width=15)
+    print 'Generating %s.' % (wc_name)
+    write_csv_file(wc_name, [header] + wc, width=15)
+
+if __name__ == '__main__':
+    parser = OptionParser(option_list=opts)
+    parser.set_defaults(**defaults)
+    (options, files) = parser.parse_args()
+
+    try:
+        data = []
+        for f in files:
+            try:
+                data.append(DataFile(f))
+            except IOError as err:
+                print "Skipping %s.\n(%s)" % (f, err)
+        groups = group_by_scheduler(data)
+        for sched in groups:
+            write_profiles(groups[sched])
+    except KeyboardInterrupt:
+        pass