Initial skeleton of Subzero.

szdiff.py

Index: szdiff.py
diff --git a/szdiff.py b/szdiff.py
new file mode 100755
index 0000000000000000000000000000000000000000..70805972910db1b60ef1cc71f9d5b185ca84f08d
--- /dev/null
+++ b/szdiff.py
@@ -0,0 +1,61 @@
+#!/usr/bin/env python2
+
+# Runs llvm2ice on an input .ll file, and compares the output against
+# the input.
+#
+# Before comparing, the input file is massaged to remove comments,
+# blank lines, global variable definitions, external function
+# declarations, and possibly other patterns that llvm2ice does not
+# handle.
+#
+# The output file and the massaged input file are compared line by
+# line for differences.  However, there is a regex defined such that
+# if the regex matches a line in the input file, that line and the
+# corresponding line in the output file are ignored.  This lets us
+# ignore minor differences such as inttoptr and ptrtoint, and printing
+# of floating-point constants.
+#
+# On success, no output is produced.  On failure, each mismatch is
+# printed as two lines, one starting with "SZ" and one starting with
+# "LL".

[JF, 2014/03/21 02:11:30]

I'd expect this to be a triply-quoted string instead of a comment. Python style.

[Jim Stichnoth, 2014/03/21 21:33:08]

Done.

+
+import argparse
+import subprocess
+import re
+
+if __name__ == '__main__':
+    argparser = argparse.ArgumentParser()

[JF, 2014/03/21 02:11:30]

description='...'

[Jim Stichnoth, 2014/03/21 21:33:08]

Done.

+    argparser.add_argument('llfile')

[JF, 2014/03/21 02:11:30]

type=argparse.FileType('r') would allow you to read the file straight up.
type=string would at least make it a string.

Also, required=True, metavar='FILE', help='...'

[Jim Stichnoth, 2014/03/21 21:33:08]

Done.  (BTW 'type=string' is not valid, and the default type is a string.)
Done.  (BTW 'required' is invalid for positional arguments.)

+    argparser.add_argument('--llvm2ice')

[JF, 2014/03/21 02:11:30]

Same as above, but with default='./llvm2ice' and required=False

[Jim Stichnoth, 2014/03/21 21:33:08]

Done.

+    args = argparser.parse_args()
+
+    infile = args.llfile
+    llvm2ice = args.llvm2ice if args.llvm2ice else './llvm2ice'

[JF, 2014/03/21 02:11:30]

This can be dropped with the above default=

[Jim Stichnoth, 2014/03/21 21:33:08]

Done.

+    command = [llvm2ice, '-verbose', 'inst', '-notranslate', infile]
+    p = subprocess.Popen(command, stdout=subprocess.PIPE)
+    sz_out, stderr = p.communicate()
+    sed_command = ['sed', '-e', 's/;.*//', '-e', 's/ *$//', infile]
+    grep1_command = ['grep', '-v', '^ *$']
+    grep2_command = ['grep', '-v', '^declare']
+    grep3_command = ['grep', '-v', '^@']
+    sed = subprocess.Popen(sed_command, stdout=subprocess.PIPE)
+    grep1 = subprocess.Popen(grep1_command, stdin=sed.stdout,
+                             stdout=subprocess.PIPE)
+    grep2 = subprocess.Popen(grep2_command, stdin=grep1.stdout,
+                             stdout=subprocess.PIPE)
+    grep3 = subprocess.Popen(grep3_command, stdin=grep2.stdout,
+                             stdout=subprocess.PIPE)

[JF, 2014/03/21 02:11:30]

You know I sed, but in this case http://docs.python.org/2/howto/regex.html ?

[Jim Stichnoth, 2014/03/21 21:33:08]

OK OK, fixed.

+    llc_out = grep3.communicate()[0]
+    match = re.compile('|'.join([' -[0-9]+',     # negative constants
+                                 'float [-0-9]', # FP constants
+                                 'inttoptr',     # inttoptr pointer types
+                                 'ptrtoint'      # ptrtoint pointer types
+                                 ]))
+    return_code = 0
+    for (sz_line, llc_line) in zip(sz_out.splitlines(), llc_out.splitlines()):
+        if not match.search(llc_line) and sz_line != llc_line:
+            print 'SZ>' + sz_line
+            print 'LL>' + llc_line
+            return_code = 1
+    if return_code == 0: print 'Success'
+    exit(return_code)