#!/usr/bin/env python2.7 """A test case update script. This script is a utility to update LLVM opt or llc test cases with new FileCheck patterns. It can either update all of the tests in the file or a single test function. """ import argparse import itertools import os # Used to advertise this file's name ("autogenerated_note"). import string import subprocess import sys import tempfile import re # RegEx: this is where the magic happens. SCRUB_LEADING_WHITESPACE_RE = re.compile(r'^(\s+)') SCRUB_WHITESPACE_RE = re.compile(r'(?!^(| \w))[ \t]+', flags=re.M) SCRUB_TRAILING_WHITESPACE_RE = re.compile(r'[ \t]+$', flags=re.M) SCRUB_X86_SHUFFLES_RE = ( re.compile( r'^(\s*\w+) [^#\n]+#+ ((?:[xyz]mm\d+|mem) = .*)$', flags=re.M)) SCRUB_X86_SP_RE = re.compile(r'\d+$%(esp|rsp)$') SCRUB_X86_RIP_RE = re.compile(r'[.\w]+$%rip$') SCRUB_KILL_COMMENT_RE = re.compile(r'^ *#+ +kill:.*\n') SCRUB_IR_COMMENT_RE = re.compile(r'\s*;.*') RUN_LINE_RE = re.compile('^\s*;\s*RUN:\s*(.*)$') IR_FUNCTION_RE = re.compile('^\s*define\s+(?:internal\s+)?[^@]*@([\w-]+)\s*\(') LLC_FUNCTION_RE = re.compile( r'^_?(?P[^:]+):[ \t]*#+[ \t]*@(?P=func)\n[^:]*?' r'(?P^##?[ \t]+[^:]+:.*?)\s*' r'^\s*(?:[^:\n]+?:\s*\n\s*\.size|\.cfi_endproc|\.globl|\.comm|\.(?:sub)?section)', flags=(re.M | re.S)) OPT_FUNCTION_RE = re.compile( r'^\s*define\s+(?:internal\s+)?[^@]*@(?P[\w-]+?)\s*\(' r'(\s+)?[^{]*\{\n(?P.*?)\}', flags=(re.M | re.S)) CHECK_PREFIX_RE = re.compile('--check-prefix=(\S+)') CHECK_RE = re.compile(r'^\s*;\s*([^:]+?)(?:-NEXT|-NOT|-DAG|-LABEL)?:') IR_VALUE_DEF_RE = re.compile(r'\s+%(.*) =') # Invoke the tool that is being tested. def invoke_tool(args, cmd_args, ir): with open(ir) as ir_file: stdout = subprocess.check_output(args.tool_binary + ' ' + cmd_args, shell=True, stdin=ir_file) # Fix line endings to unix CR style. stdout = stdout.replace('\r\n', '\n') return stdout # FIXME: Separate the x86-specific scrubbers, so this can be used for other targets. def scrub_asm(asm): # Detect shuffle asm comments and hide the operands in favor of the comments. asm = SCRUB_X86_SHUFFLES_RE.sub(r'\1 {{.*#+}} \2', asm) # Generically match the stack offset of a memory operand. asm = SCRUB_X86_SP_RE.sub(r'{{[0-9]+}}(%\1)', asm) # Generically match a RIP-relative memory operand. asm = SCRUB_X86_RIP_RE.sub(r'{{.*}}(%rip)', asm) # Strip kill operands inserted into the asm. asm = SCRUB_KILL_COMMENT_RE.sub('', asm) return asm def scrub_body(body, tool_basename): # Scrub runs of whitespace out of the assembly, but leave the leading # whitespace in place. body = SCRUB_WHITESPACE_RE.sub(r' ', body) # Expand the tabs used for indentation. body = string.expandtabs(body, 2) # Strip trailing whitespace. body = SCRUB_TRAILING_WHITESPACE_RE.sub(r'', body) if tool_basename == "llc": body = scrub_asm(body) return body # Build up a dictionary of all the function bodies. def build_function_body_dictionary(raw_tool_output, prefixes, func_dict, verbose, tool_basename): if tool_basename == "llc": func_regex = LLC_FUNCTION_RE else: func_regex = OPT_FUNCTION_RE for m in func_regex.finditer(raw_tool_output): if not m: continue func = m.group('func') scrubbed_body = scrub_body(m.group('body'), tool_basename) if func.startswith('stress'): # We only use the last line of the function body for stress tests. scrubbed_body = '\n'.join(scrubbed_body.splitlines()[-1:]) if verbose: print >>sys.stderr, 'Processing function: ' + func for l in scrubbed_body.splitlines(): print >>sys.stderr, ' ' + l for prefix in prefixes: if func in func_dict[prefix] and func_dict[prefix][func] != scrubbed_body: if prefix == prefixes[-1]: print >>sys.stderr, ('WARNING: Found conflicting asm under the ' 'same prefix: %r!' % (prefix,)) else: func_dict[prefix][func] = None continue func_dict[prefix][func] = scrubbed_body # Create a FileCheck variable name based on an IR name. def get_value_name(var): if var.isdigit(): var = 'TMP' + var var = var.replace('.', '_') return var.upper() # Create a FileCheck variable from regex. def get_value_definition(var): return '[[' + get_value_name(var) + ':%.*]]' # Use a FileCheck variable. def get_value_use(var): return '[[' + get_value_name(var) + ']]' # Replace IR value defs and uses with FileCheck variables. def genericize_check_lines(lines): lines_with_def = [] vars_seen = [] for line in lines: # An IR variable named '%.' matches the FileCheck regex string. line = line.replace('%.', '%dot') m = IR_VALUE_DEF_RE.match(line) if m: vars_seen.append(m.group(1)) line = line.replace('%' + m.group(1), get_value_definition(m.group(1))) lines_with_def.append(line) # A single def isn't worth replacing? #if len(vars_seen) < 2: # return lines output_lines = [] vars_seen.sort(key=len, reverse=True) for line in lines_with_def: for var in vars_seen: line = line.replace('%' + var, get_value_use(var)) output_lines.append(line) return output_lines def add_checks(output_lines, prefix_list, func_dict, func_name, tool_basename): # Select a label format based on the whether we're checking asm or IR. if tool_basename == "llc": check_label_format = "; %s-LABEL: %s:" else: check_label_format = "; %s-LABEL: @%s(" printed_prefixes = [] for checkprefixes, _ in prefix_list: for checkprefix in checkprefixes: if checkprefix in printed_prefixes: break if not func_dict[checkprefix][func_name]: continue # Add some space between different check prefixes, but not after the last # check line (before the test code). #if len(printed_prefixes) != 0: # output_lines.append(';') printed_prefixes.append(checkprefix) output_lines.append(check_label_format % (checkprefix, func_name)) func_body = func_dict[checkprefix][func_name].splitlines() # For IR output, change all defs to FileCheck variables, so we're immune # to variable naming fashions. if tool_basename == "opt": func_body = genericize_check_lines(func_body) # Handle the first line of the function body as a special case because # it's often just noise (a useless asm comment or entry label). if func_body[0].startswith("#") or func_body[0].startswith("entry:"): is_blank_line = True else: output_lines.append('; %s: %s' % (checkprefix, func_body[0])) is_blank_line = False for func_line in func_body[1:]: if func_line.strip() == '': is_blank_line = True continue # Do not waste time checking IR comments. if tool_basename == "opt": func_line = SCRUB_IR_COMMENT_RE.sub(r'', func_line) # Skip blank lines instead of checking them. if is_blank_line == True: output_lines.append('; %s: %s' % (checkprefix, func_line)) else: output_lines.append('; %s-NEXT: %s' % (checkprefix, func_line)) is_blank_line = False # Add space between different check prefixes and also before the first # line of code in the test function. output_lines.append(';') break return output_lines def should_add_line_to_output(input_line, prefix_set): # Skip any blank comment lines in the IR. if input_line.strip() == ';': return False # Skip any blank lines in the IR. #if input_line.strip() == '': # return False # And skip any CHECK lines. We're building our own. m = CHECK_RE.match(input_line) if m and m.group(1) in prefix_set: return False return True def main(): parser = argparse.ArgumentParser(description=__doc__) parser.add_argument('-v', '--verbose', action='store_true', help='Show verbose output') parser.add_argument('--tool-binary', default='llc', help='The tool used to generate the test case') parser.add_argument( '--function', help='The function in the test file to update') parser.add_argument('tests', nargs='+') args = parser.parse_args() autogenerated_note = ('; NOTE: Assertions have been autogenerated by ' + os.path.basename(__file__)) tool_basename = os.path.basename(args.tool_binary) if (tool_basename != "llc" and tool_basename != "opt"): print >>sys.stderr, 'ERROR: Unexpected tool name: ' + tool_basename sys.exit(1) for test in args.tests: if args.verbose: print >>sys.stderr, 'Scanning for RUN lines in test file: %s' % (test,) with open(test) as f: input_lines = [l.rstrip() for l in f] run_lines = [m.group(1) for m in [RUN_LINE_RE.match(l) for l in input_lines] if m] if args.verbose: print >>sys.stderr, 'Found %d RUN lines:' % (len(run_lines),) for l in run_lines: print >>sys.stderr, ' RUN: ' + l prefix_list = [] for l in run_lines: (tool_cmd, filecheck_cmd) = tuple([cmd.strip() for cmd in l.split('|', 1)]) if not tool_cmd.startswith(tool_basename + ' '): print >>sys.stderr, 'WARNING: Skipping non-%s RUN line: %s' % (tool_basename, l) continue if not filecheck_cmd.startswith('FileCheck '): print >>sys.stderr, 'WARNING: Skipping non-FileChecked RUN line: ' + l continue tool_cmd_args = tool_cmd[len(tool_basename):].strip() tool_cmd_args = tool_cmd_args.replace('< %s', '').replace('%s', '').strip() check_prefixes = [m.group(1) for m in CHECK_PREFIX_RE.finditer(filecheck_cmd)] if not check_prefixes: check_prefixes = ['CHECK'] # FIXME: We should use multiple check prefixes to common check lines. For # now, we just ignore all but the last. prefix_list.append((check_prefixes, tool_cmd_args)) func_dict = {} for prefixes, _ in prefix_list: for prefix in prefixes: func_dict.update({prefix: dict()}) for prefixes, tool_args in prefix_list: if args.verbose: print >>sys.stderr, 'Extracted tool cmd: ' + tool_basename + ' ' + tool_args print >>sys.stderr, 'Extracted FileCheck prefixes: ' + str(prefixes) raw_tool_output = invoke_tool(args, tool_args, test) build_function_body_dictionary(raw_tool_output, prefixes, func_dict, args.verbose, tool_basename) is_in_function = False is_in_function_start = False prefix_set = set([prefix for prefixes, _ in prefix_list for prefix in prefixes]) if args.verbose: print >>sys.stderr, 'Rewriting FileCheck prefixes: %s' % (prefix_set,) output_lines = [] output_lines.append(autogenerated_note) for input_line in input_lines: if is_in_function_start: if input_line == '': continue if input_line.lstrip().startswith(';'): m = CHECK_RE.match(input_line) if not m or m.group(1) not in prefix_set: output_lines.append(input_line) continue # Print out the various check lines here. output_lines = add_checks(output_lines, prefix_list, func_dict, name, tool_basename) is_in_function_start = False if is_in_function: if should_add_line_to_output(input_line, prefix_set) == True: # This input line of the function body will go as-is into the output. # Except make leading whitespace uniform: 2 spaces. input_line = SCRUB_LEADING_WHITESPACE_RE.sub(r' ', input_line) output_lines.append(input_line) else: continue if input_line.strip() == '}': is_in_function = False continue if input_line == autogenerated_note: continue # If it's outside a function, it just gets copied to the output. output_lines.append(input_line) m = IR_FUNCTION_RE.match(input_line) if not m: continue name = m.group(1) if args.function is not None and name != args.function: # When filtering on a specific function, skip all others. continue is_in_function = is_in_function_start = True if args.verbose: print>>sys.stderr, 'Writing %d lines to %s...' % (len(output_lines), test) with open(test, 'wb') as f: f.writelines([l + '\n' for l in output_lines]) if __name__ == '__main__': main()