src/openpower/decoder/pseudo/pagereader.py

   1 # Reads OpenPOWER ISA pages from http://libre-soc.org/openpower/isa
   2 """OpenPOWER ISA page parser
   3
   4 returns an OrderedDict of namedtuple "Ops" containing details of all
   5 instructions listed in markdown files.
   6
   7 format must be strictly as follows (no optional sections) including whitespace:
   8
   9 # Compare Logical
  10
  11 X-Form
  12
  13 * cmpl BF,L,RA,RB
  14
  15     if L = 0 then a <- [0]*32 || (RA)[32:63]
  16                   b <- [0]*32 || (RB)[32:63]
  17              else a <-  (RA)
  18                   b <-  (RB)
  19     if      a <u b then c <- 0b100
  20     else if a >u b then c <- 0b010
  21     else                c <-  0b001
  22     CR[4*BF+32:4*BF+35] <- c || XER[SO]
  23
  24 Special Registers Altered:
  25
  26     CR field BF
  27     Another field
  28
  29 this translates to:
  30
  31     # heading
  32     blank
  33     Some-Form
  34     blank
  35     * instruction registerlist
  36     * instruction registerlist
  37     blank
  38     4-space-indented pseudo-code
  39     4-space-indented pseudo-code
  40     blank
  41     Special Registers Altered:
  42     4-space-indented register description
  43     blank
  44     blank(s) (optional for convenience at end-of-page)
  45 """
  46
  47 from collections import namedtuple, OrderedDict
  48 from copy import copy
  49 import os
  50
  51 opfields = ("desc", "form", "opcode", "regs", "pcode", "sregs", "page")
  52 Ops = namedtuple("Ops", opfields)
  53
  54
  55 def get_isa_dir():
  56     fdir = os.path.abspath(os.path.dirname(__file__))
  57     fdir = os.path.split(fdir)[0]
  58     fdir = os.path.split(fdir)[0]
  59     fdir = os.path.split(fdir)[0]
  60     fdir = os.path.split(fdir)[0]
  61     # print (fdir)
  62     return os.path.join(fdir, "openpower", "isa")
  63
  64
  65 class ISA:
  66
  67     def __init__(self):
  68         self.instr = OrderedDict()
  69         self.forms = {}
  70         self.page = {}
  71         self.verbose = False
  72         for pth in os.listdir(os.path.join(get_isa_dir())):
  73             if self.verbose:
  74                 print("examining", get_isa_dir(), pth)
  75             if "swp" in pth:
  76                 continue
  77             if not pth.endswith(".mdwn"):
  78                 print ("warning, file not .mdwn, skipping", pth)
  79                 continue
  80             self.read_file(pth)
  81             continue
  82             # code which helped add in the keyword "Pseudo-code:" automatically
  83             rewrite = self.read_file_for_rewrite(pth)
  84             name = os.path.join("/tmp", pth)
  85             with open(name, "w") as f:
  86                 f.write('\n'.join(rewrite) + '\n')
  87
  88     def read_file_for_rewrite(self, fname):
  89         pagename = fname.split('.')[0]
  90         fname = os.path.join(get_isa_dir(), fname)
  91         with open(fname) as f:
  92             lines = f.readlines()
  93         rewrite = []
  94
  95         l = lines.pop(0).rstrip()  # get first line
  96         rewrite.append(l)
  97         while lines:
  98             if self.verbose:
  99                 print(l)
 100             # look for HTML comment, if starting, skip line.
 101             # XXX this is braindead!  it doesn't look for the end
 102             # so please put ending of comments on one line:
 103             # <!-- line 1 comment -->
 104             # <!-- line 2 comment -->
 105             if l.startswith('<!--'):
 106                 # print ("skipping comment", l)
 107                 l = lines.pop(0).rstrip()  # get first line
 108                 continue
 109
 110             # Ignore blank lines before the first #
 111             if len(l.strip()) == 0:
 112                 continue
 113
 114             # expect get heading
 115             assert l.startswith('#'), ("# not found in line %s" % l)
 116
 117             # whitespace expected
 118             l = lines.pop(0).strip()
 119             if self.verbose:
 120                 print(repr(l))
 121             assert len(l) == 0, ("blank line not found %s" % l)
 122             rewrite.append(l)
 123
 124             # Form expected
 125             l = lines.pop(0).strip()
 126             assert l.endswith('-Form'), ("line with -Form expected %s" % l)
 127             rewrite.append(l)
 128
 129             # whitespace expected
 130             l = lines.pop(0).strip()
 131             assert len(l) == 0, ("blank line not found %s" % l)
 132             rewrite.append(l)
 133
 134             # get list of opcodes
 135             while True:
 136                 l = lines.pop(0).strip()
 137                 rewrite.append(l)
 138                 if len(l) == 0:
 139                     break
 140                 assert l.startswith('*'), ("* not found in line %s" % l)
 141
 142             rewrite.append("Pseudo-code:")
 143             rewrite.append("")
 144             # get pseudocode
 145             while True:
 146                 l = lines.pop(0).rstrip()
 147                 rewrite.append(l)
 148                 if len(l) == 0:
 149                     break
 150                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 151
 152             # "Special Registers Altered" expected
 153             l = lines.pop(0).rstrip()
 154             assert l.startswith("Special"), ("special not found %s" % l)
 155             rewrite.append(l)
 156
 157             # whitespace expected
 158             l = lines.pop(0).strip()
 159             assert len(l) == 0, ("blank line not found %s" % l)
 160             rewrite.append(l)
 161
 162             # get special regs
 163             while lines:
 164                 l = lines.pop(0).rstrip()
 165                 rewrite.append(l)
 166                 if len(l) == 0:
 167                     break
 168                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 169
 170             # expect and drop whitespace
 171             while lines:
 172                 l = lines.pop(0).rstrip()
 173                 rewrite.append(l)
 174                 if len(l) != 0 and not l.startswith('<!--'):
 175                     break
 176
 177         return rewrite
 178
 179     def read_file(self, fname):
 180         pagename = fname.split('.')[0]
 181         fname = os.path.join(get_isa_dir(), fname)
 182         with open(fname) as f:
 183             lines = f.readlines()
 184
 185         # set up dict with current page name
 186         d = {'page': pagename}
 187
 188         # line-by-line lexer/parser, quite straightforward: pops one
 189         # line off the list and checks it.  nothing complicated needed,
 190         # all sections are mandatory so no need for a full LALR parser.
 191
 192         l = lines.pop(0).rstrip()  # get first line
 193         while lines:
 194             if self.verbose:
 195                 print(l)
 196             # look for HTML comment, if starting, skip line.
 197             # XXX this is braindead!  it doesn't look for the end
 198             # so please put ending of comments on one line:
 199             # <!-- line 1 comment -->
 200             # <!-- line 2 comment -->
 201             if l.startswith('<!--'):
 202                 # print ("skipping comment", l)
 203                 l = lines.pop(0).rstrip()  # get next line
 204                 continue
 205
 206             # Ignore blank lines before the first #
 207             if len(l) == 0:
 208                 l = lines.pop(0).rstrip()  # get next line
 209                 continue
 210
 211             # expect get heading
 212             assert l.startswith('#'), ("# not found in line '%s'" % l)
 213             d['desc'] = l[1:].strip()
 214
 215             # whitespace expected
 216             l = lines.pop(0).strip()
 217             if self.verbose:
 218                 print(repr(l))
 219             assert len(l) == 0, ("blank line not found %s" % l)
 220
 221             # Form expected
 222             l = lines.pop(0).strip()
 223             assert l.endswith('-Form'), ("line with -Form expected %s" % l)
 224             d['form'] = l.split('-')[0]
 225
 226             # whitespace expected
 227             l = lines.pop(0).strip()
 228             assert len(l) == 0, ("blank line not found %s" % l)
 229
 230             # get list of opcodes
 231             li = []
 232             while True:
 233                 l = lines.pop(0).strip()
 234                 if len(l) == 0:
 235                     break
 236                 assert l.startswith('*'), ("* not found in line %s" % l)
 237                 l = l[1:].split(' ')  # lose star
 238                 l = filter(lambda x: len(x) != 0, l)  # strip blanks
 239                 li.append(list(l))
 240             opcodes = li
 241
 242             # "Pseudocode" expected
 243             l = lines.pop(0).rstrip()
 244             assert l.startswith("Pseudo-code:"), ("pseudocode found %s" % l)
 245
 246             # whitespace expected
 247             l = lines.pop(0).strip()
 248             if self.verbose:
 249                 print(repr(l))
 250             assert len(l) == 0, ("blank line not found %s" % l)
 251
 252             # get pseudocode
 253             li = []
 254             while True:
 255                 l = lines.pop(0).rstrip()
 256                 if len(l) == 0:
 257                     break
 258                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 259                 l = l[4:]  # lose 4 spaces
 260                 li.append(l)
 261             d['pcode'] = li
 262
 263             # "Special Registers Altered" expected
 264             l = lines.pop(0).rstrip()
 265             assert l.startswith("Special"), ("special not found %s" % l)
 266
 267             # whitespace expected
 268             l = lines.pop(0).strip()
 269             assert len(l) == 0, ("blank line not found %s" % l)
 270
 271             # get special regs
 272             li = []
 273             while lines:
 274                 l = lines.pop(0).rstrip()
 275                 if len(l) == 0:
 276                     break
 277                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 278                 l = l[4:]  # lose 4 spaces
 279                 li.append(l)
 280             d['sregs'] = li
 281
 282             # add in opcode
 283             for o in opcodes:
 284                 self.add_op(o, d)
 285
 286             # expect and drop whitespace
 287             while lines:
 288                 l = lines.pop(0).rstrip()
 289                 if len(l) != 0 and not l.startswith('<!--'):
 290                     break
 291
 292     def add_op(self, o, d):
 293         opcode, regs = o[0], o[1:]
 294         op = copy(d)
 295         op['regs'] = regs
 296         if len(regs) != 0:
 297             regs[0] = regs[0].split(",")
 298         op['opcode'] = opcode
 299         self.instr[opcode] = Ops(**op)
 300
 301         # create list of instructions by form
 302         form = op['form']
 303         fl = self.forms.get(form, [])
 304         self.forms[form] = fl + [opcode]
 305
 306         # create list of instructions by page
 307         page = op['page']
 308         pl = self.page.get(page, [])
 309         self.page[page] = pl + [opcode]
 310
 311     def pprint_ops(self):
 312         for k, v in self.instr.items():
 313             print("# %s %s" % (v.opcode, v.desc))
 314             print("Form: %s Regs: %s" % (v.form, v.regs))
 315             print('\n'.join(map(lambda x: "    %s" % x, v.pcode)))
 316             print("Specials")
 317             print('\n'.join(map(lambda x: "    %s" % x, v.sregs)))
 318             print()
 319         for k, v in isa.forms.items():
 320             print(k, v)
 321
 322
 323 if __name__ == '__main__':
 324     isa = ISA()
 325     isa.pprint_ops()
 326     # example on how to access cmp regs:
 327     print ("cmp regs:", isa.instr["cmp"].regs)