src/soc/decoder/pseudo/pagereader.py

   1 # Reads OpenPOWER ISA pages from http://libre-riscv.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     return os.path.join(fdir, "libreriscv", "openpower", "isa")
  62
  63
  64 class ISA:
  65
  66     def __init__(self):
  67         self.instr = OrderedDict()
  68         self.forms = {}
  69         self.page = {}
  70         for pth in os.listdir(os.path.join(get_isa_dir())):
  71             print(get_isa_dir(), pth)
  72             if "swp" in pth:
  73                 continue
  74             assert pth.endswith(".mdwn"), "only %s in isa dir" % pth
  75             self.read_file(pth)
  76             continue
  77             # code which helped add in the keyword "Pseudo-code:" automatically
  78             rewrite = self.read_file_for_rewrite(pth)
  79             name = os.path.join("/tmp", pth)
  80             with open(name, "w") as f:
  81                 f.write('\n'.join(rewrite) + '\n')
  82
  83     def read_file_for_rewrite(self, fname):
  84         pagename = fname.split('.')[0]
  85         fname = os.path.join(get_isa_dir(), fname)
  86         with open(fname) as f:
  87             lines = f.readlines()
  88         rewrite = []
  89
  90         l = lines.pop(0).rstrip()  # get first line
  91         rewrite.append(l)
  92         while lines:
  93             print(l)
  94             # look for HTML comment, if starting, skip line.
  95             # XXX this is braindead!  it doesn't look for the end
  96             # so please put ending of comments on one line:
  97             # <!-- line 1 comment -->
  98             # <!-- line 2 comment -->
  99             if l.startswith('<!--'):
 100                 print ("skipping comment", l)
 101                 continue
 102             # expect get heading
 103             assert l.startswith('#'), ("# not found in line %s" % l)
 104
 105             # whitespace expected
 106             l = lines.pop(0).strip()
 107             print(repr(l))
 108             assert len(l) == 0, ("blank line not found %s" % l)
 109             rewrite.append(l)
 110
 111             # Form expected
 112             l = lines.pop(0).strip()
 113             assert l.endswith('-Form'), ("line with -Form expected %s" % l)
 114             rewrite.append(l)
 115
 116             # whitespace expected
 117             l = lines.pop(0).strip()
 118             assert len(l) == 0, ("blank line not found %s" % l)
 119             rewrite.append(l)
 120
 121             # get list of opcodes
 122             while True:
 123                 l = lines.pop(0).strip()
 124                 rewrite.append(l)
 125                 if len(l) == 0:
 126                     break
 127                 assert l.startswith('*'), ("* not found in line %s" % l)
 128
 129             rewrite.append("Pseudo-code:")
 130             rewrite.append("")
 131             # get pseudocode
 132             while True:
 133                 l = lines.pop(0).rstrip()
 134                 rewrite.append(l)
 135                 if len(l) == 0:
 136                     break
 137                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 138
 139             # "Special Registers Altered" expected
 140             l = lines.pop(0).rstrip()
 141             assert l.startswith("Special"), ("special not found %s" % l)
 142             rewrite.append(l)
 143
 144             # whitespace expected
 145             l = lines.pop(0).strip()
 146             assert len(l) == 0, ("blank line not found %s" % l)
 147             rewrite.append(l)
 148
 149             # get special regs
 150             while lines:
 151                 l = lines.pop(0).rstrip()
 152                 rewrite.append(l)
 153                 if len(l) == 0:
 154                     break
 155                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 156
 157             # expect and drop whitespace
 158             while lines:
 159                 l = lines.pop(0).rstrip()
 160                 rewrite.append(l)
 161                 if len(l) != 0:
 162                     break
 163
 164         return rewrite
 165
 166     def read_file(self, fname):
 167         pagename = fname.split('.')[0]
 168         fname = os.path.join(get_isa_dir(), fname)
 169         with open(fname) as f:
 170             lines = f.readlines()
 171
 172         # set up dict with current page name
 173         d = {'page': pagename}
 174
 175         # line-by-line lexer/parser, quite straightforward: pops one
 176         # line off the list and checks it.  nothing complicated needed,
 177         # all sections are mandatory so no need for a full LALR parser.
 178
 179         l = lines.pop(0).rstrip()  # get first line
 180         while lines:
 181             print(l)
 182             # expect get heading
 183             assert l.startswith('#'), ("# not found in line %s" % l)
 184             d['desc'] = l[1:].strip()
 185
 186             # whitespace expected
 187             l = lines.pop(0).strip()
 188             print(repr(l))
 189             assert len(l) == 0, ("blank line not found %s" % l)
 190
 191             # Form expected
 192             l = lines.pop(0).strip()
 193             assert l.endswith('-Form'), ("line with -Form expected %s" % l)
 194             d['form'] = l.split('-')[0]
 195
 196             # whitespace expected
 197             l = lines.pop(0).strip()
 198             assert len(l) == 0, ("blank line not found %s" % l)
 199
 200             # get list of opcodes
 201             li = []
 202             while True:
 203                 l = lines.pop(0).strip()
 204                 if len(l) == 0:
 205                     break
 206                 assert l.startswith('*'), ("* not found in line %s" % l)
 207                 l = l[1:].split(' ')  # lose star
 208                 l = filter(lambda x: len(x) != 0, l)  # strip blanks
 209                 li.append(list(l))
 210             opcodes = li
 211
 212             # "Pseudocode" expected
 213             l = lines.pop(0).rstrip()
 214             assert l.startswith("Pseudo-code:"), ("pseudocode found %s" % l)
 215
 216             # whitespace expected
 217             l = lines.pop(0).strip()
 218             print(repr(l))
 219             assert len(l) == 0, ("blank line not found %s" % l)
 220
 221             # get pseudocode
 222             li = []
 223             while True:
 224                 l = lines.pop(0).rstrip()
 225                 if len(l) == 0:
 226                     break
 227                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 228                 l = l[4:]  # lose 4 spaces
 229                 li.append(l)
 230             d['pcode'] = li
 231
 232             # "Special Registers Altered" expected
 233             l = lines.pop(0).rstrip()
 234             assert l.startswith("Special"), ("special not found %s" % l)
 235
 236             # whitespace expected
 237             l = lines.pop(0).strip()
 238             assert len(l) == 0, ("blank line not found %s" % l)
 239
 240             # get special regs
 241             li = []
 242             while lines:
 243                 l = lines.pop(0).rstrip()
 244                 if len(l) == 0:
 245                     break
 246                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 247                 l = l[4:]  # lose 4 spaces
 248                 li.append(l)
 249             d['sregs'] = li
 250
 251             # add in opcode
 252             for o in opcodes:
 253                 self.add_op(o, d)
 254
 255             # expect and drop whitespace
 256             while lines:
 257                 l = lines.pop(0).rstrip()
 258                 if len(l) != 0:
 259                     break
 260
 261     def add_op(self, o, d):
 262         opcode, regs = o[0], o[1:]
 263         op = copy(d)
 264         op['regs'] = regs
 265         if len(regs) != 0:
 266             regs[0] = regs[0].split(",")
 267         op['opcode'] = opcode
 268         self.instr[opcode] = Ops(**op)
 269
 270         # create list of instructions by form
 271         form = op['form']
 272         fl = self.forms.get(form, [])
 273         self.forms[form] = fl + [opcode]
 274
 275         # create list of instructions by page
 276         page = op['page']
 277         pl = self.page.get(page, [])
 278         self.page[page] = pl + [opcode]
 279
 280     def pprint_ops(self):
 281         for k, v in self.instr.items():
 282             print("# %s %s" % (v.opcode, v.desc))
 283             print("Form: %s Regs: %s" % (v.form, v.regs))
 284             print('\n'.join(map(lambda x: "    %s" % x, v.pcode)))
 285             print("Specials")
 286             print('\n'.join(map(lambda x: "    %s" % x, v.sregs)))
 287             print()
 288         for k, v in isa.forms.items():
 289             print(k, v)
 290
 291
 292 if __name__ == '__main__':
 293     isa = ISA()
 294     isa.pprint_ops()
 295     # example on how to access cmp regs:
 296     print ("cmp regs:", isa.instr["cmp"].regs)