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                 l = lines.pop(0).rstrip()  # get first line
 102                 continue
 103
 104             # Ignore blank lines before the first #
 105             if len(l.strip()) == 0:
 106                 continue
 107
 108             # expect get heading
 109             assert l.startswith('#'), ("# not found in line %s" % l)
 110
 111             # whitespace expected
 112             l = lines.pop(0).strip()
 113             print(repr(l))
 114             assert len(l) == 0, ("blank line not found %s" % l)
 115             rewrite.append(l)
 116
 117             # Form expected
 118             l = lines.pop(0).strip()
 119             assert l.endswith('-Form'), ("line with -Form expected %s" % l)
 120             rewrite.append(l)
 121
 122             # whitespace expected
 123             l = lines.pop(0).strip()
 124             assert len(l) == 0, ("blank line not found %s" % l)
 125             rewrite.append(l)
 126
 127             # get list of opcodes
 128             while True:
 129                 l = lines.pop(0).strip()
 130                 rewrite.append(l)
 131                 if len(l) == 0:
 132                     break
 133                 assert l.startswith('*'), ("* not found in line %s" % l)
 134
 135             rewrite.append("Pseudo-code:")
 136             rewrite.append("")
 137             # get pseudocode
 138             while True:
 139                 l = lines.pop(0).rstrip()
 140                 rewrite.append(l)
 141                 if len(l) == 0:
 142                     break
 143                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 144
 145             # "Special Registers Altered" expected
 146             l = lines.pop(0).rstrip()
 147             assert l.startswith("Special"), ("special not found %s" % l)
 148             rewrite.append(l)
 149
 150             # whitespace expected
 151             l = lines.pop(0).strip()
 152             assert len(l) == 0, ("blank line not found %s" % l)
 153             rewrite.append(l)
 154
 155             # get special regs
 156             while lines:
 157                 l = lines.pop(0).rstrip()
 158                 rewrite.append(l)
 159                 if len(l) == 0:
 160                     break
 161                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 162
 163             # expect and drop whitespace
 164             while lines:
 165                 l = lines.pop(0).rstrip()
 166                 rewrite.append(l)
 167                 if len(l) != 0 and not l.startswith('<!--'):
 168                     break
 169
 170         return rewrite
 171
 172     def read_file(self, fname):
 173         pagename = fname.split('.')[0]
 174         fname = os.path.join(get_isa_dir(), fname)
 175         with open(fname) as f:
 176             lines = f.readlines()
 177
 178         # set up dict with current page name
 179         d = {'page': pagename}
 180
 181         # line-by-line lexer/parser, quite straightforward: pops one
 182         # line off the list and checks it.  nothing complicated needed,
 183         # all sections are mandatory so no need for a full LALR parser.
 184
 185         l = lines.pop(0).rstrip()  # get first line
 186         while lines:
 187             print(l)
 188             # look for HTML comment, if starting, skip line.
 189             # XXX this is braindead!  it doesn't look for the end
 190             # so please put ending of comments on one line:
 191             # <!-- line 1 comment -->
 192             # <!-- line 2 comment -->
 193             if l.startswith('<!--'):
 194                 # print ("skipping comment", l)
 195                 l = lines.pop(0).rstrip()  # get next line
 196                 continue
 197
 198             # Ignore blank lines before the first #
 199             if len(l) == 0:
 200                 l = lines.pop(0).rstrip()  # get next line
 201                 continue
 202
 203             # expect get heading
 204             assert l.startswith('#'), ("# not found in line '%s'" % l)
 205             d['desc'] = l[1:].strip()
 206
 207             # whitespace expected
 208             l = lines.pop(0).strip()
 209             print(repr(l))
 210             assert len(l) == 0, ("blank line not found %s" % l)
 211
 212             # Form expected
 213             l = lines.pop(0).strip()
 214             assert l.endswith('-Form'), ("line with -Form expected %s" % l)
 215             d['form'] = l.split('-')[0]
 216
 217             # whitespace expected
 218             l = lines.pop(0).strip()
 219             assert len(l) == 0, ("blank line not found %s" % l)
 220
 221             # get list of opcodes
 222             li = []
 223             while True:
 224                 l = lines.pop(0).strip()
 225                 if len(l) == 0:
 226                     break
 227                 assert l.startswith('*'), ("* not found in line %s" % l)
 228                 l = l[1:].split(' ')  # lose star
 229                 l = filter(lambda x: len(x) != 0, l)  # strip blanks
 230                 li.append(list(l))
 231             opcodes = li
 232
 233             # "Pseudocode" expected
 234             l = lines.pop(0).rstrip()
 235             assert l.startswith("Pseudo-code:"), ("pseudocode found %s" % l)
 236
 237             # whitespace expected
 238             l = lines.pop(0).strip()
 239             print(repr(l))
 240             assert len(l) == 0, ("blank line not found %s" % l)
 241
 242             # get pseudocode
 243             li = []
 244             while True:
 245                 l = lines.pop(0).rstrip()
 246                 if len(l) == 0:
 247                     break
 248                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 249                 l = l[4:]  # lose 4 spaces
 250                 li.append(l)
 251             d['pcode'] = li
 252
 253             # "Special Registers Altered" expected
 254             l = lines.pop(0).rstrip()
 255             assert l.startswith("Special"), ("special not found %s" % l)
 256
 257             # whitespace expected
 258             l = lines.pop(0).strip()
 259             assert len(l) == 0, ("blank line not found %s" % l)
 260
 261             # get special regs
 262             li = []
 263             while lines:
 264                 l = lines.pop(0).rstrip()
 265                 if len(l) == 0:
 266                     break
 267                 assert l.startswith('    '), ("4spcs not found in line %s" % l)
 268                 l = l[4:]  # lose 4 spaces
 269                 li.append(l)
 270             d['sregs'] = li
 271
 272             # add in opcode
 273             for o in opcodes:
 274                 self.add_op(o, d)
 275
 276             # expect and drop whitespace
 277             while lines:
 278                 l = lines.pop(0).rstrip()
 279                 if len(l) != 0 and not l.startswith('<!--'):
 280                     break
 281
 282     def add_op(self, o, d):
 283         opcode, regs = o[0], o[1:]
 284         op = copy(d)
 285         op['regs'] = regs
 286         if len(regs) != 0:
 287             regs[0] = regs[0].split(",")
 288         op['opcode'] = opcode
 289         self.instr[opcode] = Ops(**op)
 290
 291         # create list of instructions by form
 292         form = op['form']
 293         fl = self.forms.get(form, [])
 294         self.forms[form] = fl + [opcode]
 295
 296         # create list of instructions by page
 297         page = op['page']
 298         pl = self.page.get(page, [])
 299         self.page[page] = pl + [opcode]
 300
 301     def pprint_ops(self):
 302         for k, v in self.instr.items():
 303             print("# %s %s" % (v.opcode, v.desc))
 304             print("Form: %s Regs: %s" % (v.form, v.regs))
 305             print('\n'.join(map(lambda x: "    %s" % x, v.pcode)))
 306             print("Specials")
 307             print('\n'.join(map(lambda x: "    %s" % x, v.sregs)))
 308             print()
 309         for k, v in isa.forms.items():
 310             print(k, v)
 311
 312
 313 if __name__ == '__main__':
 314     isa = ISA()
 315     isa.pprint_ops()
 316     # example on how to access cmp regs:
 317     print ("cmp regs:", isa.instr["cmp"].regs)