isa_parser.py

来自「M5,一个功能强大的多处理器系统模拟器.很多针对处理器架构,性能的研究都使用它作」· Python 代码 · 共 1,867 行 · 第 1/5 页

# Copyright (c) 2003, 2004, 2005
# The Regents of The University of Michigan
# All Rights Reserved
#
# This code is part of the M5 simulator.
#
# Permission is granted to use, copy, create derivative works and
# redistribute this software and such derivative works for any
# purpose, so long as the copyright notice above, this grant of
# permission, and the disclaimer below appear in all copies made; and
# so long as the name of The University of Michigan is not used in any
# advertising or publicity pertaining to the use or distribution of
# this software without specific, written prior authorization.
#
# THIS SOFTWARE IS PROVIDED AS IS, WITHOUT REPRESENTATION FROM THE
# UNIVERSITY OF MICHIGAN AS TO ITS FITNESS FOR ANY PURPOSE, AND
# WITHOUT WARRANTY BY THE UNIVERSITY OF MICHIGAN OF ANY KIND, EITHER
# EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION THE IMPLIED
# WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
# PURPOSE. THE REGENTS OF THE UNIVERSITY OF MICHIGAN SHALL NOT BE
# LIABLE FOR ANY DAMAGES, INCLUDING DIRECT, SPECIAL, INDIRECT,
# INCIDENTAL, OR CONSEQUENTIAL DAMAGES, WITH RESPECT TO ANY CLAIM
# ARISING OUT OF OR IN CONNECTION WITH THE USE OF THE SOFTWARE, EVEN
# IF IT HAS BEEN OR IS HEREAFTER ADVISED OF THE POSSIBILITY OF SUCH
# DAMAGES.
#
# Authors: Steven K. Reinhardt
#          Korey Sewell

import os
import sys
import re
import string
import traceback
# get type names
from types import *

# Prepend the directory where the PLY lex & yacc modules are found
# to the search path.  Assumes we're compiling in a subdirectory
# of 'build' in the current tree.
sys.path[0:0] = [os.environ['M5_PLY']]

from ply import lex
from ply import yacc

#####################################################################
#
#                                Lexer
#
# The PLY lexer module takes two things as input:
# - A list of token names (the string list 'tokens')
# - A regular expression describing a match for each token.  The
#   regexp for token FOO can be provided in two ways:
#   - as a string variable named t_FOO
#   - as the doc string for a function named t_FOO.  In this case,
#     the function is also executed, allowing an action to be
#     associated with each token match.
#
#####################################################################

# Reserved words.  These are listed separately as they are matched
# using the same regexp as generic IDs, but distinguished in the
# t_ID() function.  The PLY documentation suggests this approach.
reserved = (
    'BITFIELD', 'DECODE', 'DECODER', 'DEFAULT', 'DEF', 'EXEC', 'FORMAT',
    'HEADER', 'LET', 'NAMESPACE', 'OPERAND_TYPES', 'OPERANDS',
    'OUTPUT', 'SIGNED', 'TEMPLATE'
    )

# List of tokens.  The lex module requires this.
tokens = reserved + (
    # identifier
    'ID',

    # integer literal
    'INTLIT',

    # string literal
    'STRLIT',

    # code literal
    'CODELIT',

    # ( ) [ ] { } < > , ; . : :: *
    'LPAREN', 'RPAREN',
    'LBRACKET', 'RBRACKET',
    'LBRACE', 'RBRACE',
    'LESS', 'GREATER', 'EQUALS',
    'COMMA', 'SEMI', 'DOT', 'COLON', 'DBLCOLON',
    'ASTERISK',

    # C preprocessor directives
    'CPPDIRECTIVE'

# The following are matched but never returned. commented out to
# suppress PLY warning
    # newfile directive
#    'NEWFILE',

    # endfile directive
#    'ENDFILE'
)

# Regular expressions for token matching
t_LPAREN           = r'\('
t_RPAREN           = r'\)'
t_LBRACKET         = r'\['
t_RBRACKET         = r'\]'
t_LBRACE           = r'\{'
t_RBRACE           = r'\}'
t_LESS             = r'\<'
t_GREATER          = r'\>'
t_EQUALS           = r'='
t_COMMA            = r','
t_SEMI             = r';'
t_DOT              = r'\.'
t_COLON            = r':'
t_DBLCOLON         = r'::'
t_ASTERISK	   = r'\*'

# Identifiers and reserved words
reserved_map = { }
for r in reserved:
    reserved_map[r.lower()] = r

def t_ID(t):
    r'[A-Za-z_]\w*'
    t.type = reserved_map.get(t.value,'ID')
    return t

# Integer literal
def t_INTLIT(t):
    r'(0x[\da-fA-F]+)|\d+'
    try:
        t.value = int(t.value,0)
    except ValueError:
        error(t.lexer.lineno, 'Integer value "%s" too large' % t.value)
        t.value = 0
    return t

# String literal.  Note that these use only single quotes, and
# can span multiple lines.
def t_STRLIT(t):
    r"(?m)'([^'])+'"
    # strip off quotes
    t.value = t.value[1:-1]
    t.lexer.lineno += t.value.count('\n')
    return t


# "Code literal"... like a string literal, but delimiters are
# '{{' and '}}' so they get formatted nicely under emacs c-mode
def t_CODELIT(t):
    r"(?m)\{\{([^\}]|}(?!\}))+\}\}"
    # strip off {{ & }}
    t.value = t.value[2:-2]
    t.lexer.lineno += t.value.count('\n')
    return t

def t_CPPDIRECTIVE(t):
    r'^\#[^\#].*\n'
    t.lexer.lineno += t.value.count('\n')
    return t

def t_NEWFILE(t):
    r'^\#\#newfile\s+"[\w/.-]*"'
    fileNameStack.push((t.value[11:-1], t.lexer.lineno))
    t.lexer.lineno = 0

def t_ENDFILE(t):
    r'^\#\#endfile'
    (old_filename, t.lexer.lineno) = fileNameStack.pop()

#
# The functions t_NEWLINE, t_ignore, and t_error are
# special for the lex module.
#

# Newlines
def t_NEWLINE(t):
    r'\n+'
    t.lexer.lineno += t.value.count('\n')

# Comments
def t_comment(t):
    r'//.*'

# Completely ignored characters
t_ignore           = ' \t\x0c'

# Error handler
def t_error(t):
    error(t.lexer.lineno, "illegal character '%s'" % t.value[0])
    t.skip(1)

# Build the lexer
lexer = lex.lex()

#####################################################################
#
#                                Parser
#
# Every function whose name starts with 'p_' defines a grammar rule.
# The rule is encoded in the function's doc string, while the
# function body provides the action taken when the rule is matched.
# The argument to each function is a list of the values of the
# rule's symbols: t[0] for the LHS, and t[1..n] for the symbols
# on the RHS.  For tokens, the value is copied from the t.value
# attribute provided by the lexer.  For non-terminals, the value
# is assigned by the producing rule; i.e., the job of the grammar
# rule function is to set the value for the non-terminal on the LHS
# (by assigning to t[0]).
#####################################################################

# The LHS of the first grammar rule is used as the start symbol
# (in this case, 'specification').  Note that this rule enforces
# that there will be exactly one namespace declaration, with 0 or more
# global defs/decls before and after it.  The defs & decls before
# the namespace decl will be outside the namespace; those after
# will be inside.  The decoder function is always inside the namespace.
def p_specification(t):
    'specification : opt_defs_and_outputs name_decl opt_defs_and_outputs decode_block'
    global_code = t[1]
    isa_name = t[2]
    namespace = isa_name + "Inst"
    # wrap the decode block as a function definition
    t[4].wrap_decode_block('''
StaticInstPtr
%(isa_name)s::decodeInst(%(isa_name)s::ExtMachInst machInst)
{
    using namespace %(namespace)s;
''' % vars(), '}')
    # both the latter output blocks and the decode block are in the namespace
    namespace_code = t[3] + t[4]
    # pass it all back to the caller of yacc.parse()
    t[0] = (isa_name, namespace, global_code, namespace_code)

# ISA name declaration looks like "namespace <foo>;"
def p_name_decl(t):
    'name_decl : NAMESPACE ID SEMI'
    t[0] = t[2]

# 'opt_defs_and_outputs' is a possibly empty sequence of
# def and/or output statements.
def p_opt_defs_and_outputs_0(t):
    'opt_defs_and_outputs : empty'
    t[0] = GenCode()

def p_opt_defs_and_outputs_1(t):
    'opt_defs_and_outputs : defs_and_outputs'
    t[0] = t[1]

def p_defs_and_outputs_0(t):
    'defs_and_outputs : def_or_output'
    t[0] = t[1]

def p_defs_and_outputs_1(t):
    'defs_and_outputs : defs_and_outputs def_or_output'
    t[0] = t[1] + t[2]

# The list of possible definition/output statements.
def p_def_or_output(t):
    '''def_or_output : def_format
                     | def_bitfield
                     | def_bitfield_struct
                     | def_template
                     | def_operand_types
                     | def_operands
                     | output_header
                     | output_decoder
                     | output_exec
                     | global_let'''
    t[0] = t[1]

# Output blocks 'output <foo> {{...}}' (C++ code blocks) are copied
# directly to the appropriate output section.


# Protect any non-dict-substitution '%'s in a format string
# (i.e. those not followed by '(')
def protect_non_subst_percents(s):
    return re.sub(r'%(?!\()', '%%', s)

# Massage output block by substituting in template definitions and bit
# operators.  We handle '%'s embedded in the string that don't
# indicate template substitutions (or CPU-specific symbols, which get
# handled in GenCode) by doubling them first so that the format
# operation will reduce them back to single '%'s.
def process_output(s):
    s = protect_non_subst_percents(s)
    # protects cpu-specific symbols too
    s = protect_cpu_symbols(s)
    return substBitOps(s % templateMap)

def p_output_header(t):
    'output_header : OUTPUT HEADER CODELIT SEMI'
    t[0] = GenCode(header_output = process_output(t[3]))

def p_output_decoder(t):
    'output_decoder : OUTPUT DECODER CODELIT SEMI'
    t[0] = GenCode(decoder_output = process_output(t[3]))

def p_output_exec(t):
    'output_exec : OUTPUT EXEC CODELIT SEMI'
    t[0] = GenCode(exec_output = process_output(t[3]))

# global let blocks 'let {{...}}' (Python code blocks) are executed
# directly when seen.  Note that these execute in a special variable
# context 'exportContext' to prevent the code from polluting this
# script's namespace.
def p_global_let(t):
    'global_let : LET CODELIT SEMI'
    updateExportContext()
    exportContext["header_output"] = ''
    exportContext["decoder_output"] = ''
    exportContext["exec_output"] = ''
    exportContext["decode_block"] = ''
    try:
        exec fixPythonIndentation(t[2]) in exportContext
    except Exception, exc:
        error(t.lexer.lineno,
              'error: %s in global let block "%s".' % (exc, t[2]))
    t[0] = GenCode(header_output = exportContext["header_output"],
                   decoder_output = exportContext["decoder_output"],
                   exec_output = exportContext["exec_output"],
                   decode_block = exportContext["decode_block"])

# Define the mapping from operand type extensions to C++ types and bit
# widths (stored in operandTypeMap).
def p_def_operand_types(t):
    'def_operand_types : DEF OPERAND_TYPES CODELIT SEMI'
    try:
        userDict = eval('{' + t[3] + '}')
    except Exception, exc:
        error(t.lexer.lineno,
              'error: %s in def operand_types block "%s".' % (exc, t[3]))
    buildOperandTypeMap(userDict, t.lexer.lineno)
    t[0] = GenCode() # contributes nothing to the output C++ file

# Define the mapping from operand names to operand classes and other
# traits.  Stored in operandNameMap.
def p_def_operands(t):
    'def_operands : DEF OPERANDS CODELIT SEMI'
    if not globals().has_key('operandTypeMap'):
        error(t.lexer.lineno,
              'error: operand types must be defined before operands')
    try:
        userDict = eval('{' + t[3] + '}')
    except Exception, exc:
        error(t.lexer.lineno,
              'error: %s in def operands block "%s".' % (exc, t[3]))
    buildOperandNameMap(userDict, t.lexer.lineno)
    t[0] = GenCode() # contributes nothing to the output C++ file

# A bitfield definition looks like:
# 'def [signed] bitfield <ID> [<first>:<last>]'
# This generates a preprocessor macro in the output file.
def p_def_bitfield_0(t):
    'def_bitfield : DEF opt_signed BITFIELD ID LESS INTLIT COLON INTLIT GREATER SEMI'
    expr = 'bits(machInst, %2d, %2d)' % (t[6], t[8])
    if (t[2] == 'signed'):
        expr = 'sext<%d>(%s)' % (t[6] - t[8] + 1, expr)
    hash_define = '#undef %s\n#define %s\t%s\n' % (t[4], t[4], expr)
    t[0] = GenCode(header_output = hash_define)

# alternate form for single bit: 'def [signed] bitfield <ID> [<bit>]'
def p_def_bitfield_1(t):
    'def_bitfield : DEF opt_signed BITFIELD ID LESS INTLIT GREATER SEMI'
    expr = 'bits(machInst, %2d, %2d)' % (t[6], t[6])
    if (t[2] == 'signed'):
        expr = 'sext<%d>(%s)' % (1, expr)
    hash_define = '#undef %s\n#define %s\t%s\n' % (t[4], t[4], expr)
    t[0] = GenCode(header_output = hash_define)