jedi.1

.TH JEDI 1OCTTOOLS "1 March 1990"
.SH NAME
jedi \- A symbolic encoding program
.SH SYNOPSIS
.B jedi  
[options] [input] [ > output]
.SH DESCRIPTION
.PP
.I JEDI
is a general symbolic encoding program intended for multi-level 
logic optimization by a logic synthesis system such as 
.I MISII(1OCTTOOLS).  A symbolic description is different than 
a traditional Boolean logic description in that data types are 
defined in terms of symbolic values rather than simply 0 and 1.
A special symbolic type is the Boolean set (0, 1).  More general 
symbolic types can be defined.  For example, one may wish to define 
a data type with the following values: (ADD, CMP, MUL, SUB).  To convert 
a symbolic description to a form that can be implemented in hardware, 
the symbolic values must be given binary codes (eg. ADD -> 00, CMP -> 01, 
MUL -> 10, SUB -> 11).  Depending on the choice of codes, the cost of 
the final implementation may vary dramatically (may be a factor of 2).  
Multiple symbolic data types and symbolic variables can be defined.  JEDI 
can handle both input and output variables.  The former is treated as an 
input encoding problem while the latter is treated as an output encoding 
problem.  Variables of the same symbolic type are given the same encodings.  
The input format will be explained below.
.PP
JEDI can also be used for state assignment, which is a special case 
of symbolic encoding where the internal states of a finite state 
machine are expressed symbolically (eg. st0, st1, st2, st3).  Here, 
only the state variables are encoded; the other binary variables are 
maintained.  The input format for the state assignment problem is 
basically the same as those used by the state assignment programs 
.I KISS(1OCTTOOLS),
.I MUSTANG(1OCTTOOLS),
and
.I NOVA(1OCTTOOLS).
This is so that these programs can be used interchangeably.
Currently, this is the most commonly used application of JEDI.
Since this is the case, JEDI assumes a state assignment problem
unless otherwise specified.
.PP
If JEDI is used with the general symbolic format, the output is a fully 
encoded logic description in the Berkeley PLA format suitable for further 
optimization by the programs
.I ESPRESSO(1OCTTOOLS),
and 
.I MISII(1OCTTOOLS).
The unused binary codes are automatically inserted as external don't
cares.  Therefore, the output of JEDI should be processed through the 
ESPRESSO program so that these don't care conditions can be properly 
exploited.  Other logic optimization programs can be used if they accept the
same logic interchange formats as the Berkeley logic tools.
.PP
If JEDI is used for state assignment, then it will by default produce
an extended BLIF format suitable for further processing by the
sequential logic synthesis system called SIS.  An option can be
specified to produce the PLA output format.
.SH "OPTIONS"
.IP -\fBh\fP
prints out usage information.
.IP -\fBe option\fP
r) performs random encoding.
h) performs one-hot encoding.
d) performs dynamic random encoding.
s) performs straightforward mapping.
i) encodes using the input dominant algorithm.
o) encodes using the output dominant algorithm (default).
c) encodes using a combination of input and output dominant algorithms.
y) encodes using a modified output dominant algorithm.
.IP -\fBx\fP
expands the state codes using the invalid state codes as don't cares.
.IP -\fBc\fP
uses a group based embedding algorithm.
.IP -\fBp\fP
outputs in the PLA format.
.IP -\fBg\fP
reads a general symbolic encoding format and performs general encoding.
.SH "INPUT FILE FORMAT"
Below is a general input description.  
.nf

    .i 4
    .o 4
    .ilb c1 c2 c3 presentState
    .ob nextState o1 light1 light2
    .enum States 4 2 HG HY FG FY
    .enum Colors 3 2 GREEN RED BLUE
    .itype Boolean Boolean Boolean States
    .otype States Boolean Colors Colors
    0 - - HG  HG 0 GREEN RED
    - 0 - HG  HG 0 GREEN RED
    1 1 - HG  HY 1 GREEN RED
    - - 0 HY  HY 0 YELLOW RED
    - - 1 HY  FG 1 YELLOW RED
    1 0 - FG  FG 0 RED GREEN
    0 - - FG  FY 1 RED GREEN
    - 1 - FG  FY 1 RED GREEN
    - - 0 FY  FY 0 RED YELLOW
    - - 1 FY  HG 1 RED YELLOW

.fi
The user can specify different symbolic types using the ".enum" command.  
Then, the type of the input and output variables can be declared using 
the ".itype" and ".otype" commands, respectively.  The syntax of the ".enum"
command is: .enum \fItypename\fP \fInumber_of_symbols\fP \fIcode_length\fP 
symbol1 ... symboln.  A built-in type called "Boolean" can be used for binary 
variables.  Binary variables of type "Boolean" are kept as they are.  The ".i"
command specifies the number of input variables, and the ".o" command 
specifies the number of output variables.  The ."ilb" command is used to 
give the names of the input variables.  The ".ob" command is used to give 
the names of the output variables.  These two commands are optional.
.pp
Below is a finite state machine description for the state assignment
problem.
.nf

    .i 3
    .o 5
    .s 4
    0-- HG HG 00010
    -0- HG HG 00010
    11- HG HY 10010
    --0 HY HY 00110
    --1 HY FG 10110
    10- FG FG 01000
    0-- FG FY 11000
    -1- FG FY 11000
    --0 FY FY 01001
    --1 FY HG 11001

.fi
Typical state assignment use is the following:
.nf

    % jedi -p fsm.kiss2 | espresso > fsm.pla
    % misII
    misII> read_pla -s fsm.pla
    misII> source script
    misII> print_stats -f
    misII> write_blif fsm.blif
    misII> quit
    %

.fi
.SH "ADDITIONAL COMMENT"
When reading in the encoded PLA after encoding, "read_pla -s" often
works better than "read_pla".  The -s option reads in the PLA in a
collapsed single output single-level PLA form.
.SH "SEE ALSO"
misII(1OCTTOOLS), espresso(1OCTTOOLS)
.LP
Bill Lin and A. Richard Newton, "Synthesis of Multiple Level Logic from
Symbolic High-Level Description Languages", \fIProc. VLSI 89 Conference\fR,
Munich, West Germany, August 1989.
.SH "AUTHOR"
Please direct any questions or comments to:
.nf
Bill Lin 
Cory Hall
Dept. of EECS
University of California
Berkeley, CA 94720
email: billlin@ic.Berkeley.EDU
.fi