443 lines
14 KiB
Plaintext
443 lines
14 KiB
Plaintext
.\" $Header$
|
|
.de SB
|
|
.\" SuBheader
|
|
.sp 1
|
|
.PP
|
|
.nr Sf \\n(.f
|
|
.ft B
|
|
\\$1 \\$2 \\$3 \\$4 \\$5 \\$6 \\$7 \\$8 \\$9
|
|
.ft \\n(Sf
|
|
.sp 1
|
|
..
|
|
.TH ACK 1 "$Revision$"
|
|
.ad
|
|
.SH NAME
|
|
ack \- Amsterdam Compiler Kit
|
|
.SH SYNOPSIS
|
|
\fBack\fP arguments
|
|
.br
|
|
\fBacc\fP arguments
|
|
.br
|
|
\fBapc\fP arguments
|
|
.br
|
|
\fBabc\fP arguments
|
|
.br
|
|
\fBm2\fP arguments
|
|
.br
|
|
\fBocm\fP arguments
|
|
.br
|
|
\fBf2c\fP arguments
|
|
.br
|
|
\fImachine\fP arguments
|
|
.SH DESCRIPTION
|
|
This program transforms sources in several
|
|
languages to load files for a variety of machines,
|
|
internally using several phases.
|
|
The transformation can be stopped at any phase.
|
|
Combining sources from several languages is allowed.
|
|
The run-time system of the first language mentioned,
|
|
either in the program call name or in the arguments,
|
|
is automatically included.
|
|
The libraries of all other languages mentioned,
|
|
containing most of the run-time systems,
|
|
are also automatically included.
|
|
Two types of load files can be distinguished,
|
|
\fIa.out\fP files containing machine code and \fIe.out\fP
|
|
files containing virtual EM machine code.
|
|
The last type is designed for interpretation; it is obtained by calling
|
|
the compiler as \fIem22\fP, \fIem24\fP, or \fIem44\fP.
|
|
Compilation time for interpretation is fast and gives many
|
|
runtime checks,
|
|
but execution with \fIem\fP(1) is about seven times slower.
|
|
Execution with \fIint\fP(1) is even slower than that, but available on
|
|
more machines.
|
|
Which combinations of languages and machines are allowed varies
|
|
in time and depends on the installation.
|
|
.PP
|
|
The actions of \fIack\fP are to repeatedly transform files with a
|
|
particular suffix into files with another suffix,
|
|
finally combining the results into a single file.
|
|
.PP
|
|
Different machines can use different suffices,
|
|
but the following are recognized by most machines:
|
|
.IP .p
|
|
Pascal program.
|
|
.IP .c
|
|
C module.
|
|
.IP .b
|
|
Basic program.
|
|
.IP .ocm
|
|
Occam program.
|
|
.IP .mod
|
|
Modula-2 module.
|
|
.IP .f
|
|
Fortran module.
|
|
.IP .e
|
|
EM assembly module in human readable form.
|
|
.IP .k
|
|
Compact EM assembly code.
|
|
.IP .m
|
|
Optimized compact EM assembly code.
|
|
.IP .s
|
|
Machine assembly language code.
|
|
.IP .o
|
|
Object file.
|
|
.PP
|
|
\fIAck\fP accepts the following flags:
|
|
.SB machine used
|
|
.IP \-m\fImachine\fP
|
|
This flag tells \fIack\fP to generate a load file for \fImachine\fP.
|
|
\fIMachine\fP can also be used as the program call
|
|
name, instead of \fIack\fP.
|
|
e.g. \fIack \-mi86 file.p\fP is equivalent to \fIi86
|
|
file.p\fP.
|
|
.SB output files
|
|
.IP \-o
|
|
Use the next argument as the name of the resulting file.
|
|
\fIAck\fP produces \fIa.out\fP or \fIe.out\fP by default.
|
|
This flag can always be used when \fIack\fP produces a single output file,
|
|
as in
|
|
.br
|
|
.ti +5
|
|
\fBack -c.s main.c -o new.s\fP.
|
|
.br
|
|
The output is produced on \fInew.s\fP instead of \fImain.s\fP.
|
|
.IP \-c\fI.suffix\fP
|
|
.IP \-c
|
|
\fIAck\fP tries to transform each source into a file with the \fIsuffix\fP.
|
|
When no \fIsuffix\fP is specified \fIack\fP stops just
|
|
before the phase where it combines all arguments into a load file,
|
|
thereby transforming the sources into \fI.k\fP, \fI.s\fP,
|
|
\&\fI.o\fP or \fI.m\fP files.
|
|
One extra \fIsuffix\fP is recognized here, \fI.i\fP,
|
|
this tells \fIack\fP to only preprocess all human readable sources,
|
|
producing files with \fIsuffix\fP \fI.i\fP.
|
|
.br
|
|
Note: \fIack\fP refuses to overwrite argument \fI.e\fP files.
|
|
.IP \-t
|
|
Preserve all intermediate files.
|
|
If two \fB\-t\fP are used,
|
|
\fIack\fP also preserves core dumps and output of failed transformations.
|
|
.SB messages
|
|
.IP \-w
|
|
Suppress all warning messages.
|
|
.IP \-v
|
|
Verbose.
|
|
Print information while juggling with files.
|
|
.IP \-KR
|
|
Tell the C-compiler to conform to "Kernighan & Ritchie" C.
|
|
.IP \-ansi
|
|
Tell \fIack\fP to use the ANSI C compiler instead of the old one.
|
|
.IP \-fp
|
|
Use the software floating point package, if present.
|
|
.SB preprocessing
|
|
.IP \-I\fIdir\fP
|
|
\&\`#include\' files whose names do not begin with \`/\' are
|
|
always sought first in the directory of the \fIfile\fP argument,
|
|
then in the directories named in \fB\-I\fP options,
|
|
then in directories on a standard list.
|
|
.IP \-D\fIname=def\fP
|
|
.IP \-D\fIname\fP
|
|
Define the \fIname\fP to the preprocessor,
|
|
as if by \`#define\'.
|
|
If no definition is given the \fIname\fP is defined as 1.
|
|
.IP \-U\fIname\fP
|
|
Remove any initial definition of \fIname\fP, before
|
|
preprocessing.
|
|
.IP \-P
|
|
Don't generate line directives.
|
|
.IP \-C
|
|
Leave C-comments in.
|
|
.SB debugging
|
|
.IP \-p
|
|
This flag tells both the Pascal and C front ends to include
|
|
code enabling the user to do some monitoring/debugging.
|
|
Each time a routine is entered the routine \fBprocentry\fP
|
|
is called and just before each return \fBprocexit\fP is called.
|
|
These routines are supplied with one parameter, a pointer
|
|
to a string containing the name of the routine.
|
|
.SB optimizing
|
|
.IP \-O
|
|
.IP \-O\fInum\fP
|
|
.IP \-O\fIopt1,opt2,...\fP
|
|
Try to use the named optimizers, or, if no optimizers are given,
|
|
try to use the optimizers with optimization level <= \fInum\fP (default 1).
|
|
For optimizer names, see the table at the end of this manual page.
|
|
Currently, only the global optimizer has a level > 1.
|
|
Higher levels will invoke more passes of the global optimizer.
|
|
For the global optimizer, it is also possible to indicate explicitly which
|
|
phases must be executed, using the following flags:
|
|
.RS
|
|
.IP \-EGO-IL
|
|
Inline substitution.
|
|
.IP \-EGO-CS
|
|
Common subexpression elimination.
|
|
.IP \-EGO-SR
|
|
Strength reduction.
|
|
.IP \-EGO-UD
|
|
Use definition analysis.
|
|
.IP \-EGO-LV
|
|
Live variable analysis.
|
|
.IP \-EGO-RA
|
|
Register allocation.
|
|
.IP \-EGO-SP
|
|
Stack pollution.
|
|
.IP \-EGO-BO
|
|
Branch optimization.
|
|
.IP \-EGO-CJ
|
|
Cross jumping.
|
|
.RE
|
|
.IP ""
|
|
Also, the following flags may be used:
|
|
.RS
|
|
.IP \-s<num>
|
|
Give an indication to the inline substitution phase,
|
|
how much bigger the program may get,
|
|
in percentage. This is only used as a rough indication.
|
|
The inline substitution phase will not make the program bigger when given \-s0.
|
|
.IP \-a
|
|
Indicate to the inline substitution phase that it is offered the whole program.
|
|
This allows it to throw away routines that it has substituted inline.
|
|
.IP \-Q
|
|
Give some statistics.
|
|
.IP \-T
|
|
Optimize for time.
|
|
.IP \-S
|
|
Optimize for size.
|
|
.RE
|
|
.IP ""
|
|
In principle, the optimization phases can be run in any order; a phase
|
|
may even be run more than once. However, the following rules must be obeyed:
|
|
.RS
|
|
.IP -
|
|
the Live Variable analysis phase (LV) must be run prior to the
|
|
Register Allocation phase (SA), as SA uses information produced by LV.
|
|
.IP -
|
|
SA should be the last phase.
|
|
.RE
|
|
.IP ""
|
|
Also, the following may be of use:
|
|
.RS
|
|
.IP -
|
|
Inline Substitution (IL) may create new opportunities for most other phases,
|
|
so it should be run as early as possible.
|
|
.IP -
|
|
Use Definition analysis (UD) may introduce opportunities for LV.
|
|
.IP -
|
|
Strength Reduction (SR) may create opportunities for UD.
|
|
.RE
|
|
.IP ""
|
|
The global optimizer is a combiner, so it should be offered all the source
|
|
files of the program. This is not strictly necessary, but it makes the
|
|
global optimizer more effective.
|
|
The current default optimization phases are:
|
|
.RS
|
|
.IP -
|
|
for \-O2: CJ, BO, SP;
|
|
.IP -
|
|
for \-O3: CS, SR, CJ, BO, SP, UD, LV, RA;
|
|
.IP -
|
|
for \-O4: IL, CS, SR, CJ, BO, SP, UD, LV, RA;
|
|
.IP -
|
|
for higher levels: as for \-O4.
|
|
.RE
|
|
.IP \-L
|
|
Disable the generation of code by the front ends to
|
|
record line number and source file name at run-time.
|
|
(This is the default for C and Fortran).
|
|
.SB libraries
|
|
.IP \-l\fIname\fP
|
|
Tells \fIack\fP to insert a library module at this point.
|
|
For example: the library \fImon\fP contains the
|
|
routines for systems calls needed by both C and Pascal.
|
|
.IP \-.\fIsuffix\fP
|
|
When linking multiple \fI.o\fP or \fI.m\fP files created by
|
|
separate calls of \fIack\fP together, \fIack\fP cannot deduce
|
|
the run-time system needed,
|
|
unless called as \fIapc\fP, \fIabc\fP, \fIocm\fP, \fIm2\fP, \fIf2c\fP, or \fIacc\fP.
|
|
This flag serves to tell \fIack\fP which runtime system is
|
|
needed in such a case.
|
|
For example: "ack \-c x.c ; ack \-.c x.o".
|
|
.IP \-r.\fIsuffix\fP
|
|
Most frontends and backends use one or
|
|
more run-time libraries.
|
|
These flags tell \fIack\fP to include the libraries needed when
|
|
a file with \fIsuffix\fP would be included in the arguments.
|
|
.IP \-LIB
|
|
This flag tells the peephole optimizer
|
|
.RF em_opt 6
|
|
to add information about the visibility of the names used
|
|
to each output module.
|
|
This is needed by
|
|
assembler/linkers when these modules are to be inserted
|
|
in libraries.
|
|
.SB interpreter
|
|
.IP \-{xxx}
|
|
The string starting after \`{\' and terminated by a \`}\' is passed
|
|
as an option string to the Pascal compiler and supersedes corresponding
|
|
options given in the source file.
|
|
See the ACK reference manual [4] for a list of options.
|
|
.IP "\-+xxx, \-\-xxx"
|
|
When interpreting programs, these flags are used to select some
|
|
options during interpretation, like test, profile, flow, extra and count.
|
|
A short description of these flags follows:
|
|
.RS
|
|
.IP " t(est)" 12
|
|
test for undefined, overflow, array bound etc.
|
|
.IP " f(low)"
|
|
keep track of executed source lines.
|
|
.IP " c(ount)"
|
|
count the number of times a source line is executed.
|
|
.IP " p(rofile)"
|
|
count the memory cycles executed per source line.
|
|
.RE
|
|
.IP "" 5
|
|
Test is on by default, the others are off. Normally, these
|
|
flag options are given each time the interpreter is run.
|
|
The EM assembler/linker offers the opportunity to change
|
|
the defaults per program.
|
|
The changed options are recorded in the "e.out" header.
|
|
These flags \-\- and \-+ are passed to the assembler for this purpose.
|
|
So, \-\-t and \-+pfce invert the defaults.
|
|
.SB general
|
|
.IP \-R\fIprogram=xxx\fP
|
|
Replace the \fIprogram\fP by the pathname \fIxxx\fP.
|
|
The program names referred to later in this manual are allowed here.
|
|
.IP \-R\fIprogram\-xxx\fP
|
|
The flag argument \fI\-xxx\fP is given to \fIprogram\fP.
|
|
.IP \-R\fIprogram:n\fP
|
|
Set the priority of the indicated transformation to \fIn\fP.
|
|
The default priority is 0, setting it to -1 makes it highly
|
|
unlikely the the phase will be used, setting it to 1 makes
|
|
it very likely that the phase will be used.
|
|
.IP \-k
|
|
Do not stop when an error occurs, but try to transform all
|
|
other arguments as far as possible.
|
|
.IP \-g
|
|
Instruct the front-end and back-end to produce symbolic debugger information
|
|
for \fIgrind\fP(1). This is currently supported by the following front-ends:
|
|
C, ANSI-C, Pascal, Modula-2, and the following back-ends: Motorola M68020,
|
|
Sparc, Vax, Intel 80386.
|
|
.IP \-gdb
|
|
Instruct the front-end and back-end to produce symbolic debugger information
|
|
for \fIgdb\fP. This is currently supported by the following front-ends:
|
|
C, ANSI-C, Pascal, Modula-2, and the following back-ends: Motorola M68020,
|
|
Sparc, Vax, Intel 80386.
|
|
.PP
|
|
All arguments without a suffix or with an unrecognized suffix
|
|
are passed to the loaders, as for flags.
|
|
.SH PREPROCESSOR
|
|
All C source programs are run through the preprocessor
|
|
before they are fed to the compiler proper.
|
|
On larger machines, the compiler has a built in preprocessor.
|
|
Other human readable sources (Modula-2, Pascal, or Basic programs and
|
|
machine assembly) are only preprocessed when they start with a \`#\'.
|
|
.PP
|
|
\fIAck\fP adds a few macro definitions when it calls the
|
|
preprocessor.
|
|
These macro\'s contain the word- and pointer-size and the sizes
|
|
of some basic types used by the Pascal, Basic and/or C compiler.
|
|
All sizes are in bytes.
|
|
.sp 1
|
|
.TS
|
|
tab(:);
|
|
l l l l.
|
|
EM_WSIZE:wordsize:EM_PSIZE:pointer size
|
|
EM_SSIZE:size of shorts (C):EM_LSIZE:size of longs (C+Pascal)
|
|
EM_FSIZE:size of floats (C):EM_DSIZE:size of doubles (C+Pascal)
|
|
.TE
|
|
.sp 1
|
|
The name of the \fImachine\fP or something like it when
|
|
the machine name is numeric is also defined (as 1).
|
|
As the ANSI C rules forbid this, in ANSI C, underscores are prepended to
|
|
these names.
|
|
.PP
|
|
The default directories searched for include files differ for each machine.
|
|
.SH PROGRAMS
|
|
\fIAck\fP uses one or more programs in each phase of the
|
|
transformation.
|
|
The table below gives the names \fIack\fP uses for these
|
|
programs.
|
|
Internally \fIack\fP maintains a mapping of these names to pathnames
|
|
for load files.
|
|
The table specifies which type of files are accepted by each
|
|
program as input and the file type produced as output.
|
|
.sp 1
|
|
.in +2
|
|
.TS
|
|
tab(:);
|
|
l l l l.
|
|
input:name:output:description
|
|
\&.f:f77:.c:Fortran-to-C front end
|
|
\&.c:cem:.k:C front end [4,5,6]
|
|
\&.p:pc:.k:Pascal front end [2,3,6]
|
|
\&.b:abc:.k:Basic front end [6,8]
|
|
\&.ocm:ocm:.k:Occam front end [9]
|
|
\&.mod:m2:.k:Modula-2 front end [11]
|
|
\&.e:encode:.k:Compactify EM assembly language [1]
|
|
\&.k:opt:.m:EM peephole optimizer
|
|
\&.k .m:decode:.e:Produce human readable EM assembly
|
|
\&.k .m:emass:e.out:Linker producing EM machine code [1]
|
|
\&.m:ego:.gk:EM global optimizer [10]
|
|
\&.gk:opt2:.g:Second EM peephole optimizer
|
|
\&.m .g:be:.s:backend
|
|
\&.s:asopt:.so:target optimizer
|
|
\&.s .so:asld:.out:Assembler/linker, Ack object format
|
|
\&.s .so:as:.o:Assembler, relocatable object
|
|
\&.o:ld:a.out:Linker, machine a.out format
|
|
\&.o:led:.out:Linker, Ack object format
|
|
\&.o:cv:a.out:Conversion from Ack object to machine object
|
|
.TE
|
|
.in -2
|
|
.SH "ENVIRONMENT VARIABLES"
|
|
.IP ACKDIR
|
|
If set, this environment variable overrides ack's idea of its home
|
|
directory.
|
|
.IP ACKM
|
|
If set, this environment variable overrides ack's idea of the default
|
|
machine it compiles for.
|
|
.IP ACKFE
|
|
If set, this environment variable tells ack where to get the front-end
|
|
description file.
|
|
.SH "SEE ALSO"
|
|
.PD 0
|
|
em_opt(6), em_ass(6), em_cg(6), ack.out(5)
|
|
.IP [1]
|
|
A.S. Tanenbaum, Hans van Staveren, Ed Keizer and Johan
|
|
Stevenson, \fIDescription of a machine architecture for use with
|
|
block structured languages\fP, Informatica rapport IR-81.
|
|
.IP [2]
|
|
K. Jensen and N. Wirth,
|
|
\fIPASCAL, User manual and report\fP Springer Verlag.
|
|
.IP [3]
|
|
The ISO Pascal standard proposal ISO/TC97/SC5-N462.
|
|
.IP [4]
|
|
B.W. Kernighan and D.M. Ritchie, \fIThe C Programming
|
|
language\fP, Prentice-Hall, 1978
|
|
.IP [5]
|
|
D.M. Ritchie, \fI C Reference Manual\fP
|
|
.IP [6]
|
|
Amsterdam Compiler Kit, reference manuals and UNIX manual pages.
|
|
.IP [7]
|
|
E.G. Keizer, Ack description file reference manual.
|
|
.IP [8]
|
|
M.L. Kersten a.o., \fIThe ABC compiler\fP.
|
|
.IP [9]
|
|
Kees Bot and Edwin Scheffer, \fIAn Occam Compiler\fP, IM-6.
|
|
.IP [10]
|
|
H.E. Bal, \fIThe design and implementation of the EM Global Optimizer\fP,
|
|
Informatica rapport IR-99.
|
|
.IP [11]
|
|
C.J.H. Jacobs, \fIThe ACK Modula-2 Compiler\fP.
|
|
.PD
|
|
.SH DIAGNOSTICS
|
|
.PD
|
|
The diagnostics are intended to be self\-explanatory.
|
|
.SH BUGS
|
|
Not all warning messages are superseded by \fB\-w\fP.
|
|
.br
|
|
Argument assembly files are not preprocessed when fed into the
|
|
universal assembler/loader.
|
|
.SH AUTHOR
|
|
Ed Keizer, Vrije Universiteit, Amsterdam
|