mirror of
https://github.com/Mercury-Language/mercury.git
synced 2025-12-18 15:26:31 +00:00
311d0af1e5
Estimated hours taken: 90
Constraint propagation.
compiler/constraint.m:
Push constraints left and inwards as much as possible
within a goal. This module has been completely rewritten.
compiler/deforest.m:
Push constraints within a goal before processing it.
Make specialized versions for calls with constrained outputs.
Rerun determinism inference on specialized versions
when constraint propagation has been run, because the
determinism can change from nondet to semidet.
compiler/pd_util.m:
Add pd_util__propagate_constraints, which uses constraint.m
to push constraints within a goal.
Add some documentation for the exported predicates.
compiler/pd_term.m:
Add support for checking termination of the optimization process
for constraint propagation, which differs from deforestation
in that the conjunctions selected for optimization don't
necessarily have a call at both ends.
compiler/pd_debug.m:
Print some extra information when `--debug-pd' is enabled.
compiler/mercury_compile.m:
Check whether constraint propagation should be performed when
working out whether to run the deforestation pass.
compiler/make_hlds.m:
Add `no_inline' markers to the "recursive" procedures
introduced for builtins to stop constraint propagation
attempting to specialize such procedures.
compiler/hlds_pred.m:
Don't fill in the declared determinism field of the predicates
introduced by `hlds_pred__define_new_pred', so that rerunning
determinism inference will compute a more accurate determinism.
compiler/inlining.m:
Requantify before recomputing instmap_deltas, not after.
compiler/det_report.m:
Add predicates to disable warnings when rerunning
determinism analysis after constraint propagation.
compiler/options.m:
Add documentation for `--constraint-propagation'.
Add option `--local-constraint-propagation', which makes
deforestation call constraint.m to move constraints within
a goal, but does not create specialized versions of procedures
for which there are calls with constrained outputs.
compiler/handle_options.m:
`--constraint-propagation' implies `--local-constraint-propagation'.
compiler/notes/compiler_design.html:
Change the documentation to show that constraint.m is now part
of the deforestation pass.
NEWS:
Announce the new transformation.
doc/user_guide.texi:
Document the new options.
tests/hard_coded/Mmakefile:
tests/hard_coded/constraint.{m,exp}:
tests/hard_coded/constraint_order.{m,exp}:
Test cases.
1163 lines
40 KiB
HTML
1163 lines
40 KiB
HTML
<html>
|
|
<head>
|
|
<title>
|
|
Notes On The Design Of The Mercury Compiler
|
|
</title>
|
|
</head>
|
|
|
|
<body bgcolor="#ffffff" text="#000000">
|
|
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
This file contains an overview of the design of the compiler.
|
|
|
|
See also <a href="overall_design.html">overall_design.html</a>
|
|
for an overview of how the different sub-systems (compiler,
|
|
library, runtime, etc.) fit together.
|
|
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h2> OUTLINE </h2>
|
|
|
|
<p>
|
|
|
|
The main job of the compiler is to translate Mercury into C, although it
|
|
can also translate (subsets of) Mercury to some other languages:
|
|
Mercury bytecode (for a planned bytecode interpreter), MSIL (for the
|
|
Microsoft .NET platform) and RL (the Aditi Relational Language).
|
|
|
|
<p>
|
|
|
|
The top-level of the compiler is in the file mercury_compile.m.
|
|
The basic design is that compilation is broken into the following
|
|
stages:
|
|
|
|
<ul>
|
|
<li> 1. parsing (source files -> HLDS)
|
|
<li> 2. semantic analysis and error checking (HLDS -> annotated HLDS)
|
|
<li> 3. high-level transformations (annotated HLDS -> annotated HLDS)
|
|
<li> 4. code generation (annotated HLDS -> target representation)
|
|
<li> 5. low-level optimizations
|
|
(target representation -> target representation)
|
|
<li> 6. output code (target representation -> target code)
|
|
</ul>
|
|
|
|
Note that in reality the separation is not quite as simple as that.
|
|
Although parsing is listed as step 1 and semantic analysis is listed
|
|
as step 2, the last stage of parsing actually includes some semantic checks.
|
|
And although optimization is listed as steps 3 and 5, it also occurs in
|
|
steps 2, 4, and 6. For example, elimination of assignments to dead
|
|
variables is done in mode analysis; middle-recursion optimization and
|
|
the use of static constants for ground terms is done in code
|
|
generation; and a few low-level optimizations are done in llds_out.m
|
|
as we are spitting out the C code.
|
|
|
|
<p>
|
|
|
|
In addition, the compiler is actually a multi-targeted compiler
|
|
with several different back-ends. When you take the different
|
|
back-ends into account, the structure looks like this:
|
|
|
|
<ul type=disc>
|
|
<li> front-end
|
|
<ul type=disc>
|
|
<li> 1. parsing (source files -> HLDS)
|
|
<li> 2. semantic analysis and error checking (HLDS -> annotated HLDS)
|
|
<li> 3. high-level transformations (annotated HLDS -> annotated HLDS)
|
|
</ul>
|
|
<li> back-ends
|
|
<ul type=disc>
|
|
<li> a. LLDS back-end
|
|
<ul type=disc>
|
|
<li> 4a. code generation (annotated HLDS -> LLDS)
|
|
<li> 5a. low-level optimizations (LLDS -> LLDS)
|
|
<li> 6a. output code (LLDS -> C)
|
|
</ul>
|
|
<li> b. MLDS back-end
|
|
<ul type=disc>
|
|
<li> 4b. code generation (annotated HLDS -> MLDS)
|
|
<li> 5b. MLDS transformations (MLDS -> MLDS)
|
|
<li> 6b. output code
|
|
(MLDS -> C or MLDS -> MSIL
|
|
or eventually MLDS -> Java, etc.)
|
|
</ul>
|
|
<li> c. RL back-end
|
|
<ul type=disc>
|
|
<li> 4c. code generation (annotated HLDS -> RL)
|
|
<li> 5c. low-level optimizations (RL -> RL)
|
|
<li> 6c. output code (RL -> RL-bytecode)
|
|
</ul>
|
|
<li> d. bytecode back-end
|
|
<ul type=disc>
|
|
<li> 4d. code generation (annotated HLDS -> bytecode)
|
|
</ul>
|
|
</ul>
|
|
</ul>
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h2> DETAILED DESIGN </h2>
|
|
|
|
This section describes the role of each module in the compiler.
|
|
For more information about the design of a particular module,
|
|
see the documentation at the start of that module's source code.
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
<p>
|
|
|
|
The action is co-ordinated from mercury_compile.m.
|
|
|
|
<p>
|
|
|
|
<h3> Option handling </h3>
|
|
|
|
<p>
|
|
|
|
The command-line options are defined in the module options.m.
|
|
mercury_compile.m calls library/getopt.m, passing the predicates
|
|
defined in options.m as arguments, to parse them. It then invokes
|
|
handle_options.m to postprocess the option set. The results are
|
|
stored in the io__state, using the type globals defined in globals.m.
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> FRONT END </h3>
|
|
<h4> 1. Parsing </h4>
|
|
|
|
<p>
|
|
|
|
<ul>
|
|
|
|
<li> lexical analysis (library/lexer.m)
|
|
|
|
<li> stage 1 parsing - convert strings to terms. <br>
|
|
|
|
library/parser.m contains the code to do this, while
|
|
library/term.m and library/varset.m contain the term and varset
|
|
data structures that result, and predicates for manipulating them.
|
|
|
|
<li> stage 2 parsing - convert terms to `items' (declarations, clauses, etc.)
|
|
<br>
|
|
|
|
The result of this stage is a parse tree that has a one-to-one
|
|
correspondence with the source code. The parse tree data structure
|
|
definition is in prog_data.m, while the code to create it is in
|
|
prog_io.m and its submodules prog_io_dcg.m (which handles clauses
|
|
using Definite Clause Grammar notation), prog_io_goal.m (which handles
|
|
goals), prog_io_pragma.m (which handles pragma declarations),
|
|
prog_io_typeclass.m (which handles typeclass and instance declarations)
|
|
and prog_io_util.m (which defines predicates and types needed by the
|
|
other prog_io*.m modules. The data structure for insts is stored in
|
|
its own module, inst.m.
|
|
|
|
<p>
|
|
|
|
The modules prog_out.m and mercury_to_mercury.m contain predicates
|
|
for printing the parse tree. prog_util.m contains some utility
|
|
predicates for manipulating the parse tree.
|
|
|
|
<li> imports and exports are handled at this point (modules.m) <br>
|
|
|
|
modules.m has the code to write out `.int', `.int2', `.int3',
|
|
`.d' and `.dep' files.
|
|
|
|
<li> module qualification of types, insts and modes <br>
|
|
|
|
module_qual.m - <br>
|
|
Adds module qualifiers to all types insts and modes,
|
|
checking that a given type, inst or mode exists and that
|
|
there is only possible match. This is done here because
|
|
it must be done before the `.int' and `.int2' interface files
|
|
are written. This also checks whether imports are really needed
|
|
in the interface.
|
|
<br>
|
|
Notes on module qualification:
|
|
<ul>
|
|
<li> all types, typeclasses, insts and modes occuring in pred, func,
|
|
type, typeclass and mode declarations are module qualified by
|
|
module_qual.m.
|
|
<li> all types, insts and modes occuring in lambda expressions,
|
|
explicit type qualifications, and clause mode annotations
|
|
are module qualified in make_hlds.m.
|
|
<li> constructors occuring in predicate and function mode declarations
|
|
are module qualified during type checking.
|
|
<li> predicate and function calls and constructors within goals
|
|
are module qualified during mode analysis.
|
|
</ul>
|
|
|
|
|
|
<li> reading and writing of optimization interfaces
|
|
(intermod.m and trans_opt.m). <br>
|
|
|
|
<module>.opt contains clauses for exported preds suitable for
|
|
inlining or higher-order specialization. The `.opt' file for the
|
|
current module is written after type-checking. `.opt' files
|
|
for imported modules are read here.
|
|
<module>.opt contains termination analysis information
|
|
for exported preds (eventually it ought to contain other
|
|
"transitive" information too, e.g. for optimization, but
|
|
currently it is only used for termination analysis).
|
|
`.trans_opt' files for imported modules are read here.
|
|
The `.trans_opt' file for the current module is written
|
|
after the end of semantic analysis.
|
|
|
|
<li> expansion of equivalence types (equiv_type.m) <br>
|
|
|
|
This is really part of type-checking, but is done
|
|
on the item_list rather than on the HLDS because it
|
|
turned out to be much easier to implement that way.
|
|
|
|
<li> conversion to superhomogeneous form and into HLDS <br>
|
|
|
|
make_hlds.m transforms the code into superhomogeneous form,
|
|
and at the same time converts the parse tree into the HLDS.
|
|
It expands away universal quantification
|
|
(using `all [Vs] G' ===> `not (some [Vs] (not G))')
|
|
and implication (using `A => B' ===> `not(A, not B)').
|
|
It converts `pragma import', `pragma c_code' and `pragma fact_table'
|
|
declarations into clauses with HLDS `pragma_c_code'
|
|
instructions for bodies.
|
|
The `pragma fact_table' conversion is done by calling fact_table.m
|
|
which also reads the facts from the declared file and compiles them
|
|
into a separate C file for which the `pragma_c_code' contains lookup
|
|
code.
|
|
make_hlds.m also calls make_tags.m which chooses the data
|
|
representation for each discriminated union type by
|
|
assigning tags to each functor.
|
|
</ul>
|
|
|
|
<p>
|
|
|
|
The result at this stage is the High Level Data Structure,
|
|
which is defined in four files:
|
|
|
|
<ol>
|
|
<li> hlds_data.m defines the parts of the HLDS concerned with
|
|
function symbols, types, insts, modes and determinisms;
|
|
<li> hlds_goal.m defines the part of the HLDS concerned with the
|
|
structure of goals, including the annotations on goals;
|
|
<li> hlds_pred.m defines the part of the HLDS concerning
|
|
predicates and procedures;
|
|
<li> hlds_module.m defines the top-level parts of the HLDS,
|
|
including the type module_info.
|
|
</ol>
|
|
|
|
The module hlds_out.m contains predicates to dump the HLDS to a file.
|
|
The module goal_util.m contains predicates for renaming variables
|
|
in an HLDS goal.
|
|
|
|
<p>
|
|
|
|
<h4> 2. Semantic analysis and error checking </h4>
|
|
|
|
<p>
|
|
|
|
Any pass which can report errors or warnings must be part of this stage,
|
|
so that the compiler does the right thing for options such as
|
|
`--halt-at-warn' (which turns warnings into errors) and
|
|
`--error-check-only' (which makes the compiler only compile up to this stage).
|
|
|
|
<p>
|
|
|
|
<dl>
|
|
|
|
<dt> implicit quantification
|
|
|
|
<dd>
|
|
quantification.m handles implicit quantification and computes
|
|
the set of non-local variables for each sub-goal.
|
|
It also expands away bi-implication (unlike the expansion
|
|
of implication and universal quantification, this expansion
|
|
cannot be done until after quantification).
|
|
This pass is called from the `transform' predicate in make_hlds.m.
|
|
|
|
<dt> checking typeclass instances (check_typeclass.m)
|
|
<dd>
|
|
check_typeclass.m both checks that instance declarations satisfy all
|
|
the appropriate superclass constraints and
|
|
performs a source-to-source transformation on the
|
|
methods methods from the instance declarations.
|
|
The transformed code is checked for type, mode, uniqueness, purity
|
|
and determinism correctness by the later passes, which has the effect
|
|
of checking the correctness of the instance methods themselves
|
|
(ie. that the instance methods match those expected by the typeclass
|
|
declaration).
|
|
During the transformation,
|
|
pred_ids and proc_ids are assigned to the methods for each instance.
|
|
|
|
In
|
|
addition, while checking that the superclasses of a class are satisfied
|
|
by the instance declaration, a set of constraint_proofs are built up
|
|
for the superclass constraints. These are used by polymorphism.m when
|
|
generating the base_typeclass_info for the instance.
|
|
|
|
<dt> type checking
|
|
|
|
<dd>
|
|
<ul>
|
|
<li> typecheck.m handles type checking, overloading resolution &
|
|
module name resolution, and almost fully qualifies all predicate
|
|
and functor names. It sets the map(var, type) field in the
|
|
pred_info. However, typecheck.m doesn't figure out the pred_id
|
|
for function calls or calls to overloaded predicates; that can't
|
|
be done in a single pass of typechecking, and so it is done
|
|
later on (in post_typecheck.m, for both preds and function calls)
|
|
Typeclass constraints are checked here, and
|
|
any redundant constraints that are eliminated are recorded (as
|
|
constraint_proofs) in the pred_info for future reference. When it has
|
|
finished, typecheck.m calls clause_to_proc.m to make duplicate copies
|
|
of the clauses for each different mode of a predicate; all later
|
|
stages work on procedures, not predicates.
|
|
<li> type_util.m contains utility predicates dealing with types
|
|
that are used in a variety of different places within the compiler
|
|
<li> post_typecheck.m may also be considered to logically be a part
|
|
of typechecking, but it is actually called from purity
|
|
analysis (see below). It contains the stuff related to
|
|
type checking that can't be done in the main type checking pass.
|
|
It also removes assertions from further processing.
|
|
</ul>
|
|
|
|
<dt> assertions
|
|
|
|
<dd>
|
|
assertion.m is the abstract interface to the assertion table.
|
|
Currently all the compiler does is type check the assertions and
|
|
record for each predicate that is used in an assertion, which
|
|
assertion it is used in. The set up of the assertion table occurs
|
|
in post_typecheck__finish_assertion.
|
|
|
|
<dt> purity analysis
|
|
|
|
<dd>
|
|
purity.m is responsible for purity checking, as well as
|
|
defining the <CODE>purity</CODE> type and a few public
|
|
operations on it. It also calls post_typecheck.m to
|
|
complete the handling of predicate
|
|
overloading for cases which typecheck.m is unable to handle,
|
|
and to check for unbound type variables.
|
|
Elimination of double negation is also done here; that needs to
|
|
be done after quantification analysis and before mode analysis.
|
|
|
|
<dt> polymorphism transformation
|
|
|
|
<dd>
|
|
polymorphism.m handles introduction of type_info arguments for
|
|
polymorphic predicates and introduction of typeclass_info arguments
|
|
for typeclass-constrained predicates.
|
|
This phase needs to come before mode analysis so that mode analysis
|
|
can properly reorder code involving existential types.
|
|
(It also needs to come before simplification so that simplify.m's
|
|
optimization of goals with no output variables doesn't do the
|
|
wrong thing for goals whose only output is the type_info for
|
|
an existentially quantified type parameter.)
|
|
<p>
|
|
This phase also
|
|
converts higher-order predicate terms into lambda expressions,
|
|
and copies the clauses to the proc_infos in preparation for
|
|
mode analysis.
|
|
<p>
|
|
The polymorphism.m module also exports some utility routines that
|
|
are used by other modules. These include some routines for generating
|
|
code to create type_infos, which are used by simplify.m and magic.m
|
|
when those modules introduce new calls to polymorphic procedures.
|
|
|
|
<dt> mode analysis
|
|
|
|
<dd>
|
|
<ul>
|
|
<li> modes.m is the main mode analysis module.
|
|
It checks that the code is mode-correct, reordering it
|
|
if necessary, and annotates each goal with a delta-instmap
|
|
that specifies the changes in instantiatedness of each
|
|
variable over that goal.
|
|
<li> modecheck_unify.m is the sub-module which analyses
|
|
unification goals.
|
|
It also module qualifies data constructors.
|
|
<li> modecheck_call.m is the sub-module which analyses calls.
|
|
|
|
<p>
|
|
|
|
The following sub-modules are used:
|
|
<dl>
|
|
<dt> mode_info.m
|
|
<dd>
|
|
(the main data structure for mode analysis)
|
|
<dt> delay_info.m
|
|
<dd>
|
|
(a sub-component of the mode_info data
|
|
structure used for storing the information
|
|
for scheduling: which goals are currently
|
|
delayed, what variables they are delayed on, etc.)
|
|
<dt> instmap.m
|
|
<dd>
|
|
Defines the instmap and instmap_delta ADTs
|
|
which store information on what instantiations
|
|
a set of variables may be bound to.
|
|
<dt> inst_match.m
|
|
<dd>
|
|
This contains the code for examining insts and
|
|
checking whether they match.
|
|
<dt> inst_util.m
|
|
<dd>
|
|
This contains the code for creating new insts from
|
|
old ones: unifying them, merging them and so on.
|
|
<dt> mode_errors.m
|
|
<dd>
|
|
This module contains all the code to
|
|
print error messages for mode errors
|
|
</dl>
|
|
<li> mode_util.m contains miscellaneous useful predicates dealing
|
|
with modes (many of these are used by lots of later stages
|
|
of the compiler)
|
|
<li> mode_debug.m contains utility code for tracing the actions
|
|
of the mode checker.
|
|
</ul>
|
|
|
|
<dt> indexing and determinism analysis
|
|
|
|
<dd>
|
|
<ul>
|
|
<li> switch_detection.m transforms into switches those disjunctions
|
|
in which several disjuncts test the same variable against different
|
|
function symbols.
|
|
<li> cse_detection.m looks for disjunctions in which each disjunct tests
|
|
the same variable against the same function symbols, and hoists any
|
|
such unifications out of the disjunction.
|
|
If cse_detection.m modifies the code,
|
|
it will re-run mode analysis and switch detection.
|
|
<li> det_analysis.m annotates each goal with its determinism;
|
|
it inserts cuts in the form of "some" goals wherever the determinisms
|
|
and delta instantiations of the goals involved make it necessary.
|
|
Any errors found during determinism analysis are reported by
|
|
det_report.m.
|
|
Det_util.m contains utility predicates used in several modules.
|
|
</ul>
|
|
|
|
<dt> checking of unique modes (unique_modes.m)
|
|
|
|
<dd>
|
|
unique_modes.m checks that non-backtrackable unique modes were
|
|
not used in a context which might require backtracking.
|
|
Note that what unique_modes.m does is quite similar to
|
|
what modes.m does, and unique_modes calls lots of predicates
|
|
defined in modes.m to do it.
|
|
|
|
<dt> stratification checking
|
|
|
|
<dd>
|
|
The module stratify.m implements the `--warn-non-stratification'
|
|
warning, which is an optional warning that checks for loops
|
|
through negation.
|
|
|
|
<dt> simplification (simplify.m)
|
|
|
|
<dd>
|
|
simplify.m finds and exploits opportunities for simplifying the
|
|
internal form of the program, both to optimize the code and to
|
|
massage the code into a form the code generator will accept.
|
|
It also warns the programmer about any constructs that are so simple
|
|
that they should not have been included in the program in the first
|
|
place. (That's why this pass needs to be part of semantic analysis:
|
|
because it can report warnings.)
|
|
simplify.m converts complicated unifications into procedure calls.
|
|
simplify.m calls common.m which looks for (a) construction unifications
|
|
that construct a term that is the same as one that already exists,
|
|
or (b) repeated calls to a predicate with the same inputs, and replaces
|
|
them with assignment unifications.
|
|
simplify.m also attempts to partially evaluate calls to builtin
|
|
procedures if the inputs are all constants (see const_prop.m),
|
|
|
|
</dl>
|
|
|
|
<h4> 3. High-level transformations </h4>
|
|
|
|
<p>
|
|
|
|
The first pass of this stage does tabling transformations (table_gen.m).
|
|
This involves the insertion of several calls to tabling predicates
|
|
defined in mercury_builtin.m and the addition of some scaffolding structure.
|
|
Note that this pass can change the evaluation methods of some procedures to
|
|
eval_table_io, so it should come before any passes that require definitive
|
|
evaluation methods (e.g. inlining).
|
|
|
|
<p>
|
|
|
|
The next pass of this stage is a code simplification, namely
|
|
removal of lambda expressions (lambda.m):
|
|
|
|
<ul>
|
|
<li>
|
|
lambda.m converts lambda expressions into higher-order predicate
|
|
terms referring to freshly introduced separate predicates.
|
|
This pass needs to come after unique_modes.m to ensure that
|
|
the modes we give to the introduced predicates are correct.
|
|
It also needs to come after polymorphism.m since polymorphism.m
|
|
doesn't handle higher-order predicate constants.
|
|
</ul>
|
|
|
|
(Is there any good reason why lambda.m comes after table_gen.m?)
|
|
|
|
<p>
|
|
|
|
The next pass is termination analysis. The various modules involved are:
|
|
|
|
<ul>
|
|
<li>
|
|
termination.m is the control module. It sets the argument size and
|
|
termination properties of builtin and compiler generated procedures,
|
|
invokes term_pass1.m and term_pass2.m
|
|
and writes .trans_opt files and error messages as appropriate.
|
|
<li>
|
|
term_pass1.m analyzes the argument size properties of user-defined procedures,
|
|
<li>
|
|
term_pass2.m analyzes the termination properties of user-defined procedures.
|
|
<li>
|
|
term_traversal.m contains code common to the two passes.
|
|
<li>
|
|
term_errors.m defines the various kinds of termination errors
|
|
and prints the messages appropriate for each.
|
|
<li>
|
|
term_util.m defines the main types used in termination analysis
|
|
and contains utility predicates.
|
|
<li>
|
|
lp.m is used by term_pass1.m. It implements the Linear Programming
|
|
algorithm for optimizing a set of linear constraints with respect to
|
|
a linear cost function.
|
|
</ul>
|
|
|
|
<p>
|
|
|
|
Most of the remaining HLDS-to-HLDS transformations are optimizations:
|
|
|
|
<ul>
|
|
<li> specialization of higher-order and polymorphic predicates where the
|
|
value of the higher-order/type_info/typeclass_info arguments are known
|
|
(higher_order.m)
|
|
|
|
<li> attempt to introduce accumulators (accumulator.m). This optimizes
|
|
procedures whose tail consists of independent associative computations
|
|
or independant chains of commutative computations into a tail
|
|
recursive form by the introduction of accumulators. If lco is turned
|
|
on it can also transform some procedures so that only construction
|
|
unifications are after the recursive call. This pass must come before
|
|
lco, unused_args (eliminating arguments makes it hard to relate the
|
|
code back to the assertion) and inlining (can make the associative
|
|
call disappear).
|
|
|
|
<li> inlining (i.e. unfolding) of simple procedures (inlining.m)
|
|
|
|
<li> deforestation and partial evaluation (deforest.m). This optimizes
|
|
multiple traversals of data structures within a conjunction, and
|
|
avoids creating intermediate data structures. It also performs
|
|
loop unrolling where the clause used is known at compile time.
|
|
deforest.m makes use of the following sub-modules
|
|
(`pd_' stands for "partial deduction"):
|
|
<ul>
|
|
<li>
|
|
constraint.m transforms goals so that goals which can fail
|
|
are executed earlier.
|
|
<li>
|
|
pd_cost.m contains some predicates to estimate the improvement
|
|
caused by deforest.m.
|
|
<li>
|
|
pd_debug.m produces debugging output.
|
|
<li>
|
|
pd_info.m contains a state type for deforestation.
|
|
<li>
|
|
pd_term.m contains predicates to check that the deforestation algorithm
|
|
terminates.
|
|
<li>
|
|
pd_util.m contains various utility predicates.
|
|
</ul>
|
|
|
|
<li> issue warnings about unused arguments from predicates, and create
|
|
specialized versions without them (unused_args.m); type_infos are
|
|
often unused.
|
|
|
|
<li> delay_construct.m pushes construction unifications to the right in
|
|
semidet conjunctions, in an effort to reduce the probability that it will
|
|
need to be executed.
|
|
|
|
<li> unneeded_code.m looks for goals whose results are either not needed
|
|
at all, or needed in some branches of computation but not others. Provided
|
|
that the goal in question satisfies some requirements (e.g. it is pure,
|
|
it cannot fail etc), it either deletes the goal or moves it to the
|
|
computation branches where its output is needed.
|
|
|
|
<li> elimination of dead procedures (dead_proc_elim.m). Inlining, higher-order
|
|
specialization and the elimination of unused args can make procedures dead
|
|
even the user doesn't, and automatically constructed unification and
|
|
comparison predicates are often dead as well.
|
|
|
|
<li> conversion of Aditi procedures into disjunctive normal form (dnf.m).
|
|
The supplementary magic sets and context transformations are only defined
|
|
for predicates in DNF.
|
|
|
|
<li> supplementary magic sets or supplementary context transformation of
|
|
Aditi procedures (magic.m, magic_util.m, context.m).
|
|
The magic sets or context transformations must be applied to convert the
|
|
program to a form for which Aditi-RL bytecode can be generated.
|
|
|
|
<li> reducing the number of variables that have to be saved across
|
|
procedure calls (saved_vars.m). We do this by putting the code that
|
|
generates the value of a variable just before the use of that variable,
|
|
duplicating the variable and the code that produces it if necessary,
|
|
provided the cost of doing so is smaller than the cost of saving and
|
|
restoring the variable would be.
|
|
|
|
</ul>
|
|
|
|
<p>
|
|
|
|
The module transform.m contains stuff that is supposed to be useful
|
|
for high-level optimizations (but which is not yet used).
|
|
|
|
<p>
|
|
|
|
The last HLDS-to-HLDS transformation implements deep profiling
|
|
(deep_profiling.m).
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> a. LLDS BACK-END </h3>
|
|
|
|
<h4> 4a. Code generation. </h4>
|
|
|
|
<p>
|
|
|
|
<dl>
|
|
<dt> pre-passes to annotate the HLDS
|
|
|
|
<dd>
|
|
Before code generation there are a few more passes which
|
|
annotate the HLDS with information used for code generation:
|
|
|
|
<dl>
|
|
<dt> choosing registers for procedure arguments (arg_info.m)
|
|
<dd>
|
|
Currently uses one of two simple algorithms, but
|
|
we may add other algorithms later.
|
|
<dt> annotation of goals with liveness information (liveness.m)
|
|
<dd>
|
|
This records the birth and death of each variable
|
|
in the HLDS goal_info.
|
|
<dt> allocation of stack slots
|
|
<dd>
|
|
This is done by live_vars.m, which works
|
|
out which variables need to be saved on the
|
|
stack when (trace.m determines what variables
|
|
are needed for debugging purposes).
|
|
It then uses graph_colour.m to determine
|
|
a good allocation of variables to stack slots.
|
|
<dt> migration of builtins following branched structures
|
|
<dd>
|
|
This transformation, which is performed by
|
|
follow_code.m, improves the results of follow_vars.
|
|
<dt> allocating the follow vars (follow_vars.m)
|
|
<dd>
|
|
Traverses backwards over the HLDS, annotating some
|
|
goals with information about what locations variables
|
|
will be needed in next. This allows us to generate
|
|
more efficient code by putting variables in the right
|
|
spot directly. This module is not called from
|
|
mercury_compile.m; it is called from store_alloc.m.
|
|
<dt> allocating the store map (store_alloc.m)
|
|
<dd>
|
|
Annotates each branched goal with variable location
|
|
information so that we can generate correct code
|
|
by putting variables in the same spot at the end
|
|
of each branch.
|
|
<dt> computing goal paths (goal_path.m)
|
|
<dd>
|
|
The goal path of a goal defines its position in
|
|
the procedure body. This transformation attaches
|
|
its goal path to every goal, for use by the debugger.
|
|
</dl>
|
|
|
|
<dt> code generation
|
|
|
|
<dd>
|
|
Code generation converts HLDS into LLDS.
|
|
For the LLDS back-end, this is also the point at which we
|
|
insert code to handle debugging and trailing, and to do
|
|
heap reclamation on failure.
|
|
The main code generation module is code_gen.m.
|
|
It handles conjunctions and negations, but calls sub-modules
|
|
to do most of the other work:
|
|
|
|
<ul>
|
|
<li> ite_gen.m (if-then-elses)
|
|
<li> call_gen.m (predicate calls and also calls to
|
|
out-of-line unification procedures)
|
|
<li> disj_gen.m (disjunctions)
|
|
<li> par_conj.m (parallel conjunctions)
|
|
<li> unify_gen.m (unifications)
|
|
<li> switch_gen.m (switches), which has sub-modules
|
|
<ul>
|
|
<li> dense_switch.m
|
|
<li> lookup_switch.m
|
|
<li> string_switch.m
|
|
<li> tag_switch.m
|
|
<li> switch_util.m (also used by MLDS back-end)
|
|
</ul>
|
|
<li> commit_gen.m (commits)
|
|
<li> pragma_c_gen.m (embedded C code)
|
|
</ul>
|
|
<p>
|
|
|
|
It also calls middle_rec.m to do middle recursion optimization.
|
|
|
|
<p>
|
|
|
|
The code generation modules make use of
|
|
<dl>
|
|
<dt> code_info.m
|
|
<dd>
|
|
The main data structure for the code generator.
|
|
<dt> code_exprn.m
|
|
<dd>
|
|
This defines the exprn_info type, which is
|
|
a sub-component of the code_info data structure
|
|
which holds the information about
|
|
the contents of registers and
|
|
the values/locations of variables.
|
|
<dt> exprn_aux.m
|
|
<dd>
|
|
Various preds which use exprn_info.
|
|
<dt> code_util.m
|
|
<dd>
|
|
Some miscellaneous preds used for code generation.
|
|
<dt> code_aux.m
|
|
<dd>
|
|
Some miscellaneous preds which, unlike those in
|
|
code_util, use code_info.
|
|
<dt> continuation_info.m
|
|
<dd>
|
|
For accurate garbage collection, collects
|
|
information about each live value after calls,
|
|
and saves information about procedures.
|
|
<dt> trace.m
|
|
<dd>
|
|
Inserts calls to the runtime debugger.
|
|
</dl>
|
|
|
|
<dt> code generation for `pragma export' declarations (export.m)
|
|
<dd> This is handled seperately from the other parts of code generation.
|
|
mercury_compile.m calls the procedures `export__produce_header_file'
|
|
and `export__get_pragma_exported_procs' to produce C code fragments
|
|
which declare/define the C functions which are the interface stubs
|
|
for procedures exported to C.
|
|
|
|
<dt> generation of constants for RTTI data structures
|
|
<dd> This could also be considered a part of code generation,
|
|
but for the LLDS back-end this is currently done as part
|
|
of the output phase (see below).
|
|
</dl>
|
|
|
|
<p>
|
|
|
|
The result of code generation is the Low Level Data Structure (llds.m),
|
|
which may also contains some data structures whose types are defined in rtti.m.
|
|
The code for each procedure is generated as a tree of code fragments
|
|
which is then flattened (tree.m).
|
|
|
|
<p>
|
|
|
|
<h4> 5a. Low-level optimization (LLDS). </h4>
|
|
|
|
<p>
|
|
|
|
The various LLDS-to-LLDS optimizations are invoked from optimize.m.
|
|
They are:
|
|
|
|
<ul>
|
|
<li> optimization of jumps to jumps (jumpopt.m)
|
|
|
|
<li> elimination of duplicate code sequences (dupelim.m)
|
|
|
|
<li> optimization of stack frame allocation/deallocation (frameopt.m)
|
|
|
|
<li> filling branch delay slots (delay_slot.m)
|
|
|
|
<li> dead code and dead label removal (labelopt.m)
|
|
|
|
<li> peephole optimization (peephole.m)
|
|
|
|
<li> introduction of local C variables (use_local_vars.m) <br>
|
|
|
|
</ul>
|
|
|
|
<p>
|
|
|
|
Several of these optimizations (frameopt and use_local_vars)
|
|
use livemap.m, a module that finds the set of locations live at each label.
|
|
|
|
Use_local_vars numbering also introduces
|
|
references to temporary variables in extended basic blocks
|
|
in the LLDS representation of the C code.
|
|
The transformation to insert the block scopes
|
|
and declare the temporary variables is performed by wrap_blocks.m.
|
|
|
|
<p>
|
|
|
|
Depending on which optimization flags are enabled,
|
|
optimize.m may invoke many of these passes multiple times.
|
|
|
|
<p>
|
|
|
|
Some of the low-level optimization passes use basic_block.m,
|
|
which defines predicates for converting sequences of instructions to
|
|
basic block format and back, as well as opt_util.m, which contains
|
|
miscellaneous predicates for LLDS-to-LLDS optimization.
|
|
|
|
<p>
|
|
|
|
<h4> 6a. Output C code </h4>
|
|
|
|
<ul>
|
|
<li> type_ctor_info.m generates the type_ctor_gen_info structures that list
|
|
items of information (including unification, index and compare predicates)
|
|
associated with each declared type constructor that go into the static
|
|
type_ctor_info data structure. If the type_ctor_gen_info structure is not
|
|
eliminated as inaccessible, this module adds the corresponding type_ctor_info
|
|
structure to the RTTI data structures defined in rtti.m,
|
|
which are part of the LLDS.
|
|
|
|
<li> base_typeclass_info.m generates the base_typeclass_info structures that
|
|
list the methods of a class for each instance declaration. These are added to
|
|
the RTTI data structures, which are part of the LLDS.
|
|
|
|
<li> stack_layout.m generates the stack_layout structures for
|
|
accurate garbage collection. Tables are created from the data
|
|
collected in continuation_info.m.
|
|
|
|
Stack_layout.m uses prog_rep.m and static_term.m to generate representations
|
|
of procedure bodies for use by the declarative debugger.
|
|
|
|
<li> Type_ctor_info structures and stack_layout structures both contain
|
|
pseudo_type_infos, which are type_infos with holes for type variables;
|
|
these are generated by pseudo_type_info.m.
|
|
|
|
<li> llds_common.m extracts static terms from the main body of the LLDS, and
|
|
puts them at the front. If a static term originally appeared several times,
|
|
it will now appear as a single static term with multiple references to it.
|
|
|
|
<li> transform_llds.m is responsible for doing any source to source
|
|
transformations on the llds which are required to make the C output
|
|
acceptable to various C compilers. Currently computed gotos can have
|
|
their maximum size limited to avoid a fixed limit in lcc.
|
|
|
|
<li> Final generation of C code is done in llds_out.m, which subcontracts the
|
|
output of RTTI structures to rtti_out.m.
|
|
</ul>
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> b. MLDS BACK-END </h3>
|
|
|
|
The original LLDS code generator generates very low-level code,
|
|
since the LLDS was designed to map easily to RISC architectures.
|
|
We're currently developing a new back-end that generates much higher-level
|
|
code, suitable for generating Java, high-level C, etc.
|
|
This back-end uses the Medium Level Data Structure (mlds.m) as its
|
|
intermediate representation.
|
|
|
|
<h4> 3b. pre-passes to annotate/transform the HLDS </h4>
|
|
|
|
Before code generation there is a pass which annotates the HLDS with
|
|
information used for code generation:
|
|
|
|
<ul>
|
|
<li> mark_static_terms.m marks construction unifications
|
|
which can be implemented using static constants rather
|
|
than heap allocation.
|
|
</ul>
|
|
|
|
For the MLDS back-end, we've tried to keep the code generator simple.
|
|
So we prefer to do things as HLDS to HLDS transformations where possible,
|
|
rather than complicating the HLDS to MLDS code generator.
|
|
So we have a pass which transforms the HLDS to handle trailing:
|
|
|
|
<ul>
|
|
<li> add_trail_ops.m inserts code to manipulate the trail,
|
|
in particular ensuring that we apply the appropriate
|
|
trail operations before each choice point, when execution
|
|
resumes after backtracking, and whenever we do a commit.
|
|
The trail operations are represented as (and implemented as)
|
|
calls to impure procedures defined in library/private_builtin.m.
|
|
</ul>
|
|
|
|
<h4> 4b. MLDS code generation </h4>
|
|
<ul>
|
|
<li> ml_code_gen.m converts HLDS code to MLDS.
|
|
The following sub-modules are used to handle different constructs:
|
|
<dl>
|
|
<dt> ml_unify_gen.m
|
|
<dt> ml_call_gen.m
|
|
<dt> ml_switch_gen.m, which in turn has sub-modules
|
|
<ul>
|
|
<li> ml_dense_switch.m
|
|
<li> ml_string_switch.m
|
|
<li> ml_tag_switch.m
|
|
<li> switch_util.m (also used by MLDS back-end)
|
|
</ul>
|
|
<dl>
|
|
The module ml_code_util.m provides utility routines for
|
|
MLDS code generation. The module ml_util.m provides some
|
|
general utility routines for the MLDS.
|
|
<li> ml_type_gen.m converts HLDS types to MLDS.
|
|
<li> type_ctor_info.m and base_typeclass_info.m generate
|
|
the RTTI data structures defined in rtti.m and pseudo_type_info.m
|
|
(those four modules are shared with the LLDS back-end)
|
|
and then mlds_to_rtti.m converts these to MLDS.
|
|
</ul>
|
|
|
|
<h4> 5b. MLDS transformations </h4>
|
|
<ul>
|
|
<li> ml_tailcall.m annotates the MLDS with information about tailcalls.
|
|
<li> ml_optimize.m does MLDS->MLDS optimizations
|
|
<li> ml_elim_nested.m transforms the MLDS to eliminate nested functions.
|
|
</ul>
|
|
|
|
<h4> 6b. MLDS output </h4>
|
|
|
|
There are currently two backends that generate code from MLDS, one
|
|
generates C/C++ code, the other generates Microsoft's Intermediate
|
|
Language (MSIL or IL).
|
|
<p>
|
|
<ul>
|
|
<li>mlds_to_c.m converts MLDS to C/C++ code.
|
|
</ul>
|
|
<p>
|
|
The MLDS->IL backend is broken into several submodules.
|
|
<ul>
|
|
<li> mlds_to_ilasm.m converts MLDS to IL assembler and writes it to a .il file.
|
|
<li> mlds_to_il.m converts MLDS to IL
|
|
<li> ilds.m contains representations of IL
|
|
<li> ilasm.m contains output routines for writing IL to assembler.
|
|
<li> il_peephole.m performs peephole optimization on IL instructions.
|
|
</ul>
|
|
After IL assembler has been emitted, ILASM in invoked to turn the .il
|
|
file into a .dll or .exe.
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> c. Aditi-RL BACK-END </h3>
|
|
|
|
<h4> 4c. Aditi-RL generation </h4>
|
|
|
|
<ul>
|
|
<li> rl_gen.m converts HLDS to RL.
|
|
|
|
<li> rl_relops.m generates RL for relational operations such as joins.
|
|
|
|
<li> rl_info.m defines a state type.
|
|
|
|
<li> rl.m defines the the representation of Aditi-RL used within the
|
|
Mercury compiler. There are some slight differences between rl.m
|
|
and Aditi-RL to make optimization easier.
|
|
|
|
<li> rl_dump.m contains predicates to write the types defined in rl.m.
|
|
</ul>
|
|
|
|
<h4> 5c. Aditi-RL optimization </h4>
|
|
|
|
<ul>
|
|
<li> rl_opt.m invokes the RL optimization passes.
|
|
|
|
<li> rl_block.m converts an RL procedure into basic blocks, and performs
|
|
other tasks such as detecting the loops in those basic blocks.
|
|
|
|
<li> rl_analyse.m contains a generic data-flow analysis procedure for
|
|
RL procedures.
|
|
|
|
<li> rl_liveness.m uses rl_analyse.m to insert code to initialise relations
|
|
and clear references to them when they are no longer needed.
|
|
|
|
<li> rl_loop.m moves invariants out of loops.
|
|
|
|
<li> rl_block_opt.m performs common subexpression elimination and instruction
|
|
merging on basic blocks.
|
|
|
|
<li> rl_key.m computes upper and lower bounds on the non-locals of a goal
|
|
for which the goal could succeed, which is useful for determining when
|
|
an index could be used for a relational operation.
|
|
|
|
<li> rl_sort.m introduces sort-merge and indexed versions of relational
|
|
operations where possible, and optimizes away unnecessary sorting and
|
|
indexing.
|
|
|
|
<li> rl_stream.m detects relations which do not need to be materialised.
|
|
</ul>
|
|
|
|
<h4> 6c. Output Aditi-RL code </h4>
|
|
|
|
<ul>
|
|
<li> rl_out.m converts from the instructions defined in rl.m
|
|
to bytecode either as data in the <module>.c file or
|
|
to <module>.rlo and outputs a text representation to
|
|
<module>.rla.
|
|
|
|
<li> rl_exprn.m is called from rl_out.m to convert top down Mercury
|
|
code (in HLDS form) to Aditi bytecode.
|
|
|
|
<li> rl_code.m contains the definition of the bytecodes interpreted
|
|
by Aditi. This file is automatically generated from the Aditi sources.
|
|
|
|
<li> rl_file.m contains routines to output the bytecodes defined in rl_code.m.
|
|
</ul>
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> d. BYTECODE BACK-END </h3>
|
|
|
|
<p>
|
|
|
|
The Mercury compiler can translate Mercury programs into bytecode for
|
|
interpretation by a bytecode interpreter. The intent of this is to
|
|
achieve faster turn-around time during development. However, the
|
|
bytecode interpreter has not yet been written.
|
|
|
|
<ul>
|
|
<li> bytecode.m defines the internal representation of bytecodes, and contains
|
|
the predicates to emit them in two forms. The raw bytecode form is emitted
|
|
into <filename>.bytecode for interpretation, while a human-readable
|
|
form is emitted into <filename>.bytedebug for visual inspection.
|
|
|
|
<li> bytecode_gen.m contains the predicates that translate HLDS into bytecode.
|
|
|
|
<li> bytecode_data.m contains the predicates that translate ints, strings
|
|
and floats into bytecode. This is also used by rl_code.m.
|
|
</ul>
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> SMART RECOMPILATION </h3>
|
|
|
|
<p>
|
|
|
|
The Mercury compiler can record program dependency information
|
|
to avoid unnecessary recompilations when an imported module's
|
|
interface changes in a way which does not invalidate previously
|
|
compiled code.
|
|
|
|
<ul>
|
|
<li> recompilation.m contains types used by the other smart
|
|
recompilation modules.
|
|
|
|
<li> recompilation_version.m generates version numbers for program items
|
|
in interface files.
|
|
|
|
<li> recompilation_usage.m works out which program items were used
|
|
during a compilation.
|
|
|
|
<li> recompilation_check.m is called before recompiling a module.
|
|
It uses the information written by recompilation_version.m and
|
|
recompilation_usage.m to work out whether the recompilation is
|
|
actually needed.
|
|
|
|
<li> timestamp.m contains an ADT representing timestamps used
|
|
by smart recompilation.
|
|
</ul>
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> MISCELLANEOUS </h3>
|
|
|
|
<dl>
|
|
<dt> builtin_ops:
|
|
<dd>
|
|
This module defines the types unary_op and binary_op
|
|
which are used by several of the different back-ends:
|
|
bytecode.m, llds.m, and mlds.m.
|
|
|
|
<dt> c_util:
|
|
<dd>
|
|
This module defines utility routines useful for generating
|
|
C code. It is used by both llds_out.m and mlds_to_c.m.
|
|
|
|
<dt> det_util:
|
|
<dd>
|
|
This module contains utility predicates needed by the parts
|
|
of the semantic analyzer and optimizer concerned with
|
|
determinism.
|
|
|
|
<dt> special_pred.m, unify_proc.m:
|
|
<dd>
|
|
These modules contain stuff for handling the special
|
|
compiler-generated predicates which are generated for
|
|
each type: unify/2, compare/3, and index/1 (used in the
|
|
implementation of compare/3).
|
|
|
|
<dt> dependency_graph.m:
|
|
<dd>
|
|
This contains predicates to compute the call graph for a
|
|
module, and to print it out to a file.
|
|
(The call graph file is used by the profiler.)
|
|
The call graph may eventually also be used by det_analysis.m,
|
|
inlining.m, and other parts of the compiler which could benefit
|
|
from traversing the predicates in a module in a bottom-up or
|
|
top-down fashion with respect to the call graph.
|
|
|
|
<dt> passes_aux.m
|
|
<dd>
|
|
Contains code to write progress messages, and higher-order
|
|
code to traverse all the predicates defined in the current
|
|
module and do something with each one.
|
|
|
|
<dt> opt_debug.m:
|
|
<dd>
|
|
Utility routines for debugging the LLDS-to-LLDS optimizations.
|
|
|
|
<dt> error_util.m:
|
|
<dd>
|
|
Utility routines for printing nicely formatted error messages.
|
|
</dl>
|
|
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
<h3> CURRENTLY USELESS </h3>
|
|
|
|
<p>
|
|
|
|
<dl>
|
|
|
|
<dt> lco.m:
|
|
<dd>
|
|
This finds predicates whose implementations would benefit
|
|
from last call optimization modulo constructor application.
|
|
It does not apply the optimization and will not until the
|
|
mode system is capable of expressing definite aliasing.
|
|
|
|
</dl>
|
|
|
|
<p>
|
|
<hr>
|
|
<!---------------------------------------------------------------------------->
|
|
|
|
Last update was $Date: 2001-08-11 14:09:51 $ by $Author: stayl $@cs.mu.oz.au. <br>
|
|
</body>
|
|
</html>
|