mirror of
https://github.com/Mercury-Language/mercury.git
synced 2026-04-24 05:43:53 +00:00
dadf30718d6084962be37dae8b94de41aaee90e2
115 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
Zoltan Somogyi
|
5f50259d16 |
Write to explicitly named streams in many modules.
Right now, most parts of the compiler write to the "current output stream".
This was a pragmatic choice at the time, but has not aged well. The problem
is that the answer to the question "where is the current output stream going?"
is not obvious in *all* places in the compiler (although it is obvious in
most). When using such implicit streams, finding where the output is going
to in a given predicate requires inspecting not just the ancestors of that
predicate, but also all their older siblings (since any of them could have
changed the current stream), *including* their entire call trees. This is
usually an infeasible task. By constrast, if we explicitly pass streams
to all output operations, we need only follow the places where the variable
representing that stream is bound, which the mode system makes easy.
This diff switches large parts of the compiler over to doing output only
to explicitly passed streams, never to the implicit "current output stream".
The parts it switches over are the parts that rely to a significant degree
on the innermost change, which is to the "output" typeclass in
parse_tree_out_info.m. This is the part that has to be switched over to
explicit streams first, because (a) many modules such as mercury_to_mercury.m
rely on the output typeclass, and (b) most other modules that do output
call predicates in these modules. Starting anywhere else would be like
building a skyscraper starting at the top.
This typeclass, output(U), has two instances: output(io), and output(string),
so you could output either to the current output stream, or to a string.
To allow the specification of the destination stream in the first case,
this diff changes the typeclass to output(S, U) with a functional dependency
from U to S, with the two instances being output(io.text_output_stream, io)
and output(unit, string). (The unit arg is ignored in the second case.)
There is a complication with the output typeclass method, add_list, that
outputs a list of items. The complication is that each item is output
by a predicate supplied by the caller, but the separator between the items
(usually a comma) is output by add_list itself. We don't want to give
callers of this method the opportunity to screw up by specifying (possibly
implicitly) two different output streams for these two purposes, so we want
(a) the caller to tell add_list where to put the separators, and then
(b) for add_list, not its caller, tell the user-supplied predicate what
stream to write to. This works only if the stream argument is just before
the di,uo pair of I/O state arguments, which differs from our usual practice
of passing the stream at or near the left edge of the argument list,
not near the right. The result of this complication is that two categories
of predicates that are and are not used to print items in a list differ
in where they put the stream in their argument lists. This makes it easy
to pass the stream in the wrong argument position if you call a predicate
without looking up its signature, and may require *changing* the argument
order when a predicate is used to print an item in a list for the first time.
A complete switch over to always passing the stream just before !IO
would fix this inconsistency, but is far to big a change to make all at once.
compiler/parse_tree_out_info.m:
Make the changes described above.
Add write_out_list, which is a variant of io.write_list specifically
designed to address the "complication" described above. It also has
the arguments in an order that is better suited for higher-order use.
Make the same change to argument order in the class method add_list
as well.
Almost all of the following changes consist of passing an extra stream
argument to output predicates. In some places, where I thought this would
aid readability, I replaced sequences of calls to output predicates
with a single io.format.
compiler/prog_out.m:
This module had many predicates that wrote things to the current output
stream. This diff adds versions of these predicates that take an
explicit stream argument.
If the originals are still needed after the changes to the other modules,
keep them, but add "_to_cur_stream" to the end of their names.
Otherwise, delete them. (Many of the changes below replace
write_xyz(..., !IO) with io.write_string(Stream, xyz_to_string(...), !IO),
especially when write_xyz did nothing except call xyz_to_string
and wrote out the result.)
compiler/c_util.m:
Add either an explicit stream argument to the argument list, or a
"_current_stream" suffix to the name, of every predicate defined
in this module that does output.
Add a new predicate to print out the block comment containing
input for mkinit. This factors out common code in the LLDS and MLDS
backends.
compiler/name_mangle.m:
Delete all predicates that used to write to the current output stream,
after replacing them if necessary with functions that return a string,
which the caller can print to wherever it wants. (The "if necessary"
part is there because some of the "replacement" functions already
existed.)
When converting a proc_label to a string, *always* require the caller
to say whether the label prefix should be added to the string,
instead of silently assuming "yes, add it", as calls to one of the old,
now deleted predicates had it.
compiler/file_util.m:
Add output_to_file_stream, a version of output_to_file which
simply passes the output file stream it opens to the predicate
that is intended to define the contents of the newly created or
updated file. The existing output_to_file, which instead sets
and resets the current output stream around the equivalent
predicate call, is still needed e.g. by the MLDS backend,
but hopefully for not too long.
compiler/mercury_to_mercury.m:
compiler/parse_tree_out.m:
compiler/parse_tree_out_clause.m:
compiler/parse_tree_out_inst.m:
compiler/parse_tree_out_pragma.m:
compiler/parse_tree_out_pred_decl.m:
compiler/parse_tree_out_term.m:
compiler/parse_tree_out_type_repn.m:
Change the code writing out parse trees to explicitly pass a stream
to every predicate that does output.
In some places, this allows us to avoid changing the identity
of the current output stream.
compiler/hlds_out.m:
compiler/hlds_out_goal.m:
compiler/hlds_out_mode.m:
compiler/hlds_out_module.m:
compiler/hlds_out_pred.m:
compiler/hlds_out_util.m:
compiler/intermod.m:
Change the code writing out HLDS code to explicitly pass a stream
to every predicate that does output. (The changes to these modules
belong in this diff because these modules call many of the output
predicates in the parse tree package.)
In hlds_out_util.m, delete some write_to_xyz(...) predicates that wrote
the result of xyz_to_string(...) to the current output stream.
Replace calls to the deleted predicates with calls to io.write_string
with the string being written being computed by xyz_to_string.
Add a predicate to hlds_out_util.m that outputs a comment containing
the current context, if it is valid. This factors out code that used
to be common to several of the other modules.
In a few places in hlds_out_module.m, the new code generates a
slighly different set of blank lines, but this should not be a problem.
compiler/layout_out.m:
compiler/llds_out_code_addr.m:
compiler/llds_out_data.m:
compiler/llds_out_file.m:
compiler/llds_out_global.m:
compiler/llds_out_instr.m:
compiler/llds_out_util.m:
compiler/opt_debug.m:
compiler/rtti_out.m:
Change the code writing out the LLDS to explicitly pass a stream
to every predicate that does output. (The changes to these modules
belong in this diff because layout_out.m and rtti_out.m call
many of the output predicates in the parse tree package,
and through them, the rest of the LLDS backend is affected as well.)
compiler/make.module_dep_file.m:
compiler/mercury_compile_main.m:
compiler/mercury_compile_middle_passes.m:
Replace code that sets and resets the current output stream
with code that simply passes an explicit output stream to a
predicate that now *takes* an explicit stream as an argument.
compiler/accumulator.m:
compiler/add_clause.m:
compiler/code_gen.m:
compiler/code_loc_dep.m:
compiler/cse_detection.m:
compiler/delay_partial_inst.m:
compiler/dep_par_conj.m:
compiler/det_analysis.m:
compiler/error_msg_inst.m:
compiler/export.m:
compiler/format_call.m:
compiler/goal_expr_to_goal.m:
compiler/ite_gen.m:
compiler/lco.m:
compiler/liveness.m:
compiler/lp_rational.m:
compiler/mercury_compile_front_end.m:
compiler/mercury_compile_llds_back_end.m:
compiler/mlds_to_c_file.m:
compiler/mlds_to_c_global.m:
compiler/mode_debug.m:
compiler/mode_errors.m:
compiler/modes.m:
compiler/optimize.m:
compiler/passes_aux.m:
compiler/pd_debug.m:
compiler/pragma_c_gen.m:
compiler/proc_gen.m:
compiler/prog_ctgc.m:
compiler/push_goals_together.m:
compiler/rat.m:
compiler/recompilation.m:
compiler/recompilation.usage.m:
compiler/recompilation.version.m:
compiler/rtti.m:
compiler/saved_vars.m:
compiler/simplify_goal_conj.m:
compiler/stack_opt.m:
compiler/structure_reuse.analysis.m:
compiler/structure_reuse.domain.m:
compiler/structure_reuse.indirect.m:
compiler/structure_sharing.analysis.m:
compiler/superhomogeneous.m:
compiler/term_constr_build.m:
compiler/term_constr_data.m:
compiler/term_constr_fixpoint.m:
compiler/term_constr_pass2.m:
compiler/term_constr_util.m:
compiler/tupling.m:
compiler/type_assign.m:
compiler/unneeded_code.m:
compiler/write_deps_file.m:
Conform to the changes above, mostly by passing streams explicitly.
compiler/hlds_dependency_graph.m:
Conform to the changes above, mostly by passing streams explicitly.
Move a predicate's definition next it only use.
compiler/Mercury.options:
Specify --warn-implicit-stream-calls for all the modules in which
this diff has replaced all implicit streams with explicit streams.
(Unfortunately, debugging this diff has shown that --warn-implicit-
stream-calls detects only *some*, and not *all*, uses of implicit
streams.)
library/term_io.m:
Fix documentation.
|
||
|
Zoltan Somogyi
|
91560d2dd7 |
Make --everything-in-one-c-function a bool option.
It used to be a special option that just set procs_per_c_function
to the special value of zero, but this behavior is inconsistent with
optimization_options taking the max of the old and new values of
integer options. This meant that --procs-per-c-function=5 -O6
would not put all procedures into one C function.
tools/make_optimization_options_db:
tools/make_optimization_options_end:
compiler/optimization_options.m:
As above: make --everything-in-one-c-function a bool option,
named use_just_one_c_func for brevity.
compiler/options.m:
Move the code handling --everything-in-one-c-function next to the
other optimization options.
compiler/mercury_compile_llds_back_end.m:
When deciding which procedures should be put into which C functions,
use the logic: if use_just_one_c_func, then put all procs into the
same C function, otherwise, put them into one or more C functions
with up to procs_per_c_function procedures per function.
This is now the *only* place in the compiler that looks at
procs_per_c_function; everywhere else looks at use_just_one_c_func.
compiler/code_util.m:
Replace a maybe pair type that used to store the value of
procs_per_c_function with a bespoke type storing use_just_one_c_func,
which documents its meaning.
compiler/code_info.m:
Replace a bool type with a bespoke type, again documenting
its meaning.
compiler/call_gen.m:
compiler/closure_gen.m:
compiler/middle_rec.m:
compiler/pragma_c_gen.m:
compiler/proc_gen.m:
compiler/rtti_out.m:
Conform to the changes above.
compiler/llds.m:
Fix blank lines.
|
||
|
Zoltan Somogyi
|
9cbe5d2caf |
Put type_repn items for complex types into .int files.
compiler/decide_type_repn.m:
Previously, this module computed type_repn items to put into .int3 files
for a subset of the type constructors defined in the current module:
the direct_dummy, enum and notag types (the *simple* types),
and the du types whose representation is guaranteed to be
a word-aligned pointer when targeting C. (We care about pointers
being word-aligned only when applying the direct arg optimization.
This optimization is applicable only with the low level data
representation, which we use only when targeting C.)
This diff adds code to decide the representations of *all* the
type constructors defined in the current module.
This code is based on the existing code in du_type_layout.m,
which it is intended to eventually replace, but its job is more general,
because it decides the representation of each type not just for
one platform (the one we want to generate code), but for all possible
platforms. This is because we want to put the descriptions of type
representations into the module's .int file to serve as a single source
of truth for all modules that use the types defined in this module,
and the contents of .int files should be platform-independent.
For our purposes, there are six kinds of platforms, which are
distinguished along three axes: 64 vs 32 bit machines, spf vs non-spf
grades, and direct arg optimization enabled vs disabled. That is eight
combinations, but on 64 bit machines, a float takes up one word whether
that float is single or double precision, so two combinations aren't valid.
Some of the change to this module consists of generalizing the existing
code so that it can decide simple types not just when targeting .int3 files
but .int files as well. However, the bulk of it is code for deciding
the representations of non-simple types. The code is not lifted straight
from du_type_layout.m. There are two main kinds of changes.
First, I took the opportunity to simplify the algorithms used.
For example, while du_type_layout.m passes over each function symbol
in the most general kind of type twice: once to assign it a cons_tag,
and once to decide how to pack its arguments, the code here does both jobs
in one pass. Another example is that for historical reasons,
du_type_layout.m computed the amount of space needed for an argument
in one place for sub-word-sized arguments, and in another place
for more-than-word-sized arguments; decide_type_repn.m does it all
in one place.
Second, since we compute a representation for each type six times,
I tried to avoid obvious inefficiencies, but only if the code
remained simple. In the future, we may want to use an approach
based on the idea that in the process of computing the first
representation, we look out for any indication that the representation
may be different on any of the other five platforms, and if not,
we just reuse the first representation on the other five platforms as well.
However, that would be appropriate only *after* we have a simpler
system that has proven to work in practice.
There is a third, smaller change: when deciding whether an argument
is packable, we take into account not just equivalence type
definitions, but the definitions of notag types as well.
This takes advantage of the fact that if a notag type is abstract
exported, its representation is put into the relevant .int3 file
even though its definition isn't. (This is why du_type_layout.m
couldn't "see through" notag types: it couldn't depend on knowing
which types were notags.)
compiler/prog_item.m:
Change the types we use for type representation information.
Their previous definitions baked in the assumption that the only
distinction between platforms that mattered was the 64 vs 32 bit
distinction, which is not the case.
Use a more consistent naming scheme for the types we use
to represent type representation information.
Include the "dereferenced" types of the arguments in functors'
representations. (I use "dereferencing" here to mean expanding
equivalence types and throwing away any notag wrappers.).
We don't need it when generating C code using the low level
data representation, but we do need it to create constructors
when generating e.g. Java code that uses the high level data
representation.
compiler/parse_type_repn.m:
Rewrite most of this module due to the changes in prog_item.m.
compiler/parse_tree_out_type_repn.m:
A new module containing the code for writing out type representations.
The original code used to be in parse_tree_out.m, but it has been
mostly rewritten. Partly this is due the changes in prog_item.m,
but partly it is to provide much more structured output for humans,
since this makes debugging so much easier.
compiler/parse_tree.m:
Add the new module to the parse_tree package.
compiler/parse_tree_out.m:
Delete the code moved to parse_tree_out_type_repn.m.
compiler/parse_tree_out_info.m:
Provide a mechanism for selecting between output for machines
(the default) and output for humans.
compiler/hlds_data.m:
compiler/prog_data.m:
Move the ptag type from hlds_data.m to prog_data.m, to make it
accessible in prog_item.m.
Add some documentation in prog_data.m.
compiler/comp_unit_interface.m:
If the experiment1 option is enabled, invoke decide_type_repn.m
to decide what type_repn items to put into the .int file we are
generating. Otherwise, maintain the status quo.
compiler/write_module_interface_files.m:
Pass the globals to comp_unit_interface.m so it can look up experiment1.
compiler/equiv_type.m:
Add a predicate for expanding equivalence types for use by
decide_type_repn.m. This predicate expands just one type,
but reports any use of circular equivalence types in that type.
Improve the error message for circular equivalence types by *naming*
the type constructors involved. To make this possible, pass around
sets of such type constructors instead of just a boolean saying
*whether* we have found *some* circular equivalence type.
Replace bools used as changed/unchanged flag with a bespoke type.
Standardize some variable names.
compiler/options.m:
Add the developer-only option --pack-everything, which, if set,
tells decide_type_repn.m to turn on all the packing options
that currently work. This is to allow the testing of decide_type_repn.m
in the eventual intended mode of operation, even if the various
allow-packing-... options used by du_type_layout.m are set to "no".
compiler/disj_gen.m:
compiler/equiv_type_hlds.m:
compiler/llds_out_data.m:
compiler/lookup_util.m:
compiler/ml_call_gen.m:
compiler/ml_closure_gen.m:
compiler/mlds_to_c_data.m:
compiler/rtti.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
compiler/tag_switch.m:
Conform to the changes above (mostly the move of ptag to prog_data.m.)
compiler/parse_pragma.m:
Improve indentation.
tests/valid_make_int/test_repn.m:
tests/valid_make_int/test_repn_sub.m:
A fairly comprehensive test case of the new functionality.
test_repn_sub.m defines one ore more simple type constructors
of each possible kind, and test_repn.m uses them to define types
that use each possible kind of complex type representation.
tests/valid_make_int/Mmakefile:
tests/valid_make_int/Mercury.options:
Enable the new test case.
|
||
|
Zoltan Somogyi
|
4ef4402ecf |
Make --warn-inconsistent-pred-order-clauses default for the compiler.
compiler/COMP_FLAGS.in:
As above.
compiler/Mercury.options:
List the modules for we need --no-warn-inconsistent-pred-order-clauses
for now.
compiler/call_gen.m:
compiler/code_util.m:
compiler/deep_profiling.m:
compiler/equiv_type.m:
compiler/error_util.m:
compiler/exprn_aux.m:
compiler/get_dependencies.m:
compiler/global_data.m:
compiler/layout_out.m:
compiler/liveness.m:
compiler/ll_pseudo_type_info.m:
compiler/llds.m:
compiler/llds_out_code_addr.m:
compiler/llds_out_data.m:
compiler/module_cmds.m:
compiler/module_qual.m:
compiler/module_qual.qualify_items.m:
compiler/opt_debug.m:
compiler/parse_class.m:
compiler/parse_goal.m:
compiler/parse_sym_name.m:
compiler/parse_type_defn.m:
compiler/rtti_out.m:
compiler/stack_layout.m:
compiler/trace_gen.m:
Fix issues reported by --warn-inconsistent-pred-order-clauses
for these modules.
|
||
|
Zoltan Somogyi
|
c1bdd2100b | Delete unneeded $module args from aborts. | ||
|
Zoltan Somogyi
|
d49f6eab84 |
Add missing imports of parent modules.
These imports were missing from source files, but were included in imported modules' .int3 files. An upcoming change will delete these from those .int3 files. |
||
|
Peter Wang
|
5f7d3e6bb2 |
Use consistent integer types for some RTTI fields.
runtime/mercury_type_info.h:
Use unsigned integer types for a few RTTI structure fields that
are known to hold non-negative values.
Add comments for other field types that could be changed later.
compiler/rtti.m:
Use fixed size integer types for fields matching the size
and signedness of the corresponding C RTTI structure fields.
Encode type ctor flags in a uint16 instead of int.
Make type_ctor_details_num_ptags and type_ctor_details_num_functors
return a maybe value, instead of a negative value to represent no
primary tags or no function symbols, respectively.
compiler/type_ctor_info.m:
Conform to type changes.
Use uint16 to represent the "contains var" bit vector.
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
Conform to type changes.
Add comments to make it easier to find the code that writes out
each particular RTTI structure field.
compiler/ml_util.m:
Add helper functions.
compiler/add_special_pred.m:
compiler/du_type_layout.m:
compiler/erl_rtti.m:
compiler/erlang_rtti.m:
compiler/hlds_data.m:
compiler/llds_out_data.m:
compiler/ml_unify_gen_construct.m:
compiler/opt_debug.m:
compiler/pseudo_type_info.m:
compiler/stack_layout.m:
compiler/unify_gen_construct.m:
Conform to type changes.
compiler/parse_type_defn.m:
compiler/prog_data.m:
Use uint32 for functor ordinal numbers.
library/rtti_implementation.m:
Use fixed size integer types for RTTI field accessor functions,
and update callers.
java/runtime/DuArgLocn.java:
java/runtime/DuExistInfo.java:
java/runtime/DuExistLocn.java:
java/runtime/DuFunctorDesc.java:
java/runtime/TypeCtorInfo_Struct.java:
Use integer types in RTTI structure definitions for Java that match
the types in the C versions of the same structures.
runtime/mercury_dotnet.cs.in:
Use integer types in RTTI structure definitions for C# that match
the types in the C versions of the same structures.
|
||
|
Zoltan Somogyi
|
edac0e9b5c |
Move sub-word-sized args next to each other ...
... so they can be packed into the same word. |
||
|
Zoltan Somogyi
|
d1ec3e3f55 |
Pack characters as sub-word-sized arguments.
compiler/options.m:
Add new options --allow-packing-chars and --allow-packing-mini-types,
for use by developers only.
compiler/du_type_layout.m:
If the new --allow-packing-chars option is set, then allow characters
to be packed as sub-word-sized arguments.
Record the value of the --allow-packing-mini-types in the parameters
as well, without using it as yet. (That is for a future change.)
Fix the test for what counts as suboptimal packing. Given an argument
list such as <sub-word-sized args, word-sized arg, sub-word-sized args>
where the two lots of sub-word-sized args could fit into a single word,
the old code did NOT emit a "this is suboptimal" message for it. This
was because it did not consider a reduction from 2n words to 2n-1 words
to be significant, as boehm_gc would always round up the number of words
to (at least) the next multiple of 2. The problem was that the code
used to apply this test ONLY to the number of words occupied by
sub-word-sized args. In this case, better packing would make that number
go from 2 to 1, so it did emit the message, even though the *total*
number of words in the memory cell would go from 3 to 2.
We could fix this by applying the round-up-to-even the test to the total
number of words, but it is simpler to eliminate round-up-to-even entirely.
This is desirable anyway, since we *should* prefer compresssing e.g.
a 4 word cell down to 3 if possible, since this would reduce traffic
to and from (some levels of) cache, even if it would not reduce
the size of the memory allocation.
Fix some too-long lines.
compiler/prog_data.m:
Add fill_char21 as a new kind of fill for sub-word-sized arguments.
(The 21 is a reminder of the 21 bit size of Unicode chars.)
runtime/mercury_type_info.h:
Document that fill_char21 arguments are treated the same way as enum args
in the RTTI.
compiler/hlds_out_module.m:
compiler/ml_unify_gen_deconstruct.m:
compiler/ml_unify_gen_util.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
compiler/unify_gen_util.m:
Conform to the changes above.
|
||
|
Zoltan Somogyi
|
86f563a94d |
Pack subword-sized arguments next to a remote sectag.
compiler/du_type_layout.m:
If the --allow-packing-remote-sectag option is set, then try to pack
an initial subsequence of subword-sized arguments next to remote sectags.
To allow the polymorphism transformation to put the type_infos and/or
typeclass_infos it adds to a function symbol's argument list at the
*front* of that argument list, pack arguments next to remote sectags
only in function symbols that won't have any such extra arguments
added to them.
Do not write all new code for the new optimization; instead, generalize
the code that already does a very similar job for packing args next to
local sectags.
Delete the code we used to have that picked the packed representation
over the base unpacked representation only if it reduced the
"rounded-to-even" number of words. A case could be made for its usefulness,
but in the presence of the new optimization the extra code complexity
it requires is not worth it (in my opinion).
Extend the code that informs users about possible argument order
rearrangements that yield better packing to take packing next to sectags
into account.
compiler/hlds_data.m:
Provide a representation for cons_tags that use the new optimization.
Instead of adding a new cons_tag, we do this by replacing several old
cons_tags that all represent pointers to memory cells with a single
cons_tag named remote_args_tag with an argument that selects among
the old cons_tags being replaced, and adding a new alternative inside
this new type. The new alternative is remote_args_shared with a
remote_sectag whose size is rsectag_subword(...).
Instead of representing the value of the "data" field in classes
on the Java and C# backends as a strange kind of secondary tag
that is added to a memory cell by a class constructor instead of
having to be explicitly added to the front of the argument vector
by the code of a unification, represent it more directly as separate
kind of remote_args_tag. Continuing to treat it as a sectag would have
been very confusing to readers of the code of ml_unify_gen_*.m in the
presence of the new optimization.
Replacing several cons_tags that were usually treated similarly with
one cons_tag simplifies many switches. Instead of an switch with that
branches to the same switch arm for single_functor_tag, unshared_tag
and shared_remote_tag, and then switches on these three tags again
to get e.g. the primary tag of each, the new code of the switch arm
is executed for just cons_tag value (remote_args_tag), and switches
on the various kinds of remote args tags only when it needs to.
In is also more natural to pass around the argument of remote_args_tag
than to pass around a variable of type cons_tag that can be bound to only
single_functor_tag, unshared_tag or shared_remote_tag.
Add an XXX about possible further steps along these lines, such as
making a new cons_tag named something like "user_const_tag" represent
all user-visible constants.
compiler/unify_gen_construct.m:
compiler/unify_gen_deconstruct.m:
compiler/unify_gen_test.m:
compiler/unify_gen_util.m:
compiler/ml_unify_gen_construct.m:
compiler/ml_unify_gen_deconstruct.m:
compiler/ml_unify_gen_test.m:
compiler/ml_unify_gen_util.m:
Implement X = f(Yi) unifications where f uses the new representation,
i.e. some of its arguments are stored next to a remote sectag.
Some of the Yi are stored in a tagword (a word that also contains a tag,
in this case the remote secondary tag), while some are stored in other
words in a memory cell. This means that such unifications have similarities
both to unifications involving arguments being packed next to local
sectags, and to unifications involving ordinary arguments in memory cells.
Therefore wherever possible, their implemenation uses suitably generalized
versions of existing code that did those two jobs for two separate kinds of
cons_tags.
Making such generalizations possible in some cases required shifting the
boundary between predicates, moving work from a caller to a callee
or vice versa.
In unify_gen_deconstruct.m, stop using uni_vals to represent *either* a var
*or* a word in a memory cell. While this enabled us to factor out some
common code, the predicate boundaries it lead to are unsuitable for the
generalizations we now need.
Consistently use unsigned ints to represent both the whole and the parts
of words containing packed arguments (and maybe sectags), except when
comparing ptag constants with the result of applying the "tag" unop
to a word, (since that unop returns an int, at least for now).
In a few cases, avoid the recomputation of some information that we
already know. The motivation is not efficiency, since the recomputation
we avoid is usually cheap, but the simplification of the code's correctness
argument.
Use more consistent terminology in things such as variable names.
Note the possibility of further future improvements in several places.
compiler/ml_foreign_proc_gen.m:
Delete a long unused predicate.
compiler/mlds.m:
Add an XXX documenting a possible improvement.
compiler/rtti.m:
Update the compiler's internal representation of RTTI data structures
to make them able to describe secondary tags that are smaller than
a full word.
compiler/rtti_out.m:
Conform to the changes above, and delete a long-unused predicate.
compiler/type_ctor_info.m:
Use the RTTI's du_hl_rep to represent cons_tags that distinguish
between function symbols using a field in a class.
compiler/ml_type_gen.m:
Provide a specialized form of a function for code in ml_unify_gen_*.m.
Conform to the changes above.
compiler/add_special_pred.m:
compiler/bytecode_gen.m:
compiler/export.m:
compiler/hlds_code_util.m:
compiler/lco.m:
compiler/ml_closure_gen.m:
compiler/ml_switch_gen.m:
compiler/ml_tag_switch.m:
compiler/rtti_to_mlds.m:
compiler/switch_util.m:
compiler/tag_switch.m:
Conform to the changes above.
runtime/mercury_type_info.h:
Update the runtime's representation of RTTI data structures to make them
able to describe remote secondary tags that are smaller than a full word.
runtime/mercury_deconstruct.[ch]:
runtime/mercury_deconstruct.h:
runtime/mercury_deconstruct_macros.h:
runtime/mercury_ml_expand_body.h:
runtime/mercury_ml_arg_body.h:
runtime/mercury_ml_deconstruct_body.h:
runtime/mercury_ml_functor_body.h:
These modules collectively implement the predicates in deconstruct.m
in the library, and provide access to its functionality to other C code,
e.g. in the debugger. Update these to be able to handle terms with the
new data representation optimization.
This update requires a significant change in the distribution of work
between these files for the predicates deconstruct.deconstruct and
deconstruct.limited_deconstruct. We used to have mercury_ml_expand_body.h
fill in the fields of their expand_info structures (whose types are
defined in mercury_deconstruct.h) with pointers to three vectors:
(a) a vector of arg_locns with one element per argument, with a NULL
pointer being equivalent to a vector with a given element in every slot;
(b) a vector of type_infos with one element per argument, constructed
dynamically (and later freed) if necessary; and (c) a vector of argument
words. Once upon a time, before double-word and sub-word arguments,
vector (c) also had one word per argument, but that hasn't been true
for a while; we added vector (a) help the consumers of the expand_info
decode the difference. The consumers of this info always used these
vectors to build up a Mercury term containing a list of univs,
with one univ for each argument.
This structure could be stretched to handle function symbols that store
*all* their arguments in a tagword next to a local sectag, but I found
that stretching it to cover function symbols that have *some* of their
arguments packed next to a remote sectag and *some other* of their
arguments in a memory cell as usual would have required a well-nigh
incomprehensibly complex, and therefore almost undebuggable, interface
between mercury_ml_expand_body.h and the other files above. This diff
therefore changes the interface to have mercury_ml_expand_body.h
build the list of univs directly. This make its code relatively simple
and self-contained, and it should be somewhat faster then the old code
as well, since it never needs to allocate, fill in and then free
vectors of type_infos (each such typeinfo now gets put into a univ
as soon as it is constructed). The downside is that if we ever wanted
to get all the arguments at once for a purpose other than constructing
a list of univs from them, it would nevertheless require constructing
that list of univs anyway as an intermediate data structure. I don't see
this downside is significant, because (a) I don't think such a use case
is very likely, and (b) even if one arises, debuggable but a bit slow
is probably preferable to faster but very hard to debug.
Reduce the level of indentation of some of these files to make the code
easier to edit. Do this by
- not adding an indent level from switch statements to their cases; and
- not adding an indent level when a case in a switch has a local block.
Move the break or return ending a case inside that case's block,
if it has one.
runtime/mercury_deep_copy_body.h:
runtime/mercury_table_type_body.h:
Update these to enable the copying or tabling of terms whose
representations uses the new optimization.
Use the techniques listed above to reduce the level of indentation
make the code easier to edit.
runtime/mercury_tabling.c:
runtime/mercury_term_size.c:
Conform to the changes above.
runtime/mercury_unify_compare_body.h:
Make this code compile after the changes above. It does need to work
correctly, since we only ever used this code to compare the speed
of unify-by-rtti with the speed of unify-by-compiler-generated-code,
and in real life, we always use the latter. (It hasn't been updated
to work right with previous arg packing changes either.)
library/construct.m:
Update to enable the code to construct terms whose representations
uses the new optimization.
Add some sanity checks.
library/private_builtin.m:
runtime/mercury_dotnet.cs.in:
java/runtime/Sectag_Locn.java:
Update the list of possible sectag kinds.
library/store.m:
Conform to the changes above.
trace/mercury_trace_vars.c:
Conform to the changes above.
tests/hard_coded/deconstruct_arg.{m,exp,exp2}:
Extend this test to test the deconstruction of terms whose
representations uses the new optimization.
Modify some of the existing terms being tested to make them more diverse,
in order to make the output easier to navigate.
tests/hard_coded/construct_packed.{m,exp}:
A new test case to test the construction of terms whose
representations uses the new optimization.
tests/debugger/browse_packed.{m,exp}:
A new test case to test access to the fields of terms whose
representations uses the new optimization.
tests/tabling/test_packed.{m,exp}:
A new test case to test the tabling of terms whose
representations uses the new optimization.
tests/debugger/Mmakefile:
tests/hard_coded/Mmakefile:
tests/tabling/Mmakefile:
Enable the new test cases.
|
||
|
Zoltan Somogyi
|
66d20fc6d3 |
Treat apw_partial_first the same as apw_partial_shifted.
Their different treatment was to the root cause of the bug that was fixed
in commit
|
||
|
Zoltan Somogyi
|
624aaa01f1 |
Pack subword-sized arguments next to a local sectag.
compiler/du_type_layout.m:
If a new option is set, then try to represent function symbols with
only subword-sized arguments by packing those arguments into the same word
as the primary tag and (if it is needed) a secondary tag.
If there are too many such function symbols for the available number of
bits, pick the ones that need the least number of bits, in order to
allow us to use this representation for as many such function symbols
as possible.
This diff implements this packing only for types that have more than one
argument, because implementing it for types that have only one argument
has two extra complications. One is the need for another new cons_id
(see below), which would make this diff bigger and harder to review.
The other is the need to consider interactions with the direct_arg
optimization.
Don't invoke the code for deciding the representation of arguments
if either (a) the function symbol has no arguments, or (b) its cons_id
alone dictates how we will treat its argument (in such cases, there is
always exactly one).
Fix a bug in computing the number of bits needed to distinguish N things.
Store the value of the "experiment" option in the params for now,
since it has helped track down bugs in this change, and may do the same
for my next change. It costs next to nothing.
compiler/options.m:
Add an option that controls whether we allow du_type_layout to pack
arguments next to local secondary tags. The default value is "no",
since "yes" may break binary compatibility.
Add an option that controls whether we allow du_type_layout to pack
arguments next to remote secondary tags. This option is not yet used.
compiler/hlds_data.m:
Add a new cons_id, shared_local_tag_with_args, to represent function
symbols in which the arguments are packed next to a local secondary tag.
Rename the existing shared_local_tag cons_id as shared_local_tag_no_args,
to clarify the distinction.
Redesign the representation of secondary tags a bit, to meet the
requirements I discovered while implementing the new data representation.
compiler/prog_data.m:
Document the now-expanded uses of the arg_pos_width type.
compiler/ml_unify_gen.m:
compiler/unify_gen.m:
Implement unifications involving the new cons_id.
compiler/var_locn.m:
Implement deconstruction unifications involving both right-to-left data
flow and the new cons_id for the LLDS backend requires var_locn.m
to implement a new kind of assignment to a variable: one that updates
its old value. Add a predicate for this. (Previously, deconstructions
with right-to-left flow could update the old value of a word in a
memory cell, whose state var_locn.m does *not* track.)
compiler/code_loc_dep.m:
Provide the interface between unify_gen. and var_locn.m.
compiler/code_info.m:
Store the number of primary tag bits in the code_info, to save it looking
up in the globals structure, since with its new code, unify_gen.m needs it
more often now.
compiler/hlds_out_module.m:
doc/user_guide.texi:
Implement the capability of restricting the dump of the type table
to only the types defined in the module being compiled. Without this,
the type table is cluttered with information about types in other
modules, including the automatically-included builtin modules.
compiler/handle_options.m:
Add a new value of the -D option. The new value, du, asks for the
dumping out of the representations of only the locally defined types.
compiler/ml_gen_info.m:
Store the number of primary tag bits as a uint8, not as int.
compiler/ml_tag_switch.m:
compiler/switch_util.m:
compiler/tag_switch.m:
Update the code that generates switches on du types to handle
local secondary tags that must be masked off before use.
compiler/rtti.m:
Update the compiler's representation of RTTI information to account for
the new data representation.
compiler/type_ctor_info.m:
Construct the updated RTTI representation.
compiler/bytecode_gen.m:
compiler/export.m:
compiler/ml_switch_gen.m:
compiler/ml_type_gen.m:
compiler/modecheck_goal.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
Conform for the changes above.
runtime/mercury_type_info.h:
Extend the representation of du functors in the RTTI to account for
the new data representation scheme. The extensions add only to the
*ends* of structures, or to lists of enum values, with the extensions
only being used if the representation is actually used, which should
allow the updated runtime to also work with .c files that were compiled
with a compiler that does *not* have this diff. For the same reason,
make the old enum value MR_SECTAG_LOCAL a synonym for the new
MR_SECTAG_LOCAL_REST_OF_WORD, which expresses a distinction that
did not previously exist.
Delete a reference to a file that no longer exists.
runtime/mercury_dotnet.cs.in:
library/rtti_implementation.m:
Update the C# and Mercury mirrors of the types updated in
mercury_type_info.h.
runtime/mercury_deconstruct.c:
runtime/mercury_deconstruct_macros.h:
runtime/mercury_ml_expand_body.h:
Implement the deconstruction of terms using the new data representation.
runtime/mercury_deep_copy_body.h:
Implement the copying of terms using the new data representation.
runtime/mercury_table_type_body.h:
Implement the tabling of terms using the new data representation.
runtime/mercury_term_size.c:
Implement computing the size of terms using the new data representation.
runtime/mercury_unify_compare_body.h:
Implement RTTI-based unifications of terms using the new data
representation. (Or at least make a first attempt at this implementation.
We never use RTTI-based unification, so this code has not been tested,
but it is not clear that it *needs* to be tested.)
library/construct.m:
Implement the construction of terms using the new data representation.
library/private_builtin.m:
List MR_SECTAG_LOCAL_REST_OF_WORD as a synonym of MR_SECTAG_LOCAL for Java,
since rtti_to_mlds.m will now emit the new version.
Note that the new data representation is not applicable to Java (or C#),
so it should never see the other kind of sectag (MR_SECTAG_LOCAL_BITS).
tests/hard_coded/sectag_bits.{m,exp}:
tests/hard_coded/sectag_bits_test_data:
A new test case to test the reading in and writing out (and therefore
the construction and deconstruction) of terms containing arguments
packed with a local sectag.
tests/hard_coded/Mmakefile:
Enable the new test case.
|
||
|
Zoltan Somogyi
|
b06b2621b3 |
Move towards packing args with secondary tags.
compiler/hlds_data.m:
Add bespoke types to record information about local and remote secondary
tags. The one for local secondary tags includes the value of the
primary and secondary tag together, since construct unifications
need to assign this value, and it is better to compute this once,
instead leaving the target language compiler to do it, potentially
many times.
Use a wrapped uint8 to record primary tag values, and wrapped uints
to record secondary tag values. The wrap is to prevent any accidental
confusion with other values. The use of uint8 and uint has two purposes.
First, using the tighest possible representation. Tags are never negative,
and primary tags cannot exceed 7. Second, using these types in the compiler
help us eat our own dogfood; if a change causes a problem affecting
these types, its bootcheck should fail, alerting us to the problem.
Add commented-out types and fields that will be needed for packing
sub-word-sized arguments together with both local and remote secondary
tags.
compiler/du_type_layout.m:
Generate references to tags in the new format.
compiler/ml_unify_gen.m:
compiler/unify_gen.m:
compiler/modecheck_goal.m:
Conform to the changes above.
Fix an old bug: the inst corresponding to a constant with a primary
and a local secondary tag is not the secondary tag alone, but both tags
together.
compiler/bytecode.m:
compiler/bytecode_gen.m:
compiler/closure_gen.m:
compiler/disj_gen.m:
compiler/export.m:
compiler/hlds_code_util.m:
compiler/jumpopt.m:
compiler/lco.m:
compiler/llds_out_data.m:
compiler/llds_out_instr.m:
compiler/lookup_switch.m:
compiler/lookup_util.m:
compiler/ml_accurate_gc.m:
compiler/ml_call_gen.m:
compiler/ml_closure_gen.m:
compiler/ml_code_util.m:
compiler/ml_elim_nested.m:
compiler/ml_string_switch.m:
compiler/ml_switch_gen.m:
compiler/ml_tag_switch.m:
compiler/ml_type_gen.m:
compiler/mlds_dump.m:
compiler/mlds_to_c_data.m:
compiler/mlds_to_c_stmt.m:
compiler/opt_debug.m:
compiler/peephole.m:
compiler/rtti.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
compiler/string_switch.m:
compiler/switch_util.m:
compiler/tag_switch.m:
compiler/type_ctor_info.m:
Conform to the change to hlds_data.m.
In two places, in rtti_out.m and rtti_to_mlds.m, delete old code
that was needed only to implement reserved tags, which we have
stopped supporting a few months ago.
library/uint8.m:
library/uint16.m:
library/uint32.m:
library/uint64.m:
Add predicates to cast from each of these types to uint.
|
||
|
Zoltan Somogyi
|
24b98fdafe |
Pack sub-word-sized ints and dummies in terms.
Previously, the only situation in which we could pack two or more arguments
of a term into a single word was when all those arguments are enums. This diff
changes that, so that the arguments can also be sub-word-sized integers
(signed or unsigned), or values of dummy types (which occupy zero bits).
This diff also records, for each argument of a function symbol, not just
whether, and if yes, how it is packed into a word, but also at *what offset*
that word is in the term's heap cell. It is more economical to compute this
once, when the representation of the type is being decided, than to compute
it over and over again when terms with that function symbol are being
constructed or deconstructed. However, for a transition period, we compute
these offsets at *both* times, to check the consistency of the new algorithm
for computing offsets that is run at "decide representation time" with
the old algorithms run at "generate code for a unification time".
compiler/du_type_layout.m:
Make the changes described above: pack sub-word-sized integers and
dummy values into argument words, if possible, and if the relevant
new option allows it. These options are temporary. If we find no problems
with the new packing algorithm in a few weeks, we should be able to
delete them.
Allow 64 bit ints and uints to be stored in unboxed in two words
on 32 bit platforms, if the relevant new option allows it. Support
for this is not yet complete, but it makes sense to implement the
RTTI changes for both this change and one described in the above
paragraph together.
For each packed argument, record not just its width, its shift and
the mask, but also the number of bits the argument takes. Previously,
we computed this on demand from the mask, but there is no real need
for that when simply storing this info is so cheap.
For all arguments, packed or not, record its offset, relative to both
the start of the arguments, and the start of the memory cell. (The two
are different if the arguments are preceded by either a remote secondary
tag, the typeinfos and/or typeclass_infos describing some existentially
typed arguments, or both.) The reason for this is given at the top.
Centralize the decision of the parameters of packing in one predicate.
If the option --inform-suboptimal-packing is given, print an informational
message whenever the code deciding type representations finds that
reordering the arguments of a function symbol would allow it to pack
the arguments of that function symbol into less space.
compiler/options.m:
Add the option --allow-packing-ints which controls whether
du_type_layout.m will attempt to pack {int,uint}{8,16,32} arguments
alongside enum arguments.
Add the option --allow-packing-dummies which controls whether
du_type_layout.m will optimize away (in other words, represent in 0 bits)
arguments of dummy types.
Add the option --allow-double-word-ints which controls whether
du_type_layout.m will store arguments of the types int64 and uint64
unboxed in two words on 32 bit platforms, the way it currently stores
double precision floats.
All three those options are off by default, which preserves binary
compatibility with existing code. However, the first two are ready
to be switched on (the third is not).
All three options are intended to be present in the compiler
only until these changes are tested. Once we deem them sufficiently
tested, I will modify the compiler to always do the packing they control,
at which point we can delete these options. This is why they are not
documented.
Add the option --inform-suboptimal-packing, whose meaning is described
above.
doc/user_guide.texi:
Document --inform-suboptimal-packing.
compiler/prog_data.m:
For each argument of a function symbol in a type definition, use
a new type called arg_pos_width to record the extra information
mentioned above in (offsets for all arguments, and number of bits
for packed arguments).
For each function symbol that has some existential type constraints,
record the extra information mentioned for parse_type_defn.m below.
compiler/hlds_data.m:
Include the position, as well as the width, in the representation
of the arguments of function symbols.
Previously, we used the integer 0 as a tag for dummies. Add a tag to
represent dummy values, since this gives more information to any code
that sees that tag.
compiler/ml_unify_gen.m:
compiler/unify_gen.m:
Handle the packing of dummy values, and of sub-word-sized ints and uints.
Compare the cell offset of each argument computed using existing
algorithms here with the cell offset recorded in the argument's
representation, and abort if they are different.
In some cases, restructure code a bit to make it possible.
For example, for tuples and closures, this means that instead of
simply recording that each tuple argument or closure element
is a full word, we must record its correct offset as well.
Handle the new dummy_tag.
Add prelim (not yet finished) support for double-word int64s/uint64s
on 32 bit platforms.
When packing the values of two or more variables (or constants) into a
single word in a memory cell, optimize away operations that are no-ops,
such as shifting anything by zero bits, shifting the constant zero
by any number of bits, and ORing anything with zero. This makes the
generated code easier to read. It is probably also faster for us
to do it here than to write out a bigger expression, have the C compiler
read in the bigger expression, and then later make the same optimization.
In ml_unify_gen.m, avoid the unnecessary use of a list of the argument
variables' types separate from the list of the argument variables
themselves; just look up the type of each argument variable when it is
processed.
compiler/add_special_pred.m:
When creating special (unify and compare) predicates for tuples,
include the offsets in the representation of their arguments.
Delete an unused predicate.
compiler/llds.m:
Add a new way to create an rval: a cast. We use it to implement
the extraction of signed sub-word-sized integers from packed argument
words in terms. Masking the right N bits out of the packed word
leaves the other 32-N or 64-N bits as zeroes; a cast to int8_t,
int16_t or int32_t will copy the sign bit to these bits.
Likewise, when we pack signed int{8,16,32} values into words,
we cast them to their unsigned versions to throw away any sign-extension
bits in their original word-sized representations.
No similar change is needed for the MLDS, since that already had
a mechanism for casts.
compiler/mlds.m:
Note a potential simplification in the MLDS.
compiler/builtin_lib_types.m:
Add functions to return the Mercury representation of the int64
and uint64 types.
compiler/foreign.m:
Export a specialized version of an existing predicate, to allow
ml_unify_gen.m to avoid the costs of the more general version.
compiler/hlds_out_module.m:
Always print the representations of all arguments, since the
inclusion of position information in those representation means that
the representations of even all-full-word-argument terms are of potential
interest when debugging term representations.
compiler/lco.m:
Do not try to apply LCO to arguments of dummy types. (We could optimize
them differently, by filling them in before they are "computed", but
that is a separate optimization, which is of *very* low priority.)
compiler/liveness.m:
Do not include variables of dummy types in resume points.
The reason for this is that the code that establishes a resume point
returns, for each such variable, a list of *lvals* where that variable
can be found. The new code in unify_gen.m will optimize away assignments
to values of dummy types, so there is *no* lval where they can be found.
We could allocate one, but doing so would be a pessimization. Instead,
we simply don't save and restore such values. When their value (which is
always 0) is needed, we can create them out of thin air.
compiler/ml_global_data.m:
Include the target language in the ml_global_data structure, to prevent
some of its users having to look it up in the module_info.
Add notes about the specializing the implementation of arrays of
int64s/uint64s on 32 bit platforms.
compiler/check_typeclass.m:
compiler/ml_type_gen.m:
Add sanity checks of the new precomputed fields of exist_constraints.
Conform to the changes above.
compiler/mlds_to_c.m:
Add prelim (not yet finished) support for double-word int64s/uint64s
on 32 bit platforms.
Add notes about possible optimizations.
compiler/parse_type_defn.m:
When a function symbol in a type definition contains existential
arguments, precompute and store the set of constrained and unconstrained
type variables. The code in du_type_layout.m needs this information
to compute the number of slots occupied by typeinfos and typeclass_infos
in memory cells for this function symbol, and several other places
in the compiler do too. It is easier and faster to compute this
information just once, and this is the earliest time what that can be done.
compiler/type_ctor_info.m:
Use the prerecorded information about existential types to simplify
the code here
compiler/polymorphism.m:
Add an XXX about possibly using the extra info we now record in
exist_constraints to simplify the job of polymorphism.m.
compiler/pragma_c_gen.m:
compiler/var_locn.m:
Create the values of dummy variables from scratch, if needed.
compiler/rtti.m:
Replace a bool with a bespoke type.
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
When generating RTTI information for the LLDS and MLDS backends
respectively, record new kinds of arguments as needing special
treatment. These are int64s and uint64s stored unboxed in two words
on 32 bit platforms, {int,uint}{8,16,32} values packed into words,
and dummy arguments. Each of these has a special code: its own negative
negative value in the num_bits field of the argument.
Generate slightly better formatted output.
compiler/type_util.m:
Delete a predicate that isn't needed anymore.
compiler/opt_util.m:
Delete a function that hasn't been needed for a while.
Conform to the changes above.
compiler/arg_pack.m:
compiler/bytecode_gen.m:
compiler/call_gen.m:
compiler/code_util.m:
compiler/ctgc.selector.m:
compiler/dupelim.m:
compiler/dupproc.m:
compiler/equiv_type.m:
compiler/equiv_type_hlds.m:
compiler/erl_code_gen.m:
compiler/erl_rtti.m:
compiler/export.m:
compiler/exprn_aux.m:
compiler/global_data.m:
compiler/jumpopt.m:
compiler/livemap.m:
compiler/llds_out_data.m:
compiler/middle_rec.m:
compiler/ml_closure_gen.m:
compiler/ml_switch_gen.m:
compiler/ml_top_gen.m:
compiler/module_qual.qualify_items.m:
compiler/opt_debug.m:
compiler/parse_tree_out.m:
compiler/peephole.m:
compiler/recompilation.usage.m:
compiler/resolve_unify_functor.m:
compiler/stack_layout.m:
compiler/structure_reuse.direct.choose_reuse.m:
compiler/switch_util.m:
compiler/typecheck.m:
compiler/unify_proc.m:
compiler/unused_imports.m:
compiler/xml_documentation.m:
Conform to the changes above.
compiler/llds_out_util.m:
Add a comment.
compiler/ml_code_util.m:
Factor out some common code.
runtime/mercury_type_info.h:
Allocate special values of the MR_arg_bits field of the MR_DuArgLocn type
to designate arguments as two word int64/uint64s, as sub-word-sized
arguments of types {int,uint}{8,16,32}, or as arguments of dummy types.
(We already had a special value for two word float arguments.)
Document the list of places that know about this code, so that they
can be updated if and when it changes.
library/construct.m:
Handle the construction of terms with two-word int64/uint64 arguments,
with packed {int,uint}{8,16,32} arguments, and with dummy arguments.
Factor out the code common to the sectag-present and sectag-absent cases,
to make it possible to do the above in just *one* place.
library/store.m:
Add an XXX to a place that I don't think handles two word arguments
correctly. (I think this is an old bug.)
runtime/mercury_deconstruct.c:
Handle the deconstruction of terms with two-word int64/uint64 arguments,
with packed {int,uint}{8,16,32} arguments, and with dummy arguments.
runtime/mercury_deep_copy_body.h:
Handle the copying of terms with two-word int64/uint64 arguments,
with packed {int,uint}{8,16,32} arguments, and with dummy arguments.
Give a macro a more descriptive name.
runtime/mercury_type_info.c:
Handle taking the size of terms with two-word int64/uint64 arguments,
with packed {int,uint}{8,16,32} arguments, and with dummy arguments.
runtime/mercury.h:
Put related definitions next to each other.
runtime/mercury_deconstruct.h:
runtime/mercury_ml_expand_body.h:
Fix indentation.
tests/hard_coded/construct_test.{m,exp}:
Add to this test case a test of the construction, via the library's
construct.m module, of terms containing packed sub-word-sized integers,
and packed dummies.
tests/hard_coded/deconstruct_arg.{m,exp}:
Convert the source code of this test case to state variable notation,
and update the line number references (in the names of predicates created
from lambda expressions) accordingly.
tests/hard_coded/uint64_ground_term.{m,exp}:
A new test case to check that uint64 values too large to be int64 values
can be stored in static structures.
tests/hard_coded/Mmakefile:
Enable the new test case.
|
||
|
Zoltan Somogyi
|
4b98f58d9d |
Don't use reserved addresses to represent functors.
Late last year, we agreed to delete the ability to use the addresses
of reserved objects as cons_tags. After another (very short) discussion
on m-dev, this diff also deletes the ability to use small integers
(including zero) acting as pointers.
compiler/options.m:
Delete the --num-reserved-addresses option.
Add a synomym for --compiler-sufficiently-recent, with the intention
that support for the representation of reserved addresses in RTTI
code in the runtime will be deleted when all installed compilers
have this new synonym.
compiler/hlds_data.m:
Delete any mention of the reserved addresses from the cons_tag type,
since we don't have reserved addresses anymore.
Don't record for each type whether it uses reserved addresses;
no type can do so anymore.
compiler/rtti.m:
Delete the part of the RTTI representation that dealt with reserved
addresses.
compiler/add_foreign_enum.m:
compiler/add_special_pred.m:
compiler/bytecode_gen.m:
compiler/code_info.m:
compiler/du_type_layout.m:
compiler/equiv_type_hlds.m:
compiler/erl_rtti.m:
compiler/export.m:
compiler/hlds_out_module.m:
compiler/intermod.m:
compiler/ml_switch_gen.m:
compiler/ml_type_gen.m:
compiler/ml_unify_gen.m:
compiler/ml_util.m:
compiler/opt_debug.m:
compiler/prog_data.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
compiler/structure_reuse.direct.choose_reuse.m:
compiler/switch_gen.m:
compiler/switch_util.m:
compiler/type_ctor_info.m:
compiler/type_util.m:
compiler/unify_gen.m:
Conform to the changes above, mostly by deleting code that used to deal
with reserved addresses.
|
||
|
Zoltan Somogyi
|
fb97df69ed |
Make "compute type representations" a separate pass.
The ultimate purpose of this diff is to prepare for future improvements
in type representations, allowing values of some data types to be represented
more compactly than up to now.
The main way this diff does that is by creating a separate pass for deciding
how values of each type should be represented. We have traditionally decided
data representations for each type as its type definition was processed
during the make_hlds pass, but these decisions were always tentative,
and could be overridden later, e.g. when we processed foreign_type or
foreign_enum pragmas for the type. This dispersed decision making algorithm
is hard to understand, and therefore to change.
This diff centralizes decisions about type representations in a separate
pass that does nothing else. It leaves the algorithm distributed among
several files (du_type_layout.m, make_tags.m, and add_foreign_enum.m) for now,
to make reviewing this diff easier, but soon after it is committed I intend
to move all the relevant code to du_type_layout.m, to centralize the decision
code in "space" as well as in time.
For the reason why this pass runs before any of the semantic analysis
passes, instead of after all of them as I originally intended and as we
discussed on m-dev in late october 2017, see the big comment at the start of
du_type_layout.m.
As per another part of that same discussion on m-dev, this diff
makes a start on implementing a new type of item, the type_repn item,
which is intended *only* to be used in compiler-generated interface files,
*not* in source files. It is only a start because we can use these items
only *after* the creation of a separate type representation decision pass,
and this diff is already very big. The code for making the compiler understand
these items will be added later. The code for generating them will be added
later still, once the code for understanding them has been installed on
all our systems.
Since I was going to be working on the affected code anyway, this diff
also carries out two other decisions that came out of that discussion:
- the deletion of the ability to reserve a tag in a type for HAL,
either via a compiler option or via a pragma, and
- the deletion of the ability to represent a functor using the address
of a statically allocated object (which we haven't used and won't use,
because it slows down accesses to *all the other functors* of the type).
compiler/mercury_compile_front_end.m:
Invoke the new pass for making decisions about type representations
after the make_hlds pass. (We used to do only the final part of it then.)
Fix a bad dump stage name.
Add an extra check for what it means for a module to be error free.
Make a sub-switch explicit.
compiler/hlds.m:
compiler/make_hlds.m:
Move the modules that implement the new pass from the make_hlds package
to the hlds package, to give the compiler's top level access to them.
Make the same move for the modules that the new pass's modules need.
Since they are now part of hlds, they cannot reach into make_hlds,
and I think this is a cleaner solution than forwarding predicates.
Delete some forwarding predicates that are no longer needed.
compiler/notes/compiler_design.html:
Document the updated location of the moved modules.
Add an XXX to note a place where the documentation has not been
updated in the past.
compiler/du_type_layout.m:
Add code to implement the new pass.
Keep the algorithm for deciding type representations as close
to the previously used algorithm as possible, since this diff
is already big enough. (The previous algorithm was scattered across
add_type.m, add_foreign_enum.m, and make_hlds_passes.m.)
Simplifications and optimizations will come later, after this module
is merged with make_tags.m and with (at least) the foreign_enum half of
add_foreign_enum.m.
compiler/make_tags.m:
Keep the functionality of this module, which does both the first part
of deciding type representations (tentatively assigning tags to functors,
an assignment that may be overridden later), and the last part (packing
multiple adjacent less-than-word-sized enum args into a single word,
if possible.), but simplify it where possible, and note possibilities
for further improvements.
compiler/add_foreign_enum.m:
This module has two halves, one dealing with foreign_enum pragmas
and one dealing with foreign_export_enum pragmas.
Change the half that deals with foreign_enum pragmas to just build
a data structure that du_type_layout.m will need to make its decisions,
this structure being a map from type_ctors to the foreign enum
specification applicable to the current target language. Include
in this structure a component that add_foreign_enum.m itself can use
to report better error messages for duplicate foreign_enum pragmas;
this component records, for each type_ctor and language, the context
of the previous foreign_enum pragma for that combo.
Change the input for the half that deals with foreign_export_enum pragmas
to reflect the fact that it is invoked by du_type_layout.m after all
decisions about type representations have already been made.
compiler/add_special_pred.m:
Move this module from the make_hlds package to the hlds package,
since the code that adds special preds for type is now called from
du_type_layout.m.
Change the names of predicates to make clear whether they add
only the declaration of a predicate, only its definition, or both.
Don't try to pre-guess whether the implementation of a type's
compare predicate will need an index predicate. Let the code
that generates calls to the index predicate both declare and define
the index predicate. This change removes the potential for
inconsistencies between the two pieces of code.
compiler/add_pred.m:
Move this module from the make_hlds package to the hlds package,
since add_special_pred.m needs access to it.
compiler/add_type.m:
When adding a type definition to the HLDS, don't try to decide
its representation. Any such decision was tentative anyway, due
to the possibility of e.g. the later processing of foreign_type
or foreign_enum pragmas for the type. Likewise, don't try to
create the special (unify, compare) predicates for the type.
Leave both tasks to the du_type_layout pass.
Likewise, don't try to pack the representation of types, or record
no_tag types in the table of no_tag types, during the post-processing
pass either; leave both of these to du_type_layout as well.
Rename the predicate that post_processes type definitions to reflect
the two tasks left for it to do.
compiler/prog_data.m:
Do not store width information about the arguments of those data
constructors in the parse tree. That information is not computed
until later; until then, it was always filled in with dummy values.
(But see hlds_data.m below.)
Use bespoke types to represent the presence or absence of user-specified
unify and compare predicates.
Change the representation of data constructors to use a single "maybe"
type, not two lists, to denote the presence or absence of existentially
typed arguments.
Give the HLDS the ability to hold representation information about
abstract types that in the future we will get from type_repn items
in the defining modules' interface files.
Delete the uses_reserved_tag type, since we never use reserved tags
anymore.
compiler/prog_item.m:
Add the new type_repn item type, which is not used yet.
Delete the reserve_tag pragma.
Fix an earlier mistake in the wording of a context message.
compiler/hlds_data.m:
Put all the fields of hlds_du_type (the type definition variant dealing
with discriminated union types) that deal with type representation
issues in a single "maybe" field that is set to "no" before the
type representation decision pass has been run.
Add new type, constructor_repn, that stores the same information as the old
constructor type (defined in prog_data.m), PLUS the information
describing how terms with that data constructor are stored.
Likewise, add a new type ctor_arg_rep, which likewise stores
the widths of each constructor argument. When we implement
argument reordering, we would store the offset of the arg as well.
Since the parse tree representations of constructors and their arguments
don't store representation information anymore, the cons_table they
are stored in doesn't either. Make the lookup of representation information
for a given constructor possible by adding a map to the new "maybe" field
of hlds_du_type.
Provide some utility predicates.
Optimize some existing predicates.
Rename some types to better reflect their meaning.
compiler/hlds_module.m:
Provide a slot in the module_info for storing the information
gathered by make_hlds.m that is needed by the new pass.
compiler/make_hlds_separate_items.m:
When we see either a foreign_enum or a foreign_export_enum pragma,
return values of a bespoke type for them (a type defined in
hlds_module.m), instead of an item_pragma. This makes handling them
considerably easier.
compiler/make_hlds_passes.m:
With the changes in this diff, adding a type to the HLDS won't
decide its representation. Therefore delete the code that used
to loop over foreign_export_enum pragmas; in the absence of
the final type representation information, it won't work right.
Record the information that the du_type_layout pass will need
in the module_info.
compiler/add_pragma.m:
Delete the code for passing on foreign_enum and foreign_export_enum
pragmas to add_foreign_enum.m; they are now passed to add_foreign_enum.m
by du_type_layout.m.
Move a utility predicate to make_hlds_error.m, to allow add_foreign_enum.m
to call it.
compiler/make_hlds_error.m:
Add the utility predicate moved from add_pragma.m.
Move the module from the make_hlds to the hlds package.
compiler/module_qual.m:
Provide a mechanism for recording error messages about e.g. undefined
types without recording that we found an undefined type. This sounds
strange, but there is a valid use case.
When a type definition declares a functor's argument to be of an
undefined type, that error is usually fatal; we stop the compiler
from proceeding even to typechecking, since the typechecker will
probably abort with a map lookup failure. Most other references
to undefined types are similarly fatal for the same reason. However,
if e.g. a foreign_export_enum pragma refers to an undefined type,
that error *won't* be visible to the typechecker, and therefore
won't crash it. The error will still cause the compiler to exit
without generating any target language code, but at least it will be
able to run the typechecker and other semantic analysis passes.
Without this change, the compiler will report only one error in
the ee_invalid.m test case; with it, it reports *every* error
in the test case expected output.
compiler/module_qual.qualify_items.m:
Use the capability describe above for undefined types in
foreign_export_enum pragmas.
compiler/module_qual.qual_errors.m:
Delete a (somewhat incorrect) copy of a predicate in prog_item.m,
to reduce code duplication.
compiler/prog_type.m:
Add ways to represent abstract types whose representations are nevertheless
known (from type_repn items in the defining modules' interface files)
to be notag or dummy types. This will be needed to fix Mantis bug #441,
a fix that will probably be one of the first later changes to build
on this diff.
Delete a type moved to type_util.m.
compiler/type_util.m:
Provide extra versions of some predicates, with the difference between
the old and the new versions being that one requires type representations
to have been decided already, and the other one does not.
Move the definition of the ctor_defn type here from prog_type.m,
since prog_type.m itself does not use it, but type_util.m does.
Give some predicates more meaningful names.
compiler/parse_type_defn.m:
Simplify the code for parsing type definitions, to make it easier
to reuse to parse type_repn items.
Add a sanity check that requires existential constraints to have
*some* existential variables to apply to.
Allow "type_is_representable_in_n_bits" as a synonym for
"type_is_abstract_enum", since in the future we want to be able to pack
e.g. multiple int8s, not just multiple enums, into a single word.
Generate more specific error messages for some classes of malformed input.
compiler/parse_type_repn.m:
New module to parse type_repn items.
compiler/polymorphism.m:
Make some predicates that operate on type constructors take
the type constructors themselves as input arguments, not a whole type
*using* that type constructor. Put the arguments of those predicates
in a more standard order.
Note that some predicates don't belong in this module.
compiler/special_pred.m:
Make the code that decides whether a special predicate for a type
constructor can be defined lazily avoid using type representation
information. (Actually, we now make decisions about lazy vs eager
definitions after type representation is available, but that was
not so in an earlier version of this change, and the new code
is more robust.)
compiler/unify_proc.m:
When we decide to generate code for a compare predicate that needs
the type to have an index predicate, don't presume that the index
predicate has already been declared and defined; instead, declare
and define it then and there. (Index predicates are *never* called
from anywhere else.)
Pack the information needed to define a special predicate
into a single structure, to simplify the above.
Since the creation of a clause for a compare predicate may now require
the declaration and definition of an index predicate, the module_info
field of the unify_proc_info is now a writeable field.
Give some predicates and function symbols more meaningful names.
Note some problems with the existing code.
compiler/add_class.m:
compiler/add_clause.m:
compiler/add_foreign_proc.m:
compiler/add_mode.m:
compiler/add_mutable_aux_preds.m:
compiler/add_pragma_tabling.m:
compiler/add_pragma_type_spec.m:
compiler/add_solver.m:
compiler/check_typeclass.m:
compiler/code_info.m:
compiler/comp_unit_interface.m:
compiler/ctgc.selector.m:
compiler/ctgc.util.m:
compiler/default_func_mode.m:
compiler/det_report.m:
compiler/equiv_type.m:
compiler/equiv_type_hlds.m:
compiler/erl_code_gen.m:
compiler/export.m:
compiler/foreign.m:
compiler/get_dependencies.m:
compiler/goal_expr_to_goal.m:
compiler/hhf.m:
compiler/higher_order.m:
compiler/hlds_code_util.m:
compiler/hlds_out_module.m:
compiler/inst_check.m:
compiler/inst_test.m:
compiler/inst_util.m:
compiler/intermod.m:
compiler/item_util.m:
compiler/make_hlds_warn.m:
compiler/ml_accurate_gc.m:
compiler/ml_simplify_switch.m:
compiler/ml_type_gen.m:
compiler/ml_unify_gen.m:
compiler/mlds_to_cs.m:
compiler/mlds_to_java.m:
compiler/mode_util.m:
compiler/modecheck_goal.m:
compiler/module_qual.collect_mq_info.m:
compiler/modules.m:
compiler/parse_item.m:
compiler/parse_pragma.m:
compiler/parse_tree.m:
compiler/parse_tree_out.m:
compiler/parse_tree_out_pragma.m:
compiler/post_term_analysis.m:
compiler/proc_requests.m:
compiler/prog_item_stats.m:
compiler/qual_info.m:
compiler/recompilation.check.m:
compiler/recompilation.usage.m:
compiler/recompilation.version.m:
compiler/resolve_unify_functor.m:
compiler/rtti.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
compiler/simplify_goal_ite.m:
compiler/stack_opt.m:
compiler/state_var.m:
compiler/structure_reuse.direct.choose_reuse.m:
compiler/superhomogeneous.m:
compiler/switch_gen.m:
compiler/switch_util.m:
compiler/table_gen.m:
compiler/term_constr_build.m:
compiler/term_norm.m:
compiler/trailing_analysis.m:
compiler/type_constraints.m:
compiler/type_ctor_info.m:
compiler/typecheck.m:
compiler/unify_gen.m:
compiler/untupling.m:
compiler/unused_imports.m:
compiler/write_module_interface_files.m:
compiler/xml_documentation.m:
Conform to the changes above.
tests/invalid/Mmakefile:
Disable the reserve_tag test case, as it is not applicable anymore.
tests/invalid/exported_foreign_enum.{m,err_exp}:
tests/invalid/pragma_qual_error.{m,err_exp}:
Delete reserve_tag pragmas from these test cases, and its effects
from the expected outputs.
tests/invalid/bad_foreign_type.err_exp:
tests/invalid/bigtest.err_exp:
tests/invalid/foreign_enum_invalid.err_exp:
tests/invalid/type_lhs_var.err_exp:
tests/invalid/uu_type.err_exp:
tests/invalid/where_abstract_enum.err_exp:
tests/invalid/where_direct_arg.err_exp:
Expect the updated messages for some errors.
tests/valid/Mmake.valid.common:
tests/valid/Mmakefile:
Disable any reserve_tag test cases, as they are not applicable anymore.
|
||
|
Zoltan Somogyi
|
47f1df4a0a |
Split mlds_data_defn into three separate types.
We used to use mlds_data_defns to represent three related but nevertheless
distinct kinds of entities: global variables, local variables, and fields
in classes. This diff replaces the mlds_data_defn type with three separate
types: mlds_global_var_defn, mlds_local_var_defn and mlds_field_var_defn
respectively, with corresponding changes to related types, such as
mlds_data_name.
The global variables are completely separate from the other two kinds.
Local and field variables are *mostly* separate from each other, but they
are related in one way. When we flatten out nested functions, the child
nested function can no longer access its parent function's local variables,
so we pass those variables to it as fields of an environment structure.
This requires turning local variables to fields of that structure,
and the code in the flattened previously-nested function that accesses
those fields naturally wants to treat them as if they were local variables
(as indeed they sort-of were before the flattening). There are therefore
ways to convert each of local and fields vars into the other.
This restructuring makes clear several invariants of the MLDS we generate
that were previously hidden. For example, variables with certain kinds of
names (in the before-this-diff, general version of the mlds_var_name type)
could appear only as function arguments or as locals in ml_stmt_blocks,
not in ml_global_data, while for some other names the opposite was the case.
And in several cases, functions used to take a general mlds_data_defn
as argument but aborted if given the "wrong kind" of mlds_data_defn.
This diff also makes possible further simplifications. For example,
local vars should not need some flags (since e.g. they are never per-instance),
and should never need either module or type qualification, while global
variables (which are also never per-instance) should never need type
qualification (since they are not fields of a type). The definitions
in blocks should consist of local variables and (before flattening) functions,
not global variables, field variables or classes, while the members in classes
should be only field variables and functions (and maybe classes), not
global or local variables. Those changes will be in future diffs;
this is already large enough.
compiler/mlds.m:
Make the changes described above.
Use tighter types where possible.
Use (a generalized version) of the mlconst_named_const functor
to represent values of enum types defined in the runtimes
of the target platforms.
compiler/ml_global_data.m:
Store *only* global variables in fields that previously stored general
mlds_datas (that by design were always global).
Store *only* closure wrapper functions in the previous non-flat-defns
field. Before this diff, the code generator only put closure wrapper
functions in this field, but then ml_elim_nested.m put everything
resulting from the expansion of those functions back into those fields
as well, some of which were not functions. It now puts those non-function
things into the MLDS data structure directly.
compiler/ml_code_util.m:
compiler/ml_util.m:
Conform to the changes above.
Use tighter types where possible. If appropriate, change the name
of the function or predicate accordingly.
Represent references to enum constants defined in the runtime of the
target language as named constants (since they is what they are),
instead of representing them as MLDS "variables", which required
the code of mlds_to_cs.m had to special-case the treatment
of those "variables".
compiler/ml_elim_nested.m:
Conform to the changes above.
Use tighter types where possible.
Don't put the environment types resulting from flattening nested scopes
back into the non-flat-defns slot of the ml_elim_info; instead, return
them separately to code that puts them directly in the MLDS.
compiler/rtti.m:
When returning the names of enum constants in the C runtime, return also
the prefixes that you need to place in front of these to obtain their names
in the Java and C# runtimes.
compiler/mercury_compile_mlds_back_end.m:
compiler/ml_accurate_gc.m:
compiler/ml_call_gen.m:
compiler/ml_closure_gen.m:
compiler/ml_code_gen.m:
compiler/ml_commit_gen.m:
compiler/ml_disj_gen.m:
compiler/ml_foreign_proc_gen.m:
compiler/ml_gen_info.m:
compiler/ml_lookup_switch.m:
compiler/ml_optimize.m:
compiler/ml_proc_gen.m:
compiler/ml_string_switch.m:
compiler/ml_switch_gen.m:
compiler/ml_tailcall.m:
compiler/ml_type_gen.m:
compiler/ml_unify_gen.m:
compiler/mlds_to_c.m:
compiler/mlds_to_cs.m:
compiler/mlds_to_java.m:
compiler/mlds_to_target_util.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
Conform to the changes above.
Move a utility function from ml_util.m to mlds_to_target_util.m,
since it is used only in mlds_to_*.m.
|
||
|
Zoltan Somogyi
|
b96dacdcac |
Make a bunch of switches complete.
compiler/exception_analysis.m:
compiler/frameopt.m:
compiler/get_dependencies.m:
compiler/labelopt.m:
compiler/ml_global_data.m:
compiler/ml_optimize.m:
compiler/options_file.m:
compiler/parse_class.m:
compiler/rtti.m:
compiler/rtti_out.m:
compiler/type_util.m:
compiler/var_locn.m:
As above.
compiler/inst_match.m:
Fix a stray piece of code.
compiler/loop_inv.m:
Add a module qualification.
compiler/opt_util.m:
Delete two unused predicates.
|
||
|
Zoltan Somogyi
|
31ad78c1a9 |
Make c_util.m and file_util.m use explicit streams.
compiler/c_util.m:
compiler/file_util.m:
Replace every predicate that implicitly wrote to the current output stream
with two predicates: a predicate with the old name that takes an explicit
output stream parameter, and a predicate that still writes to the current
output stream, but whose name makes this fact clear.
Make some other minor improvements.
compiler/bytecode.m:
compiler/elds_to_erlang.m:
compiler/export.m:
compiler/fact_table.m:
compiler/layout_out.m:
compiler/llds_out_data.m:
compiler/llds_out_file.m:
compiler/llds_out_global.m:
compiler/llds_out_instr.m:
compiler/llds_out_util.m:
compiler/mlds_to_c.m:
compiler/mlds_to_cs.m:
compiler/mlds_to_java.m:
compiler/rtti_out.m:
Conform to the changes above.
|
||
|
Mark Brown
|
3acbf03059 |
Implement combined higher-order types and insts.
These allow types to be defined in the following manner:
:- type job ---> job(pred(int::out, io::di, io::uo) is det).
For any construction unification using this functor the argument must
have the required higher-order inst; it is a mode error if it does not.
When terms of type job with inst ground are deconstructed, the argument
is inferred to have the given inst, allowing a higher-order call in that
mode.
The new type syntax is currently only permitted as the direct argument of
a functor in a du type definition. In future it would be meaningful to
support this syntax in other locations, but that is left for a separate
change.
In order to correctly implement the construct/3 library predicate, we
need to be able to dynamically check that arguments do not violate
any constraints on the argument insts. At the moment, we conservatively
abort if any such constraints are present irrespective of whether they
are satisfied or not. Since these constraints are a new feature, no
existing code will abort in this way.
The implementation refers to the inst information associated with types
as "subtype information". This is because, generally, we think of the
combination of a type with a fully bound inst (i.e., one that describes
terms that contain no unbound variables) describes a subtype of that type.
compiler/inst_util.m:
Ensure that arguments have the necessary insts in construction
unifications.
Where available, propagate the insts into arguments rather than
using ground(shared, none).
compiler/prog_io_type_name.m:
Parse the new form of types.
compiler/unparse.m:
Unparse the new form of types.
compiler/prog_io_type_defn.m:
Allow the new form of types in functor arguments.
compiler/prog_ctgc.m:
compiler/prog_io_item.m:
compiler/prog_io_mutable.m:
compiler/prog_io_pragma.m:
compiler/prog_io_typeclass.m:
compiler/superhomogeneous.m:
Disallow the new form of types in places other than functor
arguments.
compiler/prog_data.m:
Go back to representing function types with result type appended
to the arguments. In most case this now results in simpler code.
compiler/prog_type.m:
Abstract away the representation of predicate vs function arguments
by using a predicate to construct these types.
compiler/rtti.m:
compiler/type_ctor_info.m:
Include subtype information about the arguments of a du functor
and about the argument of a notag functor. Generate this
information from the argument types.
Currently, the information is one bit which says whether or not
any subtypes exist in the arguments.
Bump the RTTI version number from the compiler side.
compiler/rtti_out.m:
Output functor subtype information for the low-level C backend.
compiler/rtti_to_mlds.m:
Include functor subtype information in the MLDS.
compiler/mlds_to_cs.m:
Add the new runtime type to the special cases.
compiler/erl_rtti.m:
compiler/erlang_rtti.m:
library/erlang_rtti_implementation.m:
Include functor subtype info in the erlang RTTI.
java/runtime/DuFunctorDesc.java:
java/runtime/FunctorSubtypeInfo.java:
Include functor subtype information in the Java runtime.
runtime/mercury_dotnet.cs.in:
Include functor subtype information in the C# runtime.
runtime/mercury_type_info.h:
Include functor subtype information in the C runtime.
Bump the RTTI version number in the runtime.
Define macros to access the new field. These macros can correctly
handle the previous RTTI version, therefore we do not need to
change the minimum version at this time.
library/private_builtin.m:
Define constants for use by the Java backend.
library/construct.m:
library/rtti_implementation.m:
Use the new RTTI to ensure we don't attempt to construct terms
that violate the new insts.
compiler/prog_rep_tables.m:
Ignore the new inst info for now.
compiler/*.m:
Changes to conform to above.
doc/reference_manual.texi:
Document the new feature.
tests/hard_coded/functor_ho_inst.{m,exp}:
tests/hard_coded/functor_ho_inst_2.{m,exp}:
tests/hard_coded/functor_ho_inst_excp.{m,exp}:
tests/hard_coded/functor_ho_inst_excp_2.{m,exp}:
Test the new functionality.
tests/invalid/combined_ho_type_inst.{m,err_exp}:
tests/invalid/combined_ho_type_inst_2.{m,err_exp}:
Test that we don't allow the new types where they are not permitted,
or are incomplete.
tests/invalid/functor_ho_inst_bad.{m,err_exp}:
tests/invalid/functor_ho_inst_bad_2.{m,err_exp}:
tests/invalid/functor_ho_inst_bad_3.{m,err_exp}:
Test that the argument inst information is enforced as required.
tests/hard_coded/Mmakefile:
tests/invalid/Mmakefile:
Run the new test cases.
|
||
|
Zoltan Somogyi
|
3cb166eb07 | Delete some unused predicates; add consider_used pragmas for some others. | ||
|
Zoltan Somogyi
|
cc9912faa8 |
Don't import anything in packages.
Packages are modules whose only job is to serve as a container for submodules. Modules like top_level.m, hlds.m, parse_tree.m and ll_backend.m are packages in this (informal) sense. Besides the include_module declarations for their submodules, most of the packages in the compiler used to import some modules, mostly other packages whose component modules their submodules may need. For example, ll_backend.m used to import parse_tree.m. This meant that modules in the ll_backend package did not have to import parse_tree.m before importing modules in the parse_tree package. However, this had a price. When we add a new module to the parse_tree package, parse_tree.int would change, and this would require the recompilation of ALL the modules in the ll_backend package, even the ones that did NOT import ANY of the modules in the parse_tree package. This happened even at one remove. Pretty much all modules in every one of the backend have to import one or more modules in the hlds package, and they therefore have import hlds.m. Since hlds.m imported transform_hlds.m, any addition of a new middle pass to the transform_hlds package required the recompilation of all backend modules, even in the usual case of the two having nothing to do with each other. This diff removes all import_module declarations from the packages, and replaces them with import_module declarations in the modules that need them. This includes only a SUBSET of their child modules and of the non-child modules that import them. |
||
|
Zoltan Somogyi
|
58e34cbbfa | Convert (C->T;E) to (if C then T else E). | ||
|
Zoltan Somogyi
|
500948d549 |
Break up mdbcomp/prim_data.m. The new modules have much better cohesion.
mdbcomp/sym_name.m:
New module, containing the part of the old prim_data.m that
dealt with sym_names.
mdbcomp/builtin_modules.m:
New module, containing the part of the old prim_data.m that
dealt with builtin modules.
mdbcomp/prim_data.m:
Remove the things that are now in the two new modules.
mdbcomp/mdbcomp.m:
deep_proiler/Mmakefile:
slice/Mmakefile:
Add the two new modules.
browser/*.m:
compiler/*.m:
deep_proiler/*.m:
mdbcomp/*.m:
slice/*.m:
Conform to the above changes.
|
||
|
Zoltan Somogyi
|
16bd4acd2f |
Shorten lines longer than 79 characters.
Estimated hours taken: 2 Branches: main compiler/*.m: Shorten lines longer than 79 characters. |
||
|
Peter Wang
|
257efbd678 |
Store double-precision `float' constructor arguments in unboxed form,
Branches: main Store double-precision `float' constructor arguments in unboxed form, in high-level C grades on 32-bit platforms, i.e. `float' (and equivalent) arguments may occupy two machine words. As the C code generated by the MLDS back-end makes use of MR_Float variables and parameters, float (un)boxing may be reduced substantially in many programs. compiler/prog_data.m: Add `double_word' as a new option for constructor argument widths, only used for float arguments as yet. compiler/make_hlds_passes.m: Set constructor arguments to have `double_word' width if required, and possible. compiler/type_util.m: Add helper predicate. compiler/builtin_ops.m: compiler/c_util.m: compiler/llds.m: Add two new binary operators used by the MLDS back-end. compiler/arg_pack.m: Handle `double_word' arguments. compiler/ml_code_util.m: Deciding whether or not a float constructor argument requires boxing now depends on the width of the field. compiler/ml_global_data.m: When a float constant appears as an initialiser of a generic array element, it is now always unboxed, irrespective of --unboxed-float. compiler/ml_type_gen.m: Take double-word arguments into account when generating structure fields. compiler/ml_unify_gen.m: Handle double-word float constructor arguments in (de)constructions. In some cases we break a float argument into its two words, so generating two assignments statements or two separate rvals. Take double-word arguments into account when calculating field offsets. compiler/mlds_to_c.m: The new binary operators require no changes here. As a special case, write `MR_float_from_dword_ptr(&X)' instead of `MR_float_from_dword(X, Y)' when X, Y are consecutive words within a field. The definition of `MR_float_from_dword_ptr' is more straightforward, and gcc produces better code than if we use the more general `MR_float_from_dword'. compiler/rtti_out.m: For double-word arguments, generate MR_DuArgLocn structures with MR_arg_bits set to -1. compiler/rtti_to_mlds.m: Handle double-word arguments in field offset calculation. compiler/unify_gen.m: Partially handle double_word arguments in LLDS back-end. compiler/handle_options.m: Set --unboxed-float when targetting Java, C# and Erlang. compiler/structure_reuse.direct.choose_reuse.m: Rename a predicate. compiler/bytecode.m: compiler/equiv_type.m: compiler/equiv_type_hlds.m: compiler/llds_to_x86_64.m: compiler/mlds_to_gcc.m: compiler/mlds_to_il.m: compiler/opt_debug.m: Conform to changes. library/construct.m: library/store.m: Handle double-word constructor arguments. runtime/mercury_conf.h.in: Clarify what `MR_BOXED_FLOAT' now means. runtime/mercury_float.h: Add helper macros for converting between doubles and word/dwords. runtime/mercury_deconstruct.c: runtime/mercury_deconstruct.h: Add a macro `MR_arg_value' and a helper function to extract a constructor argument value. This replaces `MR_unpack_arg'. runtime/mercury_type_info.h: Remove `MR_unpack_arg'. Document that MR_DuArgLocn.MR_arg_bits may be -1. runtime/mercury_deconstruct_macros.h: runtime/mercury_deep_copy_body.h: runtime/mercury_ml_arg_body.h: runtime/mercury_table_type_body.h: runtime/mercury_tabling.c: runtime/mercury_type_info.c: Handle double-word constructor arguments. tests/hard_coded/Mercury.options: tests/hard_coded/Mmakefile: tests/hard_coded/lco_double.exp: tests/hard_coded/lco_double.m: tests/hard_coded/pack_args_float.exp: tests/hard_coded/pack_args_float.m: Add test cases. trace/mercury_trace_vars.c: Conform to changes. |
||
|
Peter Wang
|
64bb189f34 |
Fix bug #204. Deconstruct of packed arguments was broken due to a mistake in
Branches: main Fix bug #204. Deconstruct of packed arguments was broken due to a mistake in the RTTI tables. compiler/prog_data.m: compiler/rtti_out.m: compiler/rtti_to_mlds.m: Correct the computation of the number of bits occupied by a packed argument, given the bit-mask. tests/hard_coded/Mmakefile: tests/hard_coded/zinc_pack_bug.exp: tests/hard_coded/zinc_pack_bug.m: Add test case. |
||
|
Peter Wang
|
0ae65de577 |
Pack consecutive enumeration arguments in discriminated union types into a
Branches: main Pack consecutive enumeration arguments in discriminated union types into a single word to reduce cell sizes. Argument packing is only enabled on C back-ends with low-level data, and reordering arguments to improve opportunities for packing is not yet attempted. The RTTI implementations for other back-ends will need to be updated, but that is best left until after any argument reordering change. Modules which import abstract enumeration types are notified so by writing declarations of the form: :- type foo where type_is_abstract_enum(NumBits). into the interface file for the module which defines the type. compiler/prog_data.m: Add an `arg_width' argument to constructor arguments. Replace `is_solver_type' by `abstract_type_details', with an extra option for abstract exported enumeration types. compiler/handle_options.m: compiler/options.m: Add an internal option `--allow-argument-packing'. compiler/make_hlds_passes.m: Determine whether and how to pack enumeration arguments, updating the `arg_width' fields of constructor arguments before constructors are added to the HLDS. compiler/mercury_to_mercury.m: compiler/modules.m: Write `where type_is_abstract_enum(NumBits)' to interface files for abstract exported enumeration types. compiler/prog_io_type_defn.m: Parse `where type_is_abstract_enum(NumBits)' attributes on type definitions. compiler/arg_pack.m: compiler/backend_libs.m: Add a new module. This mainly contains a predicate which packs rvals according to arg_widths, which is used by both LLDS and MLDS back-ends. compiler/ml_unify_gen.m: compiler/unify_gen.m: Take argument packing into account when generating code for constructions and deconstructions. Only a relatively small part of the compiler actually needs to understand argument packing. The rest works at the HLDS level with constructor arguments and variables, or at the LLDS and MLDS levels with structure fields. compiler/code_info.m: compiler/var_locn.m: Add assign_field_lval_expr_to_var and var_locn_assign_field_lval_expr_to_var. Allow more kinds of rvals in assign_cell_arg. I do not know why it was previously restricted, except that the other kinds of rvals were not encountered as cell arguments before. compiler/mlds.m: We can now rely on the compiler to pack arguments in the mlds_decl_flags type instead of doing it manually. A slight downside is that though the type is packed down to a single word cell, it will still incur a memory allocation per cell. However, I did not notice any difference in compiler speed. compiler/rtti.m: compiler/rtti_out.m: Add and output a new field for MR_DuFunctorDesc instances, which, if any arguments are packed, points to an array of MR_DuArgLocn. Each array element describes the offset in the cell at which the argument's value is held, and which bits of the word it occupies. In the more common case where no arguments are packed, the new field is simply null. compiler/rtti_to_mlds.m: Generate the new field to MR_DuFunctorDesc. compiler/structure_reuse.direct.choose_reuse.m: For now, prevent structure reuse reusing a dead cell which has a different constructor to the new cell. The code to determine whether a dead cell will hold the arguments of a new cell with a different constructor will need to be updated to account for argument packing. compiler/type_ctor_info.m: Bump RTTI version number. Conform to changes. compiler/add_type.m: compiler/check_typeclass.m: compiler/equiv_type.m: compiler/equiv_type_hlds.m: compiler/erl_rtti.m: compiler/hlds_data.m: compiler/hlds_out_module.m: compiler/intermod.m: compiler/make_tags.m: compiler/mlds_to_gcc.m: compiler/opt_debug.m: compiler/prog_type.m: compiler/recompilation.check.m: compiler/recompilation.version.m: compiler/special_pred.m: compiler/type_constraints.m: compiler/type_util.m: compiler/unify_proc.m: compiler/xml_documentation.m: Conform to changes. Reduce code duplication in classify_type_defn. compiler/hlds_goal.m: Clarify a comment. library/construct.m: Make `construct' pack arguments when necessary. Remove an old RTTI version number check as recommended in mercury_grade.h. library/store.m: Deal with packed arguments in this module. runtime/mercury_grade.h: Bump binary compatibility version number. runtime/mercury_type_info.c: runtime/mercury_type_info.h: Bump RTTI version number. Add MR_DuArgLocn structure definition. Add a macro to unpack an argument as described by MR_DuArgLocn. Add a function to determine a cell's size, since the number of arguments is no longer correct. runtime/mercury_deconstruct.c: runtime/mercury_deconstruct.h: runtime/mercury_deconstruct_macros.h: runtime/mercury_ml_arg_body.h: runtime/mercury_ml_expand_body.h: Deal with packed arguments when deconstructing. Remove an old RTTI version number check as recommended in mercury_grade.h. runtime/mercury_deep_copy_body.h: Deal with packed arguments when copying. runtime/mercury_table_type_body.h: Deal with packed arguments in tabling. runtime/mercury_dotnet.cs.in: Add DuArgLocn field to DuFunctorDesc. Argument packing is not enabled for the C# back-end yet so this is unused. trace/mercury_trace_vars.c: Deal with packed arguments in MR_select_specified_subterm, use for the `hold' command. java/runtime/DuArgLocn.java: java/runtime/DuFunctorDesc.java: Add DuArgLocn field to DuFunctorDesc. Argument packing is not enabled for the Java back-end yet so this is unused. extras/trailed_update/tr_store.m: Deal with packed arguments in this module (untested). extras/trailed_update/samples/interpreter.m: extras/trailed_update/tr_array.m: Conform to argument reordering in the array, map and other modules in previous changes. tests/hard_coded/Mercury.options: tests/hard_coded/Mmakefile: tests/hard_coded/lco_pack_args.exp: tests/hard_coded/lco_pack_args.m: tests/hard_coded/pack_args.exp: tests/hard_coded/pack_args.m: tests/hard_coded/pack_args_copy.exp: tests/hard_coded/pack_args_copy.m: tests/hard_coded/pack_args_intermod1.exp: tests/hard_coded/pack_args_intermod1.m: tests/hard_coded/pack_args_intermod2.m: tests/hard_coded/pack_args_reuse.exp: tests/hard_coded/pack_args_reuse.m: tests/hard_coded/store_ref.exp: tests/hard_coded/store_ref.m: tests/invalid/Mmakefile: tests/invalid/where_abstract_enum.err_exp: tests/invalid/where_abstract_enum.m: tests/tabling/Mmakefile: tests/tabling/pack_args_memo.exp: tests/tabling/pack_args_memo.m: Add new test cases. tests/hard_coded/deconstruct_arg.exp: tests/hard_coded/deconstruct_arg.exp2: tests/hard_coded/deconstruct_arg.m: Add constructors with packed arguments to these cases. tests/invalid/where_direct_arg.err_exp: Update expected output. |
||
|
Peter Wang
|
12281f3419 |
Implement a type representation optimisation ("direct argument functors"),
Branches: main
Implement a type representation optimisation ("direct argument functors"),
where a functor with exactly one argument can be represented by a tagged
pointer to the argument value, which itself does not require the tag bits,
e.g.
:- type maybe_foo ---> yes(foo) ; no.
:- type foo ---> foo(int, int). % aligned pointer
To ensure that all modules which could construct or deconstruct the functor
agree on the type representation, I had planned to automatically output
extra information to .int files to notify importing modules about functors
using the optimised representation:
:- type maybe_foo ---> yes(foo) ; no
where direct_arg is [yes/1].
However, the compiler does not perform enough (or any) semantic analysis
while making interface files. The fallback solution is to only use the
optimised representation when all importing modules can be guaranteed to
import both the top-level type and the argument type, namely, when both
types are exported from the same module. We also allow certain built-in
argument types; currently this only includes tuples.
Non-exported types may use the optimised representation, but when
intermodule optimisation is enabled, they may be written out to .opt files.
Then, we *do* add direct_arg attributes to .opt files to ensure that importing
modules agree on the type representation. The attributes may also be added by
Mercury programmers to source files, which will be copied directly into .int
files without analysis. They will be checked when the module is actually
compiled.
This patch includes work by Zoltan, who independently implemented a version
of this change.
compiler/hlds_data.m:
Record the direct arg functors in hlds_du_type.
Add a new option to cons_tag.
Fix some comments.
compiler/prog_data.m:
compiler/prog_io_type_defn.m:
Parse and record `direct_arg' attributes on type definitions.
compiler/prog_io_pragma.m:
Issue an error if the `direct_arg' attribute is used with a foreign
type.
compiler/make_tags.m:
compiler/mercury_compile_front_end.m:
Add a pass to convert suitable functors to use the direct argument
representation. The argument type must have been added to the type
table, so we do this after all type definitions have been added.
Move code to compute cheaper_tag_test here.
compiler/ml_unify_gen.m:
compiler/unify_gen.m:
Generate different code to construct/deconstruct direct argument
functors.
compiler/intermod.m:
Write `direct_arg' attributes to .opt files for functors
using the direct argument representation.
compiler/mercury_to_mercury.m:
Write out `direct_arg' attributes.
compiler/rtti.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
Add an option to the types which describe the location of secondary
tag options. The functors which can use the optimised representation
are a subset of those which require no secondary tag.
Output "MR_SECTAG_NONE_DIRECT_ARG" instead of "MR_SECTAG_NONE" in
RTTI structures when applicable.
compiler/add_pragma.m:
compiler/add_type.m:
compiler/bytecode_gen.m:
compiler/check_typeclass.m
compiler/code_info.m:
compiler/equiv_type.m:
compiler/export.m:
compiler/foreign.m:
compiler/hlds_code_util.m:
compiler/hlds_out_module.m:
compiler/inst_check.m:
compiler/ml_proc_gen.m:
compiler/ml_switch_gen.m:
compiler/ml_tag_switch.m:
compiler/ml_type_gen.m:
compiler/module_qual.m:
compiler/modules.m:
compiler/post_term_analysis.m:
compiler/post_typecheck.m:
compiler/recompilation.check.m:
compiler/recompilation.usage.m:
compiler/recompilation.version.m:
compiler/simplify.m:
compiler/structure_reuse.direct.choose_reuse.m:
compiler/switch_gen.m:
compiler/switch_util.m:
compiler/tag_switch.m:
compiler/term_norm.m:
compiler/type_ctor_info.m:
compiler/type_util.m:
compiler/unify_proc.m:
compiler/unused_imports.m:
compiler/xml_documentation.m:
Conform to changes.
Bump RTTI version number.
doc/reference_manual.texi:
Add commented out documentation for `direct_arg' attributes.
library/construct.m:
Handle MR_SECTAG_NONE_DIRECT_ARG in construct.construct/3.
library/private_builtin.m:
Add MR_SECTAG_NONE_DIRECT_ARG constant for Java for consistency,
though it won't be used.
runtime/mercury_grade.h:
Bump binary compatibility version number.
runtime/mercury_type_info.h:
Bump RTTI version number.
Add MR_SECTAG_NONE_DIRECT_ARG.
runtime/mercury_deconstruct.c:
runtime/mercury_deep_copy_body.h:
runtime/mercury_ml_expand_body.h:
runtime/mercury_table_type_body.h:
runtime/mercury_term_size.c:
runtime/mercury_unify_compare_body.h:
Handle MR_SECTAG_NONE_DIRECT_ARG in RTTI code.
tests/debugger/Mmakefile:
tests/debugger/chooser_tag_test.exp:
tests/debugger/chooser_tag_test.inp:
tests/debugger/chooser_tag_test.m:
tests/hard_coded/Mercury.options:
tests/hard_coded/Mmakefile:
tests/hard_coded/construct_test.exp:
tests/hard_coded/construct_test.m:
tests/hard_coded/direct_arg_cyclic1.exp:
tests/hard_coded/direct_arg_cyclic1.m:
tests/hard_coded/direct_arg_cyclic2.m:
tests/hard_coded/direct_arg_cyclic3.m:
tests/hard_coded/direct_arg_intermod1.exp:
tests/hard_coded/direct_arg_intermod1.m:
tests/hard_coded/direct_arg_intermod2.m:
tests/hard_coded/direct_arg_intermod3.m:
tests/hard_coded/direct_arg_parent.exp:
tests/hard_coded/direct_arg_parent.m:
tests/hard_coded/direct_arg_sub.m:
tests/invalid/Mmakefile:
tests/invalid/where_direct_arg.err_exp:
tests/invalid/where_direct_arg.m:
tests/invalid/where_direct_arg2.err_exp:
tests/invalid/where_direct_arg2.m:
Add test cases.
tests/invalid/ee_invalid.err_exp:
Update expected output.
|
||
|
Zoltan Somogyi
|
295415090e |
Convert almost all remaining modules in the compiler to use
Estimated hours taken: 6 Branches: main compiler/*.m: Convert almost all remaining modules in the compiler to use "$module, $pred" instead of "this_file" in error messages. In a few cases, the old error message was misleading, since it contained an incorrect, out-of-date or cut-and-pasted predicate name. tests/invalid/unresolved_overloading.err_exp: Update an expected output containing an updated error message. |
||
|
Julien Fischer
|
9f68c330f0 |
Change the argument order of many of the predicates in the map, bimap, and
Branches: main
Change the argument order of many of the predicates in the map, bimap, and
multi_map modules so they are more conducive to the use of state variable
notation, i.e. make the order the same as in the sv* modules.
Prepare for the deprecation of the sv{bimap,map,multi_map} modules by
removing their use throughout the system.
library/bimap.m:
library/map.m:
library/multi_map.m:
As above.
NEWS:
Announce the change.
Separate out the "highlights" from the "detailed listing" for
the post-11.01 NEWS.
Reorganise the announcement of the Unicode support.
benchmarks/*/*.m:
browser/*.m:
compiler/*.m:
deep_profiler/*.m:
extras/*/*.m:
mdbcomp/*.m:
profiler/*.m:
tests/*/*.m:
ssdb/*.m:
samples/*/*.m
slice/*.m:
Conform to the above change.
Remove any dependencies on the sv{bimap,map,multi_map} modules.
|
||
|
Zoltan Somogyi
|
1c3bc03415 |
Make the system compiler with --warn-unused-imports.
Estimated hours taken: 2 Branches: main, release Make the system compiler with --warn-unused-imports. browser/*.m: library/*.m: compiler/*.m: Remove unnecesary imports as flagged by --warn-unused-imports. In some files, do some minor cleanup along the way. |
||
|
Zoltan Somogyi
|
8a28e40c9b |
Add the predicates sorry, unexpected and expect to library/error.m.
Estimated hours taken: 2 Branches: main Add the predicates sorry, unexpected and expect to library/error.m. compiler/compiler_util.m: library/error.m: Move the predicates sorry, unexpected and expect from compiler_util to error. Put the predicates in error.m into the same order as their declarations. compiler/*.m: Change imports as needed. compiler/lp.m: compiler/lp_rational.m: Change imports as needed, and some minor cleanups. deep_profiler/*.m: Switch to using the new library predicates, instead of calling error directly. Some other minor cleanups. NEWS: Mention the new predicates in the standard library. |
||
|
Zoltan Somogyi
|
30aafc69a0 |
Split up three big compiler modules: llds_out.m, hlds_out.m (5000+ lines each)
Estimated hours taken: 12 Branches: main Split up three big compiler modules: llds_out.m, hlds_out.m (5000+ lines each) and deep_profiling.m (3000+ lines). Put the predicates in the resulting smaller modules into cohesive groups where possible. A few of the predicates in the original modules were unused; this diff deletes them. There are no algorithmic changes. compiler/llds_out_code_addr.m: New module containing the part of llds_out.m that outputs code addresses and labels. compiler/llds_out_data.m: New module containing the part of llds_out.m that outputs lvals, rvals and their components. compiler/llds_out_global.m: New module containing the part of llds_out.m that generates global static C data structures. compiler/llds_out_instr.m: New module containing the part of llds_out.m that outputs instructions compiler/llds_out_file.m: New module containing the top level part of llds_out.m, which coordinates the generation of a whole C source file. compiler/llds_out_util.m: New module containing the utility parts of llds_out.m. compiler/llds_out.m: Replace everything in this file with just the includes of the submodules that now have all its previous contents. compiler/hlds_llds.m: Move a predicate here from llds_out.m, since it is a utility predicate operating on a type defined here. compiler/rtti_out.m: Move a predicate here from llds_out.m, since it is a predicate generating output from a rtti type. compiler/hlds_out_mode.m: The part of hlds_out.m that deals with writing out insts and modes. compiler/hlds_out_goal.m: The part of hlds_out.m that deals with writing out goals. compiler/hlds_out_pred.m: The part of hlds_out.m that deals with writing out predicates and procedures. compiler/hlds_out_module.m: The part of hlds_out.m that deals with writing out module-wide tables. compiler/hlds_out_util.m: Parts of hlds_out.m that don't fit in anywhere else. compiler/hlds_out.m: Replace everything in this file with just the includes of the submodules that now have all its previous contents. compiler/simplify.m: compiler/hlds_goal.m: Move some insts from simplify.m to hlds_goal.m to allow hlds_out_goal.m to use them also. compiler/coverage_profiling.m: The part of deep_profiling.m that deals with coverage profiling. compiler/deep_profiling.m: Remove the code moved to coverage_profiling.m, and export the utility predicates needed by coverage_profiling.m. Remove the things moved to prog_data.m and hlds_goal.m. Put the predicates into a more logical order. compiler/hlds_goal.m: Move some predicates here from deep_profiling.m, since they belong here. compiler/prog_data.m: Move a type from deep_profiling.m here, since it belongs here. compiler/add_pragma.m: Add a predicate from llds_out.m that is used only here. compiler/*.m: Conform to the changes above. |
||
|
Zoltan Somogyi
|
d4bbcda309 |
Move all the frequently occurring layout structures and components of layout
Estimated hours taken: 40 Branches: main Move all the frequently occurring layout structures and components of layout structures into arrays where possible. By replacing N global variables holding individual layout structures or layout structure components with one global variable holding an array of them, we reduce the sizes of the symbol tables stored in object files, which should speed up both the C compiler and the linker. Measured on the modules of the library, mdbcomp and compiler directories compiled in grade asm_fast.gc.debug, this diff reduces the size of the generated C source files by 7.8%, the size of the generated object files by 10.4%, and the number of symbols in the symbol tables of those object files by a whopping 42.8%. (These improvements include, and are not on top of, the improvements in my previous similar diff.) runtime/mercury_stack_layout.h: Each label layout structure has information about the type and location of every variable that is live at that label. We store this information in three arrays: an array of pseudo-typeinfos giving the types of all these variables, and two arrays MR_ShortLvals and MR_LongLvals respectively giving their locations. (Most of the time, the location's encoded form fits into one byte (the MR_ShortLval) but sometimes it needs more bits (this is when we use MR_LongLval)). We used to store these three arrays, whose elements are different types, in a single occurrence-specific common structure, one after the other, with a cumbersome mechanism being required to access them. We now store them as segments of three separate arrays, of pseudo-typeinfos, MR_ShortLvals and MR_LongLvals respectively. This makes access simpler and faster (which will matter more to any accurate garbage collector than it does to the debugger). It also allows more scope for compression, since reusing an existing segment of one of the three arrays is easier than reusing an entire common structure, which would require the equivalent of exact matches on all three arrays. Since most label layout structures that have information about variables can encode the variables' locations using only MR_ShortLvals, create a version of the label layout structure type that omits the field used to record the whereabouts of the long location descriptors. Add macros now generated by the compiler to initialize layout structures. Simplify a one-field struct. runtime/mercury_grade.h: Increment the binary compatibility version number for debuggable executables, since .c and .o files from before and after the change to label layout structures are NOT compatible. runtime/mercury_type_info.h: Fix some binary-compatibility-related bit rot. runtime/mercury_misc.h: Move here the existing macros used by the compiler when generating references to layout arrays, and add new ones. runtime/mercury_goto.h: Delete the macros moved to mercury_misc.h. Conform to the changes in mercury_stack_layout.h. runtime/Mmakefile: Prevent the unnecessary rebuilding of mercury_conf.h. runtime/mercury_accurate_gc.c: runtime/mercury_agc_debug.c: runtime/mercury_layout_util.c: runtime/mercury_stack_trace.c: runtime/mercury_types.h: trace/mercury_trace.c: trace/mercury_trace_vars.c: Conform to the changes in mercury_stack_layout.h. runtime/mercury_wrapper.c: Improve the debug support a bit. runtime/mercury_engine.h: Fix style. compiler/layout.m: Make the change described at the top. Almost all layout structures are now in arrays. The only exceptions are those that occur rarely, and proc layouts, whose names need to be derivable from the name of the procedure itself. Instead of having a single type "layout_data" that can represent different kinds of single global variables (not array slots), have different kinds for different purposes. This makes the code clearer and allows traversals that do not have to skip over inapplicable kinds of layout structures. compiler/layout_out.m: Output the new arrays. compiler/stack_layout.m: Generate the new arrays. Previously, an individual term generated by stack_layout.m could represent several components of a layout structure, with the components separated by layout_out.m. We now do the separation in stack_layout.m itself, adding each component to the array to which it belongs. Instead of passing around a single stack_layout_info structure, pass around several smaller one. This is preferable, since I found out the hard way that including everything in one structure would give the structure 51 fields. Most parts of the module work with only one or two of these structures, which makes their role clearer. Cluster related predicates together. compiler/options.m: doc/user_guide.texi: Add an option that control whether stack_layout.m will attempt to compress the layout arrays that can meaningfully be comressed. compiler/llds.m: Remove the old distinction between a data_addr and a data_name, replacing both types with a single new one: data_id. Since different kinds of data_names were treated differently in many places, the distinction in types (which was intended to allow us to process data_addrs that wrapped data_names differently from other kinds of data_addrs) wasn't buying us anything anymore. The new data_id type allows for the possibility that the code generator wants to generate a reference to an address it does not know yet, because it is a slot in a layout array, and the slot has not been allocated yet. Add the information from which the new layout array structures will be generated to the LLDS. compiler/llds_out.m: Call layout_out.m to output the new layout arrays. Adapt the decl_id type to the replacement of data_addrs by data_ids. Don't both keeping track of the have-vs-have-not-declared status of structures that are always declared at the start. When writing out a data_addr, for some kinds of data_addr, llds_out.m would write out the name of the relevant variable, while for some other kinds, it would write out its address. This diff separates out those those things into separate predicates, each of which behaves consistently. compiler/mercury_compile_llds_back_end.m: Convey the intended contents of the new layout arrays from stack_layout.m to llds_out.m. compiler/continuation_info.m: Add a type required by the way we now generate proc_static structures for deep profiling. compiler/hlds_rtti.m: Add distinguishing prefixes to the field names of the rtti_proc_label type. compiler/code_info.m: compiler/code_util.m: compiler/erl_rtti.m: compiler/exprn_aux.m: compiler/global_data.m: compiler/ll_pseudo_type_info.m: compiler/ml_code_util.m: compiler/opt_debug.m: compiler/proc_gen.m: compiler/prog_rep.m: compiler/rtti_out.m: compiler/unify_gen.m: Conform to the changes above. tests/debugger/declarative/track_through_catch.exp: Expect procedures to be listed in the proper order. tests/EXPECT_FAIL_TESTS.asm_fast.gc.debug: tests/EXPECT_FAIL_TESTS.asm_fast.gc.profdeep: Add these files to ignore expected failues in these grades. |
||
|
Zoltan Somogyi
|
4ebe3d0d7e |
Stop storing globals in the I/O state, and divide mercury_compile.m
Estimated hours taken: 60 Branches: main Stop storing globals in the I/O state, and divide mercury_compile.m into smaller, more cohesive modules. (This diff started out as doing only the latter, but it became clear that this was effectively impossible without the former, and the former ended up accounting for the bulk of the changes.) Taking the globals out of the I/O state required figuring out how globals data flowed between pieces of code that were often widely separated. Such flows were invisible when globals could be hidden in the I/O state, but now they are visible, because the affected code now passes around globals structures explicitly. In some cases, the old flow looked buggy, as when one job invoked by mmc --make could affect the globals value of its parent or the globals value passed to the next job. I tried to fix such problems when I saw them. I am not 100% sure I succeeded in every case (I may have replaced old bugs with new ones), but at least now the flow is out in the open, and any bugs should be much easier to track down and fix. In most cases, changes the globals after the initial setup are intended to be in effect only during the invocation of a few calls. This used to be done by remembering the initial values of the to-be-changed options, changing their values in the globals in the I/O state, making the calls, and restoring the old values of the options. We now simply create a new version of the globals structure, pass it to the calls to be affected, and then discard it. In two cases, when discovering reasons why (1) smart recompilation should not be done or (2) item version numbers should not be generated, the record of the discovery needs to survive this discarding. This is why in those cases, we record the discovery by setting a mutable attached to the I/O state. We use pure code (with I/O states) both to read and to write the mutables, so this is no worse semantically than storing the information in the globals structure inside the I/O state. (Also, we were already using such a mutable for recording whether -E could add more information.) In many modules, the globals information had to be threaded through several predicates in the module. In some places, this was made more difficult by predicates being defined by many clauses. In those cases, this diff converts those predicates to using explicit disjunctions. compiler/globals.m: Stop storing the globals structure in the I/O state, and remove the predicates that accessed it there. Move a mutable and its access predicate here from handle_options.m, since here is when the mutables treated the same way are. In a couple of cases, the value of an option is available in a mutable for speed of access from inside performance-critical code. Set the values of those mutables from the option when the processing of option values is finished, not when it is starting, since otherwise the copies of each option could end up inconsistent. Validate the reuse strategy option here, since doing it during ctgc analysis (a) is too late, and (b) would require an update to the globals to be done at an otherwise inconvenient place in the code. Put the reuse strategy into the globals structure. Two fields in the globals structure were unused. One (have_printed_usage) was made redundant when the one predicate that used it itself became unused; the other (source_file_map) was effectively replaced by a mutable some time ago. Delete these fields from the globals. Give the fields of the globals structure a distinguishing prefix. Put the type declarations, predicate declarations and predicate definitions in a consistent order. compiler/source_file_map.m: Record this module's results only in the mutable (it serves as a cache), not in globals structure. Use explicitly passed globals structure for other purposes. compiler/handle_options.m: Rename handle_options as handle_given_options, since it does not process THE options to the program, but the options it is given, and even during the processing of a single module, it can be invoked up the three times in a row, each time being given different options. (It was up to four times in a row before this diff.) Make handle_given_options explicitly return the globals structure it creates. Since it does not take an old global structure as input and globals are not stored in the I/O state, it is now clear that the globals structure it returns is affected only by the default values of the options and the options it processes. Before this diff, in the presence of errors in the options, handle_options *could* return (implicitly, in the I/O state) the globals structure that happened to be in the I/O state when it was invoked. Provide a separate predicate for generating a dummy globals based only on the default values of options. This allows by mercury_compile.m to stop abusing a more general-purpose predicate from handle_options.m, which we no longer export. Remove the mutable and access predicate moved to globals.m. compiler/options.m: Document the fact that two options, smart_recompilation and generate_item_version_numbers, should not be used without seeing whether the functionalities they call for have been disabled. compiler/mercury_compile_front_end.m: compiler/mercury_compile_middle_passes.m: compiler/mercury_compile_llds_back_end.m: compiler/mercury_compile_mlds_back_end.m: compiler/mercury_compile_erl_back_end.m: New modules carved out of the old mercury_compile.m. They each cover exactly the areas suggested by their names. Each of the modules is more cohesive than the old mercury_compile.m. Their code is also arranged in a more logical order, with predicates representing compiler passes being defined in the order of their invocation. Some of these modules export predicates for use by their siblings, showing the dependencies between the groups of passes. compiler/top_level.m: compiler/notes/compiler_design.html: Add the new modules. compiler/mark_static_terms.m: Move this module from the ml_backend package to the hlds package, since (a) it does not depend on the MLDS in any way, and (b) it is also needed by a compiler pass (loop invariants) in the middle passes. compiler/hlds.m: compiler/ml_backend.m: compiler/notes/compiler_design.html: Reflect mark_static_terms.m's change of package. compiler/passes_aux.m: Move the predicates for dumping out the hLDS here from mercury_compile.m, since the new modules also need them. Look up globals in the HLDS, not the I/O state. compiler/hlds_module.m: Store the prefix (common part) of HLDS dump file names in the HLDS itself, so that the code moved to passes_aux.m can figure out the file name for a HLDS dump without doing system calls. Give the field names of some structures prefixes to avoid ambiguity. compiler/mercury_compile.m: Remove the code moved to the other modules. This module now looks after only option handling (such as deciding whether to generate .int3 files, .int files, .opt files etc), and the compilation passes up to and including the creation of the first version of the HLDS. Everything after that is subcontracted to the new modules. Simplify and make explicit the flow of globals information. When invoking predicates that could disable smart recompilation, check whether they have done so, and if yes, update the globals accordingly. When compiling via gcc, we need to link into the executable the object files of any separate C files we generate for C code foreign_procs, which we cannot translate into gcc's internal structures without becoming a C compiler as well as a Mercury compiler. Instead of adding such files to the accumulating option for extra object files in the globals structure, we return their names using the already existing mechanism we have always used to link the object files of fact tables into the executable. Give several predicates more descriptive names. Put predicates in a more logical order. compiler/make.m: compiler/make.dependencies.m: compiler/make.module_target.m: compiler/make.module_dep_file.m: compiler/make.program_target.m: compiler/make.util.m: Require callers to supply globals structures explicitly, not via the I/O state. Afterward pass them around explicitly, passing modified versions to mercury_compile.m when invoking it with module- and/or task-specific options. Due the extensive use of partial application for higher order code in these modules, passing around the globals structures explicitly is quite tricky here. There may be cases where a predicate uses an old globals structure it got from a closure instead of the updated module- and/or task-specific globals it should be using, or vice versa. However, it is just as likely that, this diff fixes old problems by preventing the implicit flow of updated-only-for-one-invocation globals structures back to the original invoking context. Although I have tried to be careful about this, it is also possible that in some places, the code is using an updated-for-an-invocation globals structure in some but not all of the places where it SHOULD be used. compiler/c_util.m: compiler/compile_target_code.m: compiler/compiler_util.m: compiler/error_util.m: compiler/file_names.m: compiler/file_util.m: compiler/ilasm.m: compiler/ml_optimize.m: compiler/mlds_to_managed.m: compiler/module_cmds.m: compiler/modules.m: compiler/options_file.m: compiler/pd_debug.m: compiler/prog_io.m: compiler/transform_llds.m: compiler/write_deps_file.m: Require callers to supply globals structures explicitly, not via the I/O state. In some cases, the explicit globals structure argument allows a predicate to dispense with the I/O states previously passed to it. In some modules, rename some predicates, types and/or function symbols to avoid ambiguity. compiler/read_modules.m: Require callers to supply globals structures explicitly, not via the I/O state. Record when smart recompilation and the generation of item version numbers should be disabled. compiler/opt_debug.m: compiler/process_util.m: Require callers to supply the needed options explicitly, not via the globals in the I/O state. compiler/analysis.m: compiler/analysis.file.m: compiler/mmc_analysis.m: Make the analysis framework's methods take their global structures as explicit arguments, not as implicit data stored in the I/O state. Stop using `with_type` and `with_inst` declarations unnecessarily. Rename some predicates to avoid ambiguity. compiler/hlds_out.m: compiler/llds_out.m: compiler/mercury_to_mercury.m: compiler/mlds_to_c.m: compiler/mlds_to_java.m: compiler/optimize.m: Make these modules stop accessing the globals from the I/O state. Do this by requiring the callers of their top predicates to explicitly supply a globals structure. To compensate for the cost of having to pass around a representation of the options, look up the values of the options of interest just once, to make further access much faster. (In the case of mlds_to_c.m, the code already did much of this, but it still had a few accesses to globals in the I/O state that this diff eliminates.) If the module exports a predicate that needs these pre-looked-up options, then export the type of this data structure and its initialization function. compiler/frameopt.m: Since this module needs only one option from the globals, pass that option instead of the globals. compiler/accumulator.m: compiler/add_clause.m: compiler/closure_analysis.m: compiler/complexity.m: compiler/deforest.m: compiler/delay_construct.m: compiler/elds_to_erlang.m: compiler/exception_analysis.m: compiler/fact_table.m: compiler/intermod.m: compiler/mode_constraints.m: compiler/mode_errors.m: compiler/pd_util.m: compiler/post_term_analysis.m: compiler/recompilation.usage.m: compiler/size_prof.usage.m: compiler/structure_reuse.analysis.m: compiler/structure_reuse.direct.choose_reuse.m: compiler/structure_reuse.direct.m: compiler/structure_sharing.analysis.m: compiler/tabling_analysis.m: compiler/term_constr_errors.m: compiler/term_constr_fixpoint.m: compiler/term_constr_initial.m: compiler/term_constr_main.m: compiler/term_constr_util.m: compiler/trailing_analysis.m: compiler/trans_opt.m: compiler/typecheck_info.m: Look up globals information from the HLDS, not the I/O state. Conform to the changes above. compiler/gcc.m: compiler/maybe_mlds_to_gcc.pp: compiler/mlds_to_gcc.m: Look up globals information from the HLDS, not the I/O state. Conform to the changes above. Convert these modules to our current programming style. compiler/termination.m: Look up globals information from the HLDS, not the I/O state. Conform to the changes above. Report some warnings with error_specs, instead of immediately printing them out. compiler/export.m: compiler/il_peephole.m: compiler/layout_out.m: compiler/rtti_out.m: compiler/liveness.m: compiler/make_hlds.m: compiler/make_hlds_passes.m: compiler/mlds_to_il.m: compiler/mlds_to_ilasm.m: compiler/recompilation.check.m: compiler/stack_opt.m: compiler/superhomogeneous.m: compiler/tupling..m: compiler/unneeded_code.m: compiler/unused_args.m: compiler/unused_import.m: compiler/xml_documentation.m: Conform to the changes above. compiler/equiv_type_hlds.m: Give the field names of a structure prefixes to avoid ambiguity. Stop using `with_type` and `with_inst` declarations unnecessarily. compiler/loop_inv.m: compiler/pd_info.m: compiler/stack_layout.m: Give the field names of some structures prefixes to avoid ambiguity. compiler/add_pragma.m: Add notes. compiler/string.m: NEWS: Add a det version of remove_suffix, for use by new code above. |
||
|
Zoltan Somogyi
|
d69ba1a1f0 |
Include the type_ctor in cons_ids for user-defined types.
Estimated hours taken: 32 Branches: main Include the type_ctor in cons_ids for user-defined types. The intention is two-fold: - It prepares for a future in which we allow more than one function symbol to with the same name to be defined in a module. - It makes the HLDS code more self-contained. In many places, processing construction and deconstruction unifications required knowing which type the cons_id belongs to, but until now, code couldn't know that unless it kept track of the type of the variable unified with the cons_id. With this diff, user-defined cons_ids are represented as cons(SymName, Arity, TypeCtor) The last field is filled in during post-typecheck. After that time, any module qualification in the SymName (which may initially be partial) is redundant, since it is also available in the TypeCtor. In the future, we could make all those SymNames be just unqualified(_) at that time. We could also replace the current maps in HLDS type definitions with full cons_id keys with just name/arity keys (since the module qualifier is a given for any given type definition), we could also support partially qualified cons_ids in source code using a map from name/arity pairs to a list of all the type_ctors that have function symbols with that name/arity, instead of our current practice of inserting all possible partially module qualified version of every cons_id into a single giant table, and we could do the same thing with the field names table. This diff also separates tuples out from user-defined types, since in many respects they are different (they don't have a single type_ctor, for starters). It also separates out character constants, since they were alreay treated specially in most places, though not in some places where they *ought* to have been treated specially. Take the opportunity to give some other cons_ids better names. compiler/prog_data.m: Make the change described above, and document it. Put the implementations of the predicates declared in each part of this module next to the declarations, instead of keeping all the code until the very end (where it was usually far from their declarations). Remove three predicates with identical definitions from inst_match.m, inst_util.m and mode_constraints.m, and put the common definition in prog_data.m. library/term_io.m: Add a new predicate that is basically a reversible version of the existing function espaced_char, since the definition of char_consts needs reversibilty. compiler/post_typecheck.m: For functors of user-defined types, record their type_ctor. For tuples and char constants, record them as such. compiler/builtin_lib_types.m: compiler/parse_tree.m: compiler/notes/compiler_design.html: New module to centralize knowledge about builtin types, specially handled library types, and their function symbols. Previously, the stuff now in this module used to be in several different places, including prog_type.m and stm_expand.m, and some of it was duplicated. mdbcomp/prim_data.m: Add some predicates now needed by builtin_lib_types.m. compiler/builtin_ops.m: Factor out some duplicated code. compiler/add_type.m: Include the relevant type_ctors in the cons_ids generated in type definitions. compiler/hlds_data.m: Document an existing type better. Rename a cons_tag in sync with its corresponding cons_id. Put some declarations into logical order. compiler/hlds_out.m: Rename a misleadingly-named predicate. compiler/prog_ctgc.m: compiler/term_constr_build.m: Add XXXs for questionable existing code. compiler/add_clause.m: compiler/add_heap_ops.m: compiler/add_pragma.m: compiler/add_pred.m: compiler/add_trail_ops.m: compiler/assertion.m: compiler/bytecode_gen.m: compiler/closure_analysis.m: compiler/code_info.m: compiler/complexity.m: compiler/ctgc_selector.m: compiler/dead_proc_elim.m: compiler/deep_profiling.m: compiler/delay_partial_inst.m: compiler/dependency_graph.m: compiler/det_analysis.m: compiler/det_report.m: compiler/distance_granularity.m: compiler/erl_rtti.m: compiler/erl_unify_gen.m: compiler/export.m: compiler/field_access.m: compiler/foreign.m: compiler/format_call.m: compiler/hhf.m: compiler/higher_order.m: compiler/hlds_code_util.m: compiler/hlds_desc.m: compiler/hlds_goal.m: compiler/implementation_defined_literals.m: compiler/inst_check.m: compiler/inst_graph.m: compiler/inst_match.m: compiler/inst_util.m: compiler/instmap.m: compiler/intermod.m: compiler/interval.m: compiler/lambda.m: compiler/lco.m: compiler/make_tags.m: compiler/mercury_compile.m: compiler/mercury_to_mercury.m: compiler/middle_rec.m: compiler/ml_closure_gen.m: compiler/ml_code_gen.m: compiler/ml_code_util.m: compiler/ml_switch_gen.m: compiler/ml_type_gen.m: compiler/ml_unify_gen.m: compiler/ml_util.m: compiler/mlds_to_c.m: compiler/mlds_to_java.m: compiler/mode_constraints.m: compiler/mode_errors.m: compiler/mode_ordering.m: compiler/mode_util.m: compiler/modecheck_unify.m: compiler/modes.m: compiler/module_qual.m: compiler/polymorphism.m: compiler/prog_ctgc.m: compiler/prog_event.m: compiler/prog_io_util.m: compiler/prog_mode.m: compiler/prog_mutable.m: compiler/prog_out.m: compiler/prog_type.m: compiler/prog_util.m: compiler/purity.m: compiler/qual_info.m: compiler/rbmm.add_rbmm_goal_infos.m: compiler/rbmm.execution_path.m: compiler/rbmm.points_to_analysis.m: compiler/rbmm.region_transformation.m: compiler/recompilation.usage.m: compiler/rtti.m: compiler/rtti_out.m: compiler/rtti_to_mlds.m: compiler/simplify.m: compiler/simplify.m: compiler/special_pred.m: compiler/ssdebug.m: compiler/stack_opt.m: compiler/stm_expand.m: compiler/stratify.m: compiler/structure_reuse.direct.detect_garbagem: compiler/superhomoegenous.m: compiler/switch_detection.m: compiler/switch_gen.m: compiler/switch_util.m: compiler/table_gen.m: compiler/term_constr_build.m: compiler/term_norm.m: compiler/try_expand.m: compiler/type_constraints.m: compiler/type_ctor_info.m: compiler/type_util.m: compiler/typecheck.m: compiler/typecheck_errors.m: compiler/unify_gen.m: compiler/unify_proc.m: compiler/unify_modes.m: compiler/untupling.m: compiler/unused_imports.m: compiler/xml_documentation.m: Minor changes, mostly to ignore the type_ctor in cons_ids in places where it is not needed, take the type_ctor from the cons_id in places where it is more convenient, conform to the new names of some cons_ids, conform to the changes in hlds_out.m, and/or add now-needed imports of builtin_lib_types.m. In some places, the handling previously applied to cons/2 (which included tuples and character constants as well as user-defined function symbols) is now applied only to user-defined function symbols or to user-defined function symbols and tuples, as appropriate, with character constants being handled more like the other kinds of constants. In inst_match.m, rename a whole bunch of predicates to avoid ambiguities. In prog_util.m, remove two predicates that did almost nothing yet were far too easy to misuse. |
||
|
Peter Wang
|
58537374bb |
Predefine MR_TypeClassConstaint_NStructs for type classes with arities up
Branches: main Predefine MR_TypeClassConstaint_NStructs for type classes with arities up to 10 (was 5). Although low-level C grades would define the structures for higher arities on demand, this is not currently implemented for high-level C grades. runtime/mercury_type_info.h: Define MR_TypeClassConstaint_NStructs for N =< 10. compiler/rtti_out.m: Bump max_always_declared_arity_type_class_constraint. tests/valid/Mmakefile: tests/valid/typeclass_constraint_arity.m: Add test case. |
||
|
Zoltan Somogyi
|
bbe6de9387 |
Make tabling statistics much more useful, and avoid core dumps while printing
Estimated hours taken: 30 Branches: main Make tabling statistics much more useful, and avoid core dumps while printing them (thereby fixing bug #23 in Mantis). runtime/mercury_tabling.h: Expand the set of statistics we can gather about tabling (starting to gather information about memory consumption). Put the statistics into logical groups: enum functors, non-enum du functors, hash tables, and the expandable table used for I/O tabling. For du types and polymorphic types, we can gather different statistics for different parts of a term. We used to use one field to gather more than one related kind of count, but these are now separated. Group the MR_ProcTableInfo struct fields relating to statistics into substructures. The substructures describing a call or answer table have their own substructures giving the statistics about the individual steps; put these in parallel with the descriptions of the steps themselves. Put those substructures into two-element arrays to allow them to be handled uniformly. runtime/mercury_types.h: Add the typedefs needed by mercury_tabling.h. runtime/mercury_tabling_macros.h: Conform to the modified classification of tabling categories. runtime/mercury_tabling.c: runtime/mercury_hash_lookup_or_add_body.h: runtime/mercury_table_int_fix_index_body.h: runtime/mercury_table_int_start_index_body.h: runtime/mercury_table_type_body.h: runtime/mercury_tabling_stats_defs.h: runtime/mercury_tabling_stats_nodefs.h: runtime/mercury_tabling_stats_undefs.h: Gather the modified and expanded set of statistics. Use more meaningful names (with MR_ prefixes) for macros. library/table_builtin.m: library/table_statistics.m: Move the material relating to statistics that used to be in table_builtin.m to the new module table_statistics.m. This is necessary because table_builtin.m is deliberately not included in the library documentation, yet users needs to know how to print out and interpret tabling statistics. Expand the code for handling statistics, adding predicates for printing them together with (hopefully explanatory) labels. Make it all work with the new runtime data structures. library/library.m: Include the new library module. mdbcomp/prim_data.m: Provide a way to refer to the tabling statistics module. compiler/hlds_pred.m: Add utility function to compute the kind of statistics we gather for a given kind of tabling step. compiler/rtti.m: Change the compiler's representation of tabling's runtime data structures to conform to the change to runtime/mercury_tabling.h. Replace a bool with a purpose-designed type (is_array), and add other similar types (call_or_answer_table, curr_or_prev_table). compiler/rtti_out.m: compiler/llds_out.m: Output the updated data structures. compiler/ml_code_gen.m: Generate the updated data structures. compiler/table_gen.m: Update the C code we generate to gather statistics. Gather the extra information needed by the updated data structures in the compiler (to fill in the updated data structures in the runtime). compiler/mlds.m: Add a new kind of constant, one that has just a name. This allows ml_code_gen.m to avoid lying about the type of constants of enum tyes (such as the kind of tabling stats we gather for a step). compiler/modules.m: Import the new table_statistics module automatically if any tabling pragma asks for statistics. compiler/add_pragma.m: compiler/hlds_out.m: compiler/llds.m: compiler/mlds_to_c.m: compiler/mlds_to_gcc.m: compiler/mlds_to_il.m: compiler/mlds_to_java.m: compiler/mlds_to_managed.m: trace/mercury_trace_cmd_developer.c: Conform to the changes above. tests/tabling/fib_stats.{m,exp}: tests/tabling/specified_stats.{m,exp}: New test cases to the test tabling statistics. tests/tabling/Mmakefile: Enable the new test cases. tests/tabling/fib.m: tests/tabling/specified.m: Clean up the code of these test cases, which were used as the basis of the new test cases. |
||
|
Julien Fischer
|
c204fac461 |
Replace some if-then-elses with switches.
Estimated hours taken: 0.2 Branches: main compiler/add_type.m: compiler/recompilation.usage.m: compiler/rtti_out.m: Replace some if-then-elses with switches. Use svmulti_map.set/4 in preference to multi_map.set/4 in the presence of state variables. |
||
|
Julien Fischer
|
2030b891e7 |
Fix an XXX left over from the change that introduced foreign enumerations.
Estimated hours taken: 1.5 Branches: main Fix an XXX left over from the change that introduced foreign enumerations. The representation of foreign tags in both the HLDS and MLDS needs to have information about which foreign language the tag belongs attached to it. In addition to fixing the above XXX (in intermod.m), this also improves sanity checking of foreign tags. compiler/hlds_data.m: Extend the enum_or_dummy type to hold the foreign language to which a foreign enumeration applies. Extend the representation of foreign tags to include their foreign language. compiler/hlds_out.m: When writing out the comment describing a d.u. type include the foreign language in the description if the type is a foreign enumeration. compiler/intermod.m: Instead of looking up the target language when writing foreign enumeration pragmas to optimization interfaces, look up the language of the foreign enumeration in the type definition. compiler/mlds.m: Extend the MLDS representation of foreign tags to include their language. (There is no point making a similar change to the LLDS since the language for that will always be C.) compiler/rtti.m: Include a foreign enumeration's language in the representation of it's RTTI data. compiler/add_pragma.m: compiler/bytecode_gen.m: compiler/erl_rtti.m: compiler/export.m: compiler/ml_switch_gen.m: compiler/ml_type_gen.m: compiler/ml_unify_gen.m: compiler/ml_util.m: compiler/mlds_to_c.m: compiler/mlds_to_gcc.m: compiler/mlds_to_il.m: compiler/mlds_to_java.m: compiler/mlds_to_managed.m: compiler/rtti_out.m: compiler/rtti_to_mlds.m: compiler/switch_util.m: compiler/type_ctor_info.m: compiler/type_util.m: compiler/unify_gen.m: compiler/unify_proc.m: Conform to the above change and strengthen sanity checking of foreign tags; in particular make sure that each backend aborts if passed foreign tags of a language that cannot be handled by that backend. |
||
|
Julien Fischer
|
1fac629e6d |
Add support for foreign enumerations to Mercury.
Estimated hours taken: 50
Branches: main
Add support for foreign enumerations to Mercury. These allow the
programmer to assign foreign language values as the representation of
enumeration constructors.
e.g.
:- type status
---> optimal
; infeasible
; unbounded
; unknown.
:- pragma foreign_enum("C", status/0, [
optimal - "STATUS_OPTIMAL",
infeasible - "STATUS_INFEASIBLE",
unbounded - "STATUS_UNBOUNDED",
unknown - "STATUS_UNKNOWN"
]).
The advantage of this is that when values of type status/0 are passed to
foreign code (C in this case) no translation is necessary. This should
simplify the task of writing bindings to foreign language libraries.
Unification and comparison for foreign enumerations are the usual
unification and comparison for enumeration types, except that the default
ordering on them is determined by the foreign representation of the
constructors. User-defined equality and comparison also work for foreign
enumeration types.
In order to implement foreign enumerations we have to introduce two
new type_ctor representations. The existing ones for enum type do not
work since they use the value of an enumeration constructor to perform
table lookups in the RTTI data structures. For foreign enumerations
we need to perform a linear search at the corresponding points. This
means that some RTTI operations related to deconstruction are more
expensive.
The dummy type optimisation is not applied to foreign enumerations as
the code generators currently initialise the arguments of non-builtin
dummy type foreign_proc arguments to zero. For unit foreign enumerations
they should be initialised to the correct foreign value. (This is could be
implemented but in practice it's probably not going to be worth it.)
Currently, foreign enumerations are only supported by the C backends.
compiler/prog_io_pragma.m:
Parse foreign_enum pragmas.
Generalise the code used to parse association lists of sym_names
and strings since this is now used by the code to parse foreign_enum
pragmas as well as that for foreign_export_enum pragmas.
Fix a typo: s/foreign_expor_enum/foreign_export_enum/
compiler/prog_item.m:
Represent foreign_enum pragmas in the parse tree.
compiler/prog_type.m:
Add a new type category for foreign enumerations.
compiler/modules.m:
Add any foreign_enum pragmas for enumeration types defined in the
interface of a module to the interface files.
Output foreign_import_module pragmas in the interface file
if any foreign_enum pragmas are included in it. This ensures that
the contents that any foreign declarations that are needed by the
foreign_enum pragmas are visible.
compiler/make_hlds_passes.m:
compiler/add_pragma.m:
Add pragma foreign_enum items to the HLDS after all the types
have been added. As they are added, error check them.
Change the constructor tag values of foreign enum types to their
foreign values.
compiler/module_qual.m:
Module qualify pragma foreign_enum items.
compiler/mercury_to_mercury.m:
Output foreign_enum pragmas.
Generalise some of the existing code for writing out association
lists in foreign_export_enum pragmas for use with foreign_enum
pragmas as well.
compiler/hlds_data.m:
Add the alternative `is_foreign_type' to the type enum_or_dummy/0.
Add new type of cons_tag, foreign_tag, whose values are directly
embedded in the target language.
compiler/intermod.m:
Write out any foreign_enum pragmas for opt_exported types.
(The XXX concerning attaching language information to foreign tags
will be addressed in a subsequent change.)
compiler/llds.m:
compiler/mlds.m:
Support new kinds of rval constants: llconst_foreign and
mlconst_foreign respectively. Both of these represent tag values
as strings that are intended to be directly embedded in the target
language.
compiler/llds_out.m:
Add code to write out the new kind of rval_const.
s/Integer/MR_Integer/ in a spot.
s/Float/MR_Float/ in a spot.
compiler/rtti.m:
compiler/rtti_out.m:
compiler/rtti_to_mlds.m:
compiler/type_ctor_info.m:
Add support the RTTI required by foreign enums.
compiler/switch_util.m:
Handle switches on foreign_enums as-per normal enumerations.
compiler/table_gen.m:
Tabling of foreign_enums is also like normal enumerations.
compiler/type_util.m:
Add a predicate that tests whether a type is a foreign enumeration.
compiler/unify_gen.m:
compiler/unify_proc.m:
compiler/ml_unify_gen.m:
Handle unification and comparison of foreign enumeration values.
They are treated like normal enumerations for the purposes of
implementing these operations.
compiler/ml_type_gen.m:
Handle foreign enumerations when generating the MLDS representation
of enumerations.
compiler/ml_util.m:
Add a function to create an initializer for an object with a
foreign tag.
compiler/mlds_to_c.m:
Handle mlconst_foreign/1 rval constants.
compiler/bytecode_gen.m:
compiler/dupproc.m:
compiler/erl_rtti.m:
compiler/exception_analysis.m:
compiler/export.m:
compiler/exprn_aux.m:
compiler/global_data.m:
compiler/hlds_out.m:
compiler/higher_order.m:
compiler/inst_match.m:
compiler/jumpopt.m:
compiler/llds_to_x86_64.m:
compiler/ml_code_util.m:
compiler/mlds_to_gcc.m:
compiler/mlds_to_il.m:
compiler/mlds_to_java.m:
compiler/mlds_to_managed.m:
compiler/opt_debug.m:
compiler/opt_util.m:
compiler/polymorphism.m:
compiler/recompilation.version.m:
compiler/term_norm.m:
compiler/trailing_analysis.m:
Conform to the above changes.
doc/reference_manual.texi:
Document the new pragma.
Fix some typos: s/pramga/pragma/, s/behavior/behaviour/
library/construct.m:
Handle the two new type_ctor reps.
Break an over-long line.
library/rtti_implementation.m:
Support the two new type_ctor reps.
(XXX The Java versions of some of this cannot be implemented until
support for foreign enumerations is added to mlds_to_java.m.)
Reformat the inst usereq/0 and extend it to include foreign enums.
runtime/mercury_type_info.h:
Add two new type_ctor reps. One for foreign enumerations and
another for foreign enumerations with user equality.
Define new types (and extend existing ones) in order to support
RTTI for foreign enumerations.
runtime/mercury_unify_compare_body.h:
Implement generic unify and compare for foreign enumerations.
(It is the same as that for regular enumerations.)
runtime/mercury_construct.[ch]:
runtime/mercury_deconstruct.h:
Handle (de)construction of foreign enumeration values.
runtime/mercury_deep_copy_body.h:
Implement deep copy for foreign enumerations.
runtime/mercury_table_type_body.h:
runtime/mercury_term_size.c:
Handle the new type_ctor representations.
java/runtime/ForeignEnumFunctorDesc.java:
Add a Java version of the MR_ForeignEnumFuntorDesc structure.
(Note: this is untested, as the java grade runtime doesn't work
anyway.)
java/runtime/TypeFunctors.java:
Add a constructor method for foreign enumerations.
(Likewise, untested.)
NEWS:
Announce pragma foreign_enum.
vim/syntax/mercury.vim:
Highlight the new pragma appropriately.
tests/hard_coded/.cvsignore:
Ignore executables generated by the new tests.
Ignore a bunch of other files create by the Mercury compiler.
tests/hard_coded/Mmakefile:
tests/hard_coded/foreign_enum_rtti.{m,exp}:
Test RTTI for foreign enumerations.
tests/hard_coded/foreign_enum_dummy.{m,exp}:
Check that dummy type optimisation is disabled for foreign
enumerations.
tests/hard_coded/Mercury.options:
tests/hard_coded/foreign_enum_mod1.{m,exp}:
tests/hard_coded/foreign_enum_mod2.m:
Test that foreign_enum pragmas are hoisted into interface files
and that they are handled correctly in optimization interfaces.
tests/invalid/Mercury.options:
tests/invalid/Mmakefile:
tests/invalid/foreign_enum_import.{m,err_exp}:
tests/invalid/foreign_enum_invalid.{m,err_exp}:
Test that errors in foreign_enum pragmas are reported.
tests/tabling/Mmakefile:
tests/hard_coded/table_foreign_enum.{m,exp}:
Test case for tabling of foreign enumerations.
|
||
|
Peter Wang
|
c948a2815d |
s/rtti_data_to_name/rtti_data_to_id in comments.
Estimated hours taken: 0 Branches: main compiler/rtti_out.m: compiler/rtti_to_mlds.m: s/rtti_data_to_name/rtti_data_to_id in comments. |
||
|
Simon Taylor
|
6de3b102ba |
Add support for deconstructing by functor number rather than name,
Estimated hours taken: 20 Branches: main Add support for deconstructing by functor number rather than name, for use by write_binary. library/deconstruct.m: runtime/mercury_deconstruct.h: runtime/mercury_deconstruct.c: runtime/mercury_ml_expand_body.h: runtime/mercury_ml_deconstruct_body.h: Add predicates deconstruct.functor_number and deconstruct.deconstruct.du, which returns a functor number suitable for use by construct.construct rather than a functor name. library/construct.m: library/term.m: browser/term_rep.m: extras/quickcheck/qcheck.m: tests/valid/agc_unbound_typevars.m: tests/valid/agc_unbound_typevars2.m: Add a function get_functor_lex, which returns the lexicographic functor number given an ordinal functor number. Add equivalence types to make it clearer which ordering is being used by which functor numbers. Remove a C-ism: num_functors now fails rather than returning -1 for types without functors. NEWS: Document the new predicates and functions. runtime/mercury_type_info.h: runtime/mercury_builtin_types.c: runtime/mercury_mcpp.h: compiler/rtti.m: compiler/rtti_out.m: compiler/type_ctor_info.m: compiler/rtti_to_mlds.m: compiler/opt_debug.m: Add a field to MR_TypeCtorInfo which contains a mapping from an ordinal functor number to a lexicographic functor number which can be passed to construct.construct. Bump MR_RTTI_VERSION. tests/hard_coded/expand.m: tests/hard_coded/expand.exp: tests/hard_coded/expand.exp2: tests/hard_coded/construct_test.m: tests/hard_coded/construct_test.exp: tests/hard_coded/construct_test_exist.m: tests/hard_coded/construct_test_exist.exp: Test cases. |
||
|
Julien Fischer
|
b4c3bb1387 |
Clean up in unused module imports in the Mercury system detected
Estimated hours taken: 3 Branches: main Clean up in unused module imports in the Mercury system detected by --warn-unused-imports. analysis/*.m: browser/*.m: deep_profiler/*.m: compiler/*.m: library/*.m: mdbcomp/*.m: profiler/*.m: slice/*.m: Remove unused module imports. Fix some minor departures from our coding standards. analysis/Mercury.options: browser/Mercury.options: deep_profiler/Mercury.options: compiler/Mercury.options: library/Mercury.options: mdbcomp/Mercury.options: profiler/Mercury.options: slice/Mercury.options: Set --no-warn-unused-imports for those modules that are used as packages or otherwise break --warn-unused-imports, e.g. because they contain predicates with both foreign and Mercury clauses and some of the imports only depend on the latter. |
||
|
Zoltan Somogyi
|
e21193c283 |
Rename a bunch of predicates and function symbols to eliminate
Estimated hours taken: 6 Branches: main browser/*.m: compiler/*.m: Rename a bunch of predicates and function symbols to eliminate ambiguities. The only real change is factoring out some common code in the mlds and llds code generators, replacing them with single definitions in switch_util.m. |
||
|
Zoltan Somogyi
|
863874df85 |
Document my recent change implementing coverage testing.
Estimated hours taken: 6 Branches: main Document my recent change implementing coverage testing. At the same time, eliminate the old hack that allowed a file containing a list of file names to be considered a trace count file. We haven't needed it since the addition of mtc_union, and it can lead to incomprensible error messages. (The presence of the old hack made documenting coverage testing harder.) In the process, fix the tools code for rerunning failed test cases only. doc/user_guide.texi: Document my recent change implementing coverage testing, and the elimination of the old hack. mdbcomp/trace_counts.m: Modify the predicates for reading in trace count files along the lines above. mdbcomp/slice_and_dice.m: Modify the predicates for reading in slices and dices along the lines above. Rename some function symbols to avoid ambiguities. compiler/tupling.m: slice/mcov.m: slice/mtc_diff.m: slice/mtc_union.m: trace/mercury_trace_declarative.c: Conform to the changes above. slice/mcov.m: Fix the usage message, which referred to this program by its old name mct. Allow the output to be restricted to a set of named modules only. This is to make testing easier. slice/mtc_diff.m: Rename the long form of the -o option from --out to --output-file, to make it consistent with the other programs. tests/run_one_test: tools/bootcheck: Modify the algorithm we use to gather trace counts for the Mercury compiler from both passed and failed test cases to run mtc_union periodically instead of gathering all the trace counts file and keeping them to the end (which takes far too much disk space). Fix an old bug: gather trace counts from executions of the Mercury compiler only. tests/debugger/Mmakefile: tests/debugger/dice.passes: Modify the dice test case to compute the union of the trace counts for the passed versions of this test case to use mtc_union to create dice.passes, instead of having dice.passes statically contain the list of the names of the passed trace count files (since that capability is deleted by this diff). tools/bootcheck: tests/Mmake.common: Fix the code for rerunning failed tests only. mdbcomp/prim_data.m: Eliminate some ambiguities in predicate names. compiler/*.m: Conform to the change to prim_data.m. compiler/error_util.m: Add reading files as a phase in error messages. compiler/mercury_compile.m: Use the new facilities in error_util for printing an error message. |
||
|
Zoltan Somogyi
|
00741b0162 |
This diff contains no algorithmic changes.
Estimated hours taken: 6 Branches: main This diff contains no algorithmic changes. It merely renames apart a bunch more function symbols to reduce ambiguity. After this diff, the summary line from the mdb command "ambiguity -f" is Total: 351 names used 975 times, maximum 31, average: 2.78 browser/*.m: compiler/*.m: Rename function symbols to eliminate ambiguities. tests/debugger/declarative/dependency.exp: tests/debugger/declarative/dependency2.exp: Update the expected out where some internal function symbol names appear in the output of the debugger. (This output is meant for implementors only.) |
||
|
Julien Fischer
|
aeeedd2c13 |
Standardize formatting of comments at the beginning of modules.
compiler/*.m: Standardize formatting of comments at the beginning of modules. |
||
|
Zoltan Somogyi
|
4fe703c7b9 |
Implement a more cache-friendly translation of lookup switches.
Estimated hours taken: 8
Branches: main
Implement a more cache-friendly translation of lookup switches. Previously,
for a switch such as the one in
:- pred p(foo::in, string::out, bar::out, float::out) is semidet.
p(d, "four", f1, 4.4).
p(e, "five", f2, 5.5).
p(f, "six", f4("hex"), 6.6).
p(g, "seven", f5(77.7), 7.7).
we generated three static cells, one for each argument, and then indexed
into each one in turn to get the values of HeadVar__2, HeadVar__3 and
HeadVar__4. The different static cells each represent a column here.
Each of the loads accessing the columns will access a different cache block,
so with this technique we expect to get as many cache misses as there are
output variables.
This diff changes the code we generate to use a vector of static cells
where each cell represents a row. The assignments to the output variables
will now access the different fields of a row, which will be next to each
other. We thus expect only one cache miss irrespective of the number of output
variables, at least up to the number of variables that actually fit into one
cache block.
compiler/global_data.m:
Provide a mechanism for creating not just single (scalar) static cells,
but arrays (vectors) of them.
compiler/lookup_switch.m:
Use the new mechanism to generate code along the lines described above.
Put the information passed between the two halves of the lookup switch
implementation (detection and code generation) into an opaque data
structure.
compiler/switch_gen.m:
Conform to the new interface of lookup_switch.m.
compiler/ll_pseudo_type_info.m:
compiler/stack_layout.m:
compiler/string_switch.m:
compiler/unify_gen.m:
compiler/var_locn.m:
Conform to the change to global_data.m.
compiler/llds.m:
Define the data structures for holding vectors of static cells. Rename
the function symbols we used to use to refer to static cells to make
clear that they apply to scalar cells only. Provide similar mechanisms
for representing static cell vectors and references to them.
Generalize heap_ref heap references to allow the index to be computed
at runtime, not compile time. For symmetry's sake, do likewise
for stack references.
compiler/llds_out.m:
Add the code required to write out static cell vectors.
Rename decl_ids to increase clarity and avoid ambiguity.
compiler/code_util.m:
compiler/exprn_aux.m:
Modify code that traverses rvals to now also traverse the new rvals
inside memory references.
compiler/name_mangle.m:
Provide the prefix for static cell vectors.
compiler/layout_out.m:
compiler/rtti_out.m:
compiler/opt_debug.m:
Conform to the change to data_addrs and decl_ids.
compiler/code_info.m:
Provide access to the new functionality in global_data.m, and conform
to the change to llds.m.
Provide a utility predicate needed by lookup_switch.m.
compiler/hlds_llds.m:
Fix the formatting of some comments.
tools/binary:
tools/binary_step:
Fix the bit rot that has set in since they were last used (the rest
of the system has changed quite a lot since then). I had to do so
to debug one part of this change.
tests/hard_coded/dense_lookup_switch2.{m,exp}:
tests/hard_coded/dense_lookup_switch3.{m,exp}:
New test cases to exercise the new algorithm.
tests/hard_coded/Mmakefile:
Enable the new test cases, as well as an old one (from 1997!)
that seems never to have been enabled.
|