* module/language/tree-il/optimize.scm (alpha-rename, peval): Add
support for lexical-set, while avoiding copy propagation and pruning
of assigned variables.
* module/language/tree-il/optimize.scm (<var>, build-var-table, peval):
Before going into peval, build a table indicating refcounts and a set?
flag for all lexicals. Add to the table when introducing new bindings
(via alpha-renaming).
* module/language/tree-il/optimize.scm (peval): Rename
`pure-expression?' to `constant-expression?', in the sense of GCC's
`pure' and `const'. A <toplevel-ref> is not constant, because it can
be mutated. A <dynref> isn't constant either, for the same reason.
* test-suite/tests/tree-il.test ("partial evaluation"): Add a test, and
update existing tests that assumed that toplevel-ref would propagate.
* module/language/tree-il/optimize.scm (peval): Add a "test" context,
which folds statically decidable values to <const>. Fold pure
expressions to <void> in "effect" contexts. Adapt the <conditional>
and <sequence> tests to simply look for <const> or <void> expressions,
respectively.
* module/language/tree-il/optimize.scm (alpha-rename): Rename the
init
expressions of a <lambda-case>.
(peval): Coalesce the <let-values> clauses.
Fix pure-expression? matching of <lambda> clauses.
Loop over and maybe-unconst the inits of a <lambda-case>.
* module/language/tree-il/optimize.scm (peval): Rename `src' to
`lv-src', and `src2' to `src'; pass `make-let-values' the right source
locations. Reindent `let*'.
* module/language/tree-il/primitives.scm (*primitive-constructors*):
Record car, cdr, vector-ref, and struct-ref as "constructors".
Comment to come later.
(*effect-free-primitives*): Update.
* test-suite/tests/tree-il.test ("partial evaluation"): Add tests.
* module/language/tree-il/optimize.scm (peval): Fix comment regarding
alpha-renaming: it's not simply the allocator that needs unique names;
rather, all transformations depend on it.
* module/language/tree-il/optimize.scm (fresh-gensyms): New helper.
(alpha-rename): Name the new gensyms using the old names as templates,
not the old gensyms. This prevents accidental collisions between
gensyms, if #{x 1}# becomes #{x 12}# instead of #{x 2}#.
* module/language/tree-il/optimize.scm (peval): Add support for
let-values. Try to inline the consumer into the body of the producer,
if there is only one return point, and we can figure out how many
values are being returned, and that number is compatible with the
consumer.
* module/language/tree-il/optimize.scm (peval): Add support for fix,
dynwind, dynlet, dynref, module-set, and toplevel-set. (Mutating a
variable directly is similar to calling a function that does so behind
our backs, so this presents no additional problem.)
* module/language/tree-il/optimize.scm (code-contains-calls?): New
procedure.
(peval): Use it and abort inlining if the residual code of a procedure
application contains recursive calls. Suggested by Wingo, Waddell,
and Dybvig. Fixes <http://debbugs.gnu.org/9542>.
* test-suite/tests/tree-il.test ("partial evaluation"): Update 2 tests
that relied on the previous behavior. Add 1 another test.
* module/language/tree-il/optimize.scm (peval)[maybe-unlambda]: New
procedures.
Use it to de-duplicate named lambdas. This fixes the scoping bug
described at <https://lists.gnu.org/archive/html/bug-guile/2011-09/msg00019.html>.
* test-suite/tests/tree-il.test ("partial evaluation"): Add tests to
reproduce the bug.
* module/language/tree-il/optimize.scm (peval): Propagate ARGS to BODY
only when all of ARGS are pure. Change APP to use `maybe-unconst' for
its arguments.
* test-suite/tests/tree-il.test ("partial evaluation"): Add tests for
mutability preservation and non-propagation of non-constant arguments
to lambdas.
* module/language/tree-il/optimize.scm (peval)[make-values]: Distinguish
between 1 or another number of values.
[mutable?, make-value-construction, maybe-unconst]: New procedures.
Use it in <let>, <letrec>, <toplevel-define>, and <lambda-case>.
* test-suite/tests/tree-il.test ("partial evaluation"): Add tests
for mutability preservation.
Thanks to William R. Cook for his excellent tutorial,
<http://softlang.uni-koblenz.de/dsl11/>.
* module/language/tree-il/optimize.scm (optimize!): Call `peval' unless
the #:partial-eval? option asks otherwise.
(peval): New procedure.
* module/language/tree-il/inline.scm: Add comment.
* module/language/tree-il/primitives.scm (*primitive-constructors*): New
variable.
(*effect-free-primitives*): Use it.
(constructor-primitive?): New primitive.
* test-suite/tests/tree-il.test (assert-tree-il->glil): Extend to
support `with-partial-evaluation', `without-partial-evaluation', and
`with-options'.
(peval): New binding.
(pass-if-peval): New macro.
("lexical refs"): Run tests without partial evaluation.
("letrec"): Likewise.
("the or hack"): Likewise.
("conditional"): Likewise, for some tests.
("sequence"): Adjust to new generated code.
("partial evaluation"): New test prefix.
* module/language/tree-il.scm (tree-il->scheme): Fix incorporation of
`lambda' in a `case-lambda'.
* test-suite/tests/tree-il.test ("tree-il->scheme"): Add a test.
* module/language/tree-il/primitives.scm (+, *, cons*): In the case of
just one argument (the identity case), expand to (values x) instead of
just x. Fixes values truncation in that case.
(values): Likewise remove (values x) -> x translation, as the compiler
will do it for us, and this fixes (values (values 1 2)).
* module/language/tree-il/compile-glil.scm (flatten-lambda-case): Handle
`values' in a push context here.
* test-suite/tests/tree-il.test ("values"): Add some tests.
http://savannah.gnu.org/bugs/?33362
* module/language/tree-il/compile-glil.scm (flatten-lambda-case): Rename
from flatten, as it really just takes a particular case. Instead of
iteratively compiling lambda cases through `comp', tail-call through
flatten-lambda-case. This allows code to see which case it's being
compiled in. Take advantage of that to limit the self-tail-call
optimization to self-calls to the same case -- otherwise we might be
jumping to a label without having reserved the right number of
locals.
(flatten-lambda): Adapt the caller.
* test-suite/tests/compiler.test ("case-lambda"): Add a test.
* module/language/glil/compile-assembly.scm: Clean up code for
subprograms (not needed, we just cache the glil) and object alists
(replaced by constants tables).
* module/language/glil/compile-assembly.scm (compile-assembly): Rework
to handle toplevel-specific code generation here, instead of in
glil->assembly. Specifically, here we build a global constant table,
and arrange for it to be the objtable of the toplevel thunk.
(compile-program): New helper, compiles a <glil-program> and returns
just the (load-program ...) form.
(compile-objtable): New helper, generates assembly to build an object
table, using some other constants table, and possibly recursing to
`compile-program' for cached GLIL programs.
(glil->assembly): Simplify, removing the toplevel? argument, and
replacing the object alist with an objtable computed in a previous
pass. Adapt to the new form of the objtable, and to use
compile-program and compile-objtable.
* module/language/glil/compile-assembly.scm (dump-constants): New
helper. Generates bytecode that will result in a vector for the
global object table being pushed on the stack. The items in the
global object table will share state as much as possible.
* module/language/glil/compile-assembly.scm (immediate?): New helper.
(build-constant-store): New helper. Walks the GLIL tree and builds up
a constant table, as a vhash.
(build-object-table): Another helper, builds a constant table for a
given GLIL program.
* module/language/assembly/compile-bytecode.scm (compile-bytecode):
Rewrite to fill a bytevector directly, instead of using bytevector
ports. `write-bytecode' itself is still present and almost the same
as before; it's just that `write-byte' et al now inline the effect of
writing a byte to a binary port.
* test-suite/tests/asm-to-bytecode.test (comp-test): Refactor to use
public interfaces.
* module/language/assembly.scm (byte-length): Silly, minor tweak: put
the fixed-length instruction case first. Seems to shave some 10% off
the time compiling psyntax.scm (when the whole rest of the system is
compiled, of course).
