* module/language/glil/compile-assembly.scm (scheme-list?): New
predicate, like `list?' but requires that the last cdr must be '(),
not #nil.
(dump-object, dump-constants): Use `list' opcode to create a list only
if it is terminated by '(). If it's terminated by #nil, we must use
the more general `cons' opcode.
* 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.
* libguile/vm-i-system.c (assert-nargs-ee/locals): New instruction, a
combination of assert-nargs-ee and reserve-locals in the case in which
nreq and nlocs can both be represented in 8 bits.
* module/language/glil/compile-assembly.scm (glil->assembly): Add
compiler case.
* doc/ref/vm.texi (Function Prologue Instructions): Update docs.
* module/system/base/language.scm (<language>): Remove the `version'
field from languages. It just wasn't useful.
* module/language/assembly/spec.scm:
* module/language/brainfuck/spec.scm:
* module/language/bytecode/spec.scm:
* module/language/ecmascript/spec.scm:
* module/language/elisp/spec.scm:
* module/language/glil/spec.scm:
* module/language/objcode/spec.scm:
* module/language/scheme/spec.scm:
* module/language/tree-il/spec.scm:
* module/language/value/spec.scm: Remove #:version from all language
definitions. Shorten some language names (e.g. "Guile Scheme" ->
"Scheme").
* module/language/glil/compile-assembly.scm (glil->assembly): Allow an
integer for `vars', which means simply to truncate the values, and not
pass binding metadata to the compiler.
* libguile/arrays.h:
* libguile/arrays.c (scm_from_contiguous_array): New public function,
like scm_from_contiguous_typed_array but for arrays of generic Scheme
values.
* libguile/vm-i-scheme.c (make-struct): Sync regs before making the
struct, so if we get a GC the regs are on the heap.
(make-array): New instruction, makes an generic (untyped) Scheme
array.
* module/language/glil/compile-assembly.scm (dump-object): Correctly
compile arrays.
* libguile/vm-i-system.c (tail-call, tail-call/nargs, tail-apply)
(tail-call/cc): Rename these back to tail-* from goto/*. We should
reserve the rename-then-goto name for when you actually do a rename
and goto, not when you shuffle the stack.
* doc/ref/vm.texi:
* module/language/glil/decompile-assembly.scm:
* module/language/tree-il/compile-glil.scm:
* test-suite/tests/tree-il.test: Adapt all callers and documentation.
* module/language/elisp/compile-tree-il.scm: Update for changes to
tree-il (lambda-case, mainly).
* module/language/elisp/spec.scm: Update GPL version to 3. Update reader
for new taking a port and environment argument.
* libguile/_scm.h: Bump objcode version.
* libguile/vm-i-system.c: Fix conflicts.
* module/Makefile.am: Fix conflicts, and add elisp modules to the build.
* module/language/glil/compile-assembly.scm (open-arity, close-arity)
(begin-arity, glil->assembly): Refactor so that arities can have
noncontiguous starts and ends. So within a prelude there is no arity
-- only before (the previous arity) or after (the new arity).
* module/system/vm/program.scm (arity:end): Add this private accessor.
Arities are expected to be in the new format. While not a change in
objcode format, it is an incompatible change, so I'll bump the objcode
cookie.
(program-arity): Check that the ip is within both bounds of the arity.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bump.
* libguile/programs.c (scm_i_program_arity): Update for new arity format.
* module/system/vm/frame.scm (vm-frame-arguments): Avoid throwing an
error in this function, which is called from the backtrace code.
* module/language/tree-il.scm (<lambda-case>): Add "inits" field, so we
don't have to parse it out of opt and kw. Adapt the traversal
procedures.
* module/language/tree-il/analyze.scm (analyze-lexicals): Analyze
lexicals in the <lambda-case> init expressions as well. Fix keyword
allocation.
* module/language/tree-il/compile-glil.scm (compile-glil): Adapt to
make-lambda-case change.
(flatten): Adapt to "inits" slot, actually init uninitialized args,
and fix bugs related to keyword arguments.
* module/language/tree-il/inline.scm (inline!): Adapt a little bit --
but with no effect.
* module/language/glil/compile-assembly.scm (glil->assembly): Flesh out
<glil-kw-prelude> compilation some more. Add a "bound?" op for
<glil-lexical>, which will push #t if the local is bound.
* module/ice-9/psyntax.scm (build-simple-lambda, build-lambda-case):
Update for new signature of make-lambda-case.
* module/ice-9/psyntax-pp.scm: Regenerated.
* module/language/brainfuck/compile-tree-il.scm (compile-body):
* module/language/ecmascript/compile-tree-il.scm (comp):
* test-suite/tests/tree-il.test ("lambda"): Update for new lambda-case
syntax.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bumparoo
* libguile/vm-i-system.c (push-rest, bind-rest): Logically there are
actually two rest binders -- one that pops, conses, and pushes, and
one that pops, conses, and local-sets. The latter is used on keyword
arguments, because the keyword arguments themselves have been shuffled
up on the stack. Renumber ops again.
* module/language/tree-il/compile-glil.scm (flatten): Attempt to handle
compilation of lambda-case with keyword arguments. Might need some
help.
* module/ice-9/psyntax.scm (build-lambda-case): An attempt to handle the
interpreted case correctly. This might need a couple iterations, but
at least it looks like the compile-glil code.
* module/ice-9/psyntax-pp.scm: Regenerated.
* module/language/glil.scm (<glil>): Rename "rest?" to "rest" in
<glil-opt-prelude> and <glil-kw-prelude>, as it is no longer a simple
boolean, but if true is an integer: the index of the local variable to
which the rest should be bound.
* module/language/glil/compile-assembly.scm (glil->assembly): Adapt to
"rest" vs "rest?". In the keyword case, use "bind-rest" instead of
"push-rest".
* test-suite/tests/tree-il.test: Update for opt-prelude change.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bump.
* libguile/vm-i-system.c (br-if-nargs-ne, br-if-args-lt)
(br-if-nargs-gt): New instructions, for use by different lambda cases.
(bind-optionals, bind-optionals/shuffle, bind-kwargs): New
instructions, for binding optional and keyword arguments. Renumber
other ops.
* module/language/ecmascript/compile-tree-il.scm (comp, comp-body):
Update for new tree-il. Use the new optional argument mechanism
instead of emulating it with rest arguments.
* module/language/glil/compile-assembly.scm (glil->assembly): Tweaks for
optional and keyword argument compilation.
* module/language/tree-il.scm (parse-tree-il, unparse-tree-il): Make the
else case optional, in the s-expression serialization of tree-il.
* module/language/tree-il/compile-glil.scm (flatten): Handle all of the
lambda-case capabilities.
* libguile/vm-i-system.c (bind-rest): Renamed from push-rest-list.
(reserve-locals): Change so that instead of reserving space for some
additional number of locals, reserve-locals takes the absolute number
of locals, including the arguments.
* module/language/glil.scm (<glil-std-prelude>, <glil-opt-prelude>)
(<glil-kw-prelude>): New GLIL constructs, to replace <glil-arity>.
* module/language/glil/compile-assembly.scm (glil->assembly): Compile
the new preludes. Some instructions are not yet implemented, though.
* module/language/tree-il/analyze.scm (analyze-lexicals): The nlocs for
a lambda will now be the total number of locals, including arguments.
* module/language/tree-il/compile-glil.scm (flatten-lambda): Update to
write the new prelude.
* module/system/vm/program.scm (program-bindings-for-ip): If a given
index doesn't have a binding at the ip given, don't cons it on the
resulting list.
* test-suite/tests/tree-il.test: Update for GLIL changes.
* libguile/objcodes.h (struct scm_objcode): Remove nargs, nrest, and
nlocs, as they are no longer needed. Also obviates the need for a
padding word.
* libguile/procs.c (scm_thunk_p): Use scm_i_program_arity for programs.
* libguile/procprop.c (scm_i_procedure_arity): Use scm_i_program_arity
for programs.
(scm_procedure_properties, scm_set_procedure_properties_x)
(scm_procedure_property, scm_set_procedure_property_x): Rework so that
non-closure properties are stored directly in a weak hash, instead of
needing a weak hash of "stand-in" closures to hold the properties. Fix
docstrings also.
* libguile/root.h (scm_stand_in_procs): Remove from the scm_sys_protects
set. Actually with libGC, we should be able to store the elements of
scm_sys_protects directly as global variables.
* libguile/gc.c (scm_init_storage): Remove scm_stand_in_procs
initialization.
* libguile/programs.c (scm_i_program_arity): New private accessor, tries
to determine the "minimum arity" of a program.
* libguile/vm.c (really_make_boot_program): Adapt to changes in
struct scm_objcode.
* module/language/assembly.scm (*program-header-len*, byte-length):
* module/language/assembly/compile-bytecode.scm (write-bytecode):
* module/language/assembly/decompile-bytecode.scm (decode-load-program):
* module/language/assembly/disassemble.scm (disassemble-load-program):
Adapt to changes in objcode.
* module/system/xref.scm (program-callee-rev-vars): Adapt to changes in
assembly.
* module/language/glil.scm: Remove nargs, nrest, and nlocs from
glil-program.
* module/language/glil/compile-assembly.scm (make-meta, glil->assembly):
* module/language/glil/decompile-assembly.scm (decompile-toplevel):
(decompile-load-program): Adapt to changes in GLIL and assembly.
* module/language/tree-il/compile-glil.scm (flatten-lambda): Adapt to
changes in GLIL.
* test-suite/tests/asm-to-bytecode.test: Adapt to assembly and bytecode
changes.
* test-suite/tests/tree-il.test: Adapt to GLIL changes.
* libguile/objcodes.h: Bump for metadata format change.
* libguile/frames.h: Rework so we don't frob the program's nargs, nlocs,
etc at runtime. Instead we don't really know what's a local var, an
argument, or an intermediate value. It's a little unfortunate, but
this will allow for case-lambda, and eventually for good polymorphic
generic dispatch; and the nlocs etc can be heuristically
reconstructed. Such a reconstruction would be better done at the
Scheme level, though.
(SCM_FRAME_STACK_ADDRESS): New macro, the pointer to the base of the
stack elements (not counting the program).
(SCM_FRAME_UPPER_ADDRESS): Repurpose to be the address of the last
element in the bookkeeping part of the stack -- i.e. to point to the
return address.
* libguile/vm-engine.h:
* libguile/vm-i-system.c: Adapt to removal of stack_base. Though we
still detect stack-smashing underflow, we don't do so as precisely as
we did before, because now we only detect overwriting of the frame
metadata.
* libguile/vm-engine.c (vm_engine): Remove the stack_base variable. It
is unnecessary, and difficult to keep track of in the face of
case-lambda. Also fix miscommented "ra" and "mvra" pushes. Push the
vp->ip as the first ra...
* libguile/vm-i-system.c (halt): ...because here we can restore the
vp->ip instead of setting ip to 0. Allows us to introspect ips all
down the stack, including in recursive VM invocations.
* libguile/frames.h:
* libguile/frames.c (scm_vm_frame_stack): Removed, because it's getting
more difficult to tell what's an argument and what's a temporary stack
element.
(scm_vm_frame_num_locals): New accessor.
(scm_vm_frame_instruction_pointer): New accessor.
(scm_vm_frame_arguments): Defer to an implementation in Scheme.
(scm_vm_frame_num_locals scm_vm_frame_local_ref)
(scm_vm_frame_local_set_x): Since we can get not-yet-active frames on
the stack now, with our current calling convention, we have to add a
heuristic here to jump over those frames -- because frames have
pointers in them, not Scheme values.
* libguile/programs.h:
* libguile/programs.c (scm_program_arity): Remove, in favor of..
(scm_program_arities): ...this, which a list of arities, in a new
format, occupying a slot in the metadata.
* module/language/assembly/decompile-bytecode.scm (decode-load-program):
Fix mv-call decompilation.
* module/system/vm/frame.scm (vm-frame-bindings, vm-frame-binding-ref)
(vm-frame-binding-set!): New functions, to access bindings by name in
a frame.
(vm-frame-arguments): Function now implemented in Scheme. Commented
fairly extensively.
* module/system/vm/program.scm (program-bindings-by-index)
(program-bindings-for-ip): New accessors, parsing the program bindings
metadata into something more useful.
(program-arities, program-arguments): In a case-lambda world, we have
to assume that programs can have multiple arities. But it's tough to
detect this algorithmically; instead we're going to require that the
program metadata include information about the arities, and the parts
of the program that that metadata applies to.
(program-lambda-list): New accessor.
(write-program): Show multiple arities.
* module/language/glil/compile-assembly.scm (glil->assembly): Add
"arities" to the state of the compiler, and add arities entries as
appropriate.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bump.
* libguile/vm-i-system.c (assert-nargs-ee, assert-nargs-ge)
(push-rest-list): New instructions, which for now don't actually do
anything. Renumber the rest of the ops in this file.
* module/language/glil.scm (<glil-arity>): New GLIL type, an entity that
checks the number of args for a block, optionally consing a rest list,
and either branching or erroring if the arity doesn't match.
* module/language/glil/compile-assembly.scm (glil->assembly): Compile
<glil-arity> to assembly. Some of these VM ops are not implemented --
notably the branching case.
* module/language/tree-il/compile-glil.scm (flatten-lambda): Emit
<glil-arity>.
* test-suite/tests/tree-il.test: Update.
* module/language/assembly/spec.scm:
* module/language/brainfuck/spec.scm:
* module/language/bytecode/spec.scm:
* module/language/ecmascript/spec.scm:
* module/language/glil/spec.scm:
* module/language/scheme/spec.scm:
* module/language/tree-il/spec.scm: Language-readers now take two
arguments: the port and the environment. This should allow for
compile-environment-specific reader behavior.
* module/system/base/compile.scm (read-and-compile):
* module/system/repl/common.scm (repl-read): Pass the environment to the
language-reader.
* module/system/repl/repl.scm (meta-reader, prompting-meta-read):
* module/system/repl/command.scm (define-meta-command): Use the second
argument to repl-reader, so we avoid frobbing current-reader.
The intent is to maintain the readability of `pmatch' invocations.
* module/language/assembly/disassemble.scm (disassemble-load-program):
Don't use wildcards in `pmatch' invocations, even when the matched
elements are unused.
* module/language/glil/decompile-assembly.scm (decompile-toplevel,
decompile-load-program): Likewise.
* module/system/xref.scm (program-callee-rev-vars): Likewise.
* module/language/assembly.scm (byte-length): Likewise.
* module/language/tree-il/compile-glil.scm (flatten): Likewise.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bump.
* libguile/vm-engine.c (vm_error_bad_wide_string_length): New error
case.
* libguile/vm-i-loader.c (load-unsigned-integer, load-integer)
(load-keyword): Remove these instructions. The former two are
obsoleted by make-int64/make-uint64, the latter via make-keyword.
(load-string): Only handle narrow strings.
(load-symbol): Only handle narrow symbols. The wide case is handled
via make-symbol.
(load-wide-string): New instruction, for wide strings.
* libguile/vm-i-system.c (define): Move here from loaders.c, as now it
just takes a sym on the stack.
(make-keyword, make-symbol): New instructions.
* module/language/assembly.scm: Remove removed instructions. No more
width byte in load-string etc.
* module/language/assembly/compile-bytecode.scm (write-bytecode): Adapt
to change in instruction set.
* module/language/glil/compile-assembly.scm (glil->assembly): Compile
define by pushing the sym then emitting (define).
(dump-object): Dump narrow and wide strings differently. Use
make-keyword and make-symbol as appropriate.
* module/language/tree-il/compile-glil.scm (flatten): When compiling a
ref to a primitive (not a call), first see if the primitive is
actually bound in the root module. (That's not the case with e.g.
bytevector-u8-ref).
* module/system/xref.scm (program-callee-rev-vars): Don't parse out
"nexts".
* test-suite/tests/asm-to-bytecode.test ("compiler"): Adapt to bytecode
format change.
* libguile/vm-i-system.c (fix-closure): New instruction, for wiring
together fixpoint procedures.
* module/Makefile.am (TREE_IL_LANG_SOURCES): Add fix-letrec.scm.
* module/language/glil/compile-assembly.scm (glil->assembly): Reindent
the <glil-lexical> case, and handle 'fix for locally-bound vars.
* module/language/tree-il.scm (<fix>): Add the <fix> tree-il type and
accessors, for fixed-point bindings. This IL construct is taken from
the Waddell paper.
(parse-tree-il, unparse-tree-il, tree-il->scheme, tree-il-fold)
(pre-order!, post-order!): Update for <fix>.
* module/language/tree-il/analyze.scm (analyze-lexicals): Update for
<fix>. The difference here is that the bindings may not be assigned,
and are not marked as such. They are not boxed.
(report-unused-variables): Update for <fix>.
* module/language/tree-il/compile-glil.scm (flatten): Compile <fix> to
GLIL.
* module/language/tree-il/fix-letrec.scm: A stub implementation of
fixing letrec -- will flesh out in a separate commit.
* module/language/tree-il/inline.scm: Fix license, it was mistakenly
added with LGPL v2.1+.
* module/language/tree-il/optimize.scm (optimize!): Run the fix-letrec!
pass.
* libguile/objcodes.c (OBJCODE_COOKIE): Bump again, as our jump offsets
are now multiplied by 8.
* libguile/vm-i-system.c (BR): Interpret the 16-bit offset as a relative
jump to the nearest 8-byte-aligned block -- increasing relative jump
range from +/-32K to +/-240K.
(mvra): Do the same for the mvra jump.
* libguile/vm.c (really_make_boot_program): Align the mvra.
* module/language/assembly.scm (align-block): New export, for aligning
blocks.
* module/language/assembly/compile-bytecode.scm (write-bytecode): Emit
jumps to the nearest 8-byte-aligned block. Effectively our range is 18
bits in either direction. I would like to do this differently -- have
long-br and long-br-if, and all the other br instructions go to 8 bits
only. But the assembler doesn't have an appropriate representation to
allow me to do this yet, so for now this is what we have.
* module/language/assembly/decompile-bytecode.scm (decode-load-program):
Decode the 19-bit jumps.
* libguile/objcodes.c (OBJCODE_COOKIE): Bump objcode cookie, as we added
to struct scm_objcode.
* libguile/objcodes.h (struct scm_objcode): Add a uint32 after metalen
and before base, so that if the structure has 8-byte alignment, base
will have 8-byte alignment too. (Before, base was 12 bytes from the
start of the structure, now it's 16 bytes.)
* libguile/vm-engine.h (ASSERT_ALIGNED_PROCEDURE): Add a check that can
be turned on with VM_ENABLE_PARANOID_ASSERTIONS.
(CACHE_PROGRAM): Call ASSERT_ALIGNED_PROCEDURE.
* libguile/vm-i-system.c (long-local-ref): Add a missing semicolon.
* libguile/vm.c (really_make_boot_program): Rework to operate directly
on a malloc'd buffer, so that the program will be 8-byte aligned.
* module/language/assembly.scm (*program-header-len*): Add another 4 for
the padding.
(object->assembly): Fix case in which we would return (make-int8 0)
instead of (make-int8:0). This would throw off compile-assembly.scm's
use of addr+.
* module/language/assembly/compile-bytecode.scm (write-bytecode): Write
out the padding int.
* module/language/assembly/decompile-bytecode.scm (decode-load-program):
And pop off the padding int too.
* module/language/glil/compile-assembly.scm (glil->assembly): Don't pack
the assembly, assume that assembly.scm has done it for us. If a
program has a meta, pad out the program so that meta will be aligned.
* test-suite/tests/asm-to-bytecode.test: Adapt to expect programs to
have the extra 4-byte padding int.
* module/language/glil/compile-assembly.scm (glil->assembly)
(dump-object): Fix an exciting bug! Subprograms of subprograms were
not being aligned correctly, because the code was generated too early.
So instead delay dumping the object table until the proper time.
* libguile/vm-i-system.c (long-local-ref, long-local-set)
(make-variable): New intructions, for handling nlocs >= 256.
* module/language/glil/compile-assembly.scm (glil->assembly): Compile
<glil-lexical> with support for nlocs >= 256.
* libguile/programs.h:
* libguile/programs.c: (SCM_PROGRAM_FREE_VARIABLES): Rename from
SCM_PROGRAM_FREE_VARS. Callers changed.
* libguile/programs.c (scm_make_program): Rename arg to
"free_variables".
(scm_program_free_variables): Rename from program-free-vars.
* libguile/vm-engine.h:
* libguile/vm-engine.c (VM_CHECK_FREE_VARIABLES): Rename from
VM_CHECK_CLOSURE.
(vm_engine, CACHE_PROGRAM): Rename closure and closure_count to free_vars and
free_vars_vount.
* libguile/vm-i-system.c (FREE_VARIABLE_REF): Rename from CLOSURE_REF.
(free-ref, free-boxed-ref, free-boxed-set): Rename from closure-ref,
closure-boxed-ref, closure-boxed-set.
(make-closure): Renamed from make-closure2.
* module/language/glil/compile-assembly.scm (glil->assembly): Hack to
never write out the the old "make-closure" instruction. Will fix
better later. Change to emit free-ref etc instead of closure-ref.
* module/language/tree-il/compile-glil.scm (flatten): Emit make-closure
instead of make-closure2, now that the old make-closure is gone.
* module/system/vm/program.scm (system): Rename program-free-vars to
program-free-variables.
* test-suite/tests/tree-il.test ("lambda"): Update for make-closure.
* module/language/glil.scm (<glil>): New GLIL type, <glil-lexical>,
which will subsume other lexical types.
* module/language/glil/compile-assembly.scm: Compile <glil-lexical>.
(make-open-binding): Change the interpretation of the second argument
-- instead of indicating an "external" var, it now indicates a boxed
var.
(open-binding): Adapt to new glil-bind format.
* module/language/tree-il/analyze.scm: Add a lot more docs.
(analyze-lexicals): Change the allocation algorithm and output format
to allow the tree-il->glil compiler to capture free variables
appropriately and to reference bound variables in boxes if necessary.
Amply documented.
* module/language/tree-il/compile-glil.scm (compile-glil): Compile
lexical variable access to <glil-lexical>. Emit variable capture and
closure creation code here, instead of leaving that task to the
GLIL->assembly compiler.
* test-suite/tests/tree-il.test: Update expected code emission.
* doc/ref/vm.texi: Document new instruction.
* libguile/vm-i-system.c: Add it to the VM.
* module/language/assembly.scm: Compile (const %nil) to (make-nil) assembly.
* module/language/glil/decompile-assembly.scm: Handle (make-nil)
* module/language/elisp/compile-tree-il.scm: Use (const %nil) for nil.
* doc/ref/vm.texi (Loading Instructions): Remove references to
load-integer and load-unsigned-integer -- they're still in the VM but
will be removed at some point.
(Data Control Instructions): Add make-int64 and make-uint64.
* libguile/vm-i-loader.c (load-unsigned-integer): Allow 8-byte values.
But this instruction is on its way out, yo.
* libguile/vm-i-system.c (make-int64, make-uint64): New instructions.
* module/language/assembly.scm (object->assembly): Write out make-int64
and make-uint64 instructions, using bytevectors to do the endianness
conversion.
(assembly->object): And pretty-print them back, for disassembly.
* module/language/glil/compile-assembly.scm: Don't generate load-integer
/ load-unsigned-integer instructions.
