* libguile/vm-engine.c (lsh, rsh, lsh/immediate, rsh/immediate): New
instructions taking unboxed bit counts.
* module/language/cps/compile-bytecode.scm (compile-function):
* module/language/cps/effects-analysis.scm:
* module/language/cps/specialize-numbers.scm (specialize-f64-unop):
(specialize-u64-unop): Add ability to specialize add/immediate, etc,
and add lsh/immediate as well.
(specialize-u64-binop, specialize-u64-shift): Move rsh/lsh
specialization to its own procedure, given that the bit count is
already unboxed.
(specialize-operations): Adapt to support more /immediate
instructions.
* module/language/cps/type-fold.scm (mul): Reify an lsh/immediate
instead of an ash.
* module/language/cps/types.scm (compute-ash-range): Add type inferrers
for lsh, rsh, and their immediate variants.
* module/system/vm/assembler.scm: Export emit-lsh and so on.
* module/language/tree-il/compile-cps.scm (convert): Convert "ash" on
immediates to rsh/immediate or lsh/immediate.
* module/language/cps.scm ($primcall): Add "param" member, which will be
a constant parameter to the primcall. The idea is that constants used
by primcalls as immediates don't need to participate in optimizations
in any way -- they should not participate in CSE, have the same
lifetime as the primcall so not part of DCE either, and don't need
slot allocation. Indirecting them through a named $const binding is
complication for no benefit. This change should eventually improve
compilation time and memory usage, once we fully take advantage of it,
as the number of labels and variables will go down.
* module/language/cps/closure-conversion.scm:
* module/language/cps/compile-bytecode.scm:
* module/language/cps/constructors.scm:
* module/language/cps/contification.scm:
* module/language/cps/cse.scm:
* module/language/cps/dce.scm:
* module/language/cps/effects-analysis.scm:
* module/language/cps/elide-values.scm:
* module/language/cps/handle-interrupts.scm:
* module/language/cps/licm.scm:
* module/language/cps/peel-loops.scm:
* module/language/cps/prune-bailouts.scm:
* module/language/cps/prune-top-level-scopes.scm:
* module/language/cps/reify-primitives.scm:
* module/language/cps/renumber.scm:
* module/language/cps/rotate-loops.scm:
* module/language/cps/self-references.scm:
* module/language/cps/simplify.scm:
* module/language/cps/slot-allocation.scm:
* module/language/cps/specialize-numbers.scm:
* module/language/cps/specialize-primcalls.scm:
* module/language/cps/split-rec.scm:
* module/language/cps/type-checks.scm:
* module/language/cps/type-fold.scm:
* module/language/cps/types.scm:
* module/language/cps/utils.scm:
* module/language/cps/verify.scm:
* module/language/tree-il/compile-cps.scm: Adapt all users.
* module/language/cps/compile-bytecode.scm (compile-function): Rename
the binary* helper back to binary, update uses, and remove logtest
branch as we no longer put logtest in test context.
* module/language/cps/primitives.scm (*comparisons*): Remove logtest.
* module/language/cps/type-fold.scm: Remove logtest folder.
(logbit?): Fold to logand.
* module/language/cps/types.scm (logtest): Update to be a type inferrer
and not a predicate inferrer.
* module/language/tree-il/peval.scm (peval): Transform logtest and
logbit? to (zero? (logand _ _)).
* module/language/tree-il/compile-cps.scm (canonicalize, convert):
Simplify handling of branching primcalls so that `convert' only ever
sees branching primcalls in a test context.
* module/language/tree-il/compile-cps.scm (convert): Fold test of
constants directly to their branches without reifying booleans, to
simplify "and" chains early.
* module/language/cps/compile-bytecode.scm (compile-function): Add
support for heap-object? in test context.
* module/language/cps/primitives.scm (*immediate-predicates*):
(*heap-type-predicates*, *comparisons*): New sets of predicates for
which the VM has branching operations.
(heap-type-predicate?): New predicate.
(*branching-primcall-arities*): Make a hash table.
(branching-primitive?, prim-arity): Adapt
to *branching-primcall-arities* being a hash table.
* module/language/cps/type-fold.scm (heap-object?): Add folder.
* module/language/tree-il/compile-cps.scm (convert): Precede heap type
checks with a heap-object? guard.
* doc/ref/vm.texi (Inlined Scheme Instructions): Add string-set!.
* libguile/vm-engine.c (string-set!): New opcode.
* module/ice-9/rdelim.scm (read-string): Reimplement in terms of a
geometrically growing list of strings, to reduce total heap usage when
reading big files.
* module/language/cps/compile-bytecode.scm (compile-function): Add
string-set! support.
* module/language/cps/types.scm (string-set!): Update for &u64 index.
* module/language/tree-il/compile-cps.scm (convert): Unbox index to
string-set!.
* module/system/vm/assembler.scm (system): Export string-set!.
Fixes <http://bugs.gnu.org/20272>.
* module/ice-9/boot-9.scm (module-generate-unique-id!)
(module-gensym): New procedures.
(module): Add 'next-unique-id' field.
(the-root-module): Inherit 'next-unique-id' value from early stub.
(make-module, make-autoload-interface): Adjust calls to
module-constructor.
* module/ice-9/psyntax.scm (gen-label, new-mark): Generate unique
identifiers from the module name and the per-module unique-id.
(build-lexical-var, generate-temporaries): Use
'module-gensym' instead of 'gensym'.
* module/ice-9/psyntax-pp.scm: Regenerate.
* module/language/tree-il/fix-letrec.scm (fix-letrec!): Use
'module-gensym' instead of 'gensym'.
* module/system/base/syntax.scm (define-record): Likewise.
(transform-record): Likewise.
Co-authored-by: Ludovic Courtès <ludo@gnu.org>
* doc/ref/vm.texi (Inlined Atomic Instructions): New section.
* libguile/vm-engine.c (VM_VALIDATE_ATOMIC_BOX, make-atomic-box)
(atomic-box-ref, atomic-box-set!, atomic-box-swap!)
(atomic-box-compare-and-swap!): New instructions.
* libguile/vm.c: Include atomic and atomics-internal.h.
(vm_error_not_a_atomic_box): New function.
* module/ice-9/atomic.scm: Register primitives with the compiler.
* module/language/cps/compile-bytecode.scm (compile-function): Add
support for atomic ops.
* module/language/cps/effects-analysis.scm: Add comment about why no
effects analysis needed.
* module/language/cps/reify-primitives.scm (primitive-module): Add case
for (ice-9 atomic).
* module/language/tree-il/primitives.scm (*effect-free-primitives*):
(*effect+exception-free-primitives*): Add atomic-box?.
* module/system/vm/assembler.scm: Add new instructions.
* test-suite/tests/atomic.test: Test with compilation and
interpretation.
* module/language/tree-il/primitives.scm
(character-comparison-expander): Expand out char<? and friends to <,
unboxing the char arguments.
* module/language/cps/types.scm:
* module/language/cps/effects-analysis.scm: Remove mention of char<? and
friends as we won't see them any more. Also fixes#24318.
* doc/ref/vm.texi (Top-Level Environment Instructions): Update
documentation.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bump, sadly.
* module/system/vm/assembler.scm (*bytecode-minor-version*): Bump.
* libguile/vm-engine.c (define!): Change to store variable in dst slot.
* module/language/tree-il/compile-cps.scm (convert):
* module/language/cps/compile-bytecode.scm (compile-function): Adapt to
define! change.
* module/language/cps/effects-analysis.scm (current-module): Fix define!
effects. Incidentally here was the bug: in Guile 2.2 you can't have
effects on different object kinds in one instruction, without
reverting to &unknown-memory-kinds.
* test-suite/tests/compiler.test ("regression tests"): Add a test.
* module/language/tree-il/peval.scm (peval): Don't
inline (call-with-values foo (lambda (x) exp)) to (let ((x (foo)))
exp). The idea is that call-with-values sets up an explicit context
in which we are requesting an explicit return arity, and that dropping
extra values when there's not a rest argument is the wrong thing.
Fixes#13966.
* test-suite/tests/peval.test ("partial evaluation"): Update test.
* module/language/tree-il/peval.scm (peval): In test context,
fold (let ((x EXP)) (if x x ALT)) to (if EXP #t ALT). This reduces
the number of boolean literals that the compiler has to reify, by
causing EXP to evaluate in test context instead of value context.
Also, rotate `let' out of the test part of conditionals, for the same
reason.
* libguile/vm-engine.c (logsub): New op.
* module/language/cps/effects-analysis.scm (logsub):
* module/language/cps/types.scm (logsub):
* module/system/vm/assembler.scm (system): Add support for the new op.
* module/language/tree-il/compile-cps.scm (canonicalize):
Rewrite (logand x (lognot y)) to (logsub x y).
* libguile/vm-engine.c (bv-s8-ref, bv-s16-ref, bv-s32-ref, bv-s64-ref):
Unbox index and return unboxed S32 value.
(bv-s8-set!, bv-s16-set!, bv-s32-set!, bv-s64-set!): Unbox index and
take unboxed S32 value.
(bv-u8-ref, bv-u16-ref, bv-u32-ref, bv-u64-ref)
(bv-s8-set!, bv-s16-set!, bv-s32-set!, bv-s64-set!): Likewise, but
with unsigned values.
(bv-f32-ref, bv-f32-set!, bv-f64-ref, bv-f64-set!): Use memcpy to
access the value so we don't have to think about alignment. GCC will
inline this to a single instruction on architectures that support
unaligned access.
* libguile/vm.c (vm_error_out_of_range_uint64)
(vm_error_out_of_range_int64): New helpers.
* module/language/cps/slot-allocation.scm (compute-var-representations):
All bytevector ref operations produce untagged values.
* module/language/cps/types.scm (define-bytevector-accessors): Update
for bytevector untagged indices and values.
* module/language/cps/utils.scm (compute-constant-values): Fix s64
case.
* module/language/tree-il/compile-cps.scm (convert): Box results of all
bytevector accesses, and unbox incoming indices and values.
* libguile/vm-engine.c: Remove add1 and sub1 instructions. Will replace
with add/immediate and sub/immediate.
* module/language/tree-il/peval.scm (peval): If we reify a new
<primcall>, expand it. Removes 1- and similar primcalls.
* module/language/tree-il/primitives.scm: Don't specialize (+ x 1) to 1+.
(expand-primcall): New export, does a single primcall expansion.
(expand-primitives): Use the new helper.
* module/language/cps/effects-analysis.scm:
* module/language/cps/primitives.scm:
* module/language/cps/types.scm:
* module/system/vm/assembler.scm: Remove support for add1 and sub1 CPS
primitives.
* test-suite/tests/peval.test ("partial evaluation"): Adapt tests that
expect 1+/1- to expect +/-.
* module/language/tree-il/compile-cps.scm (convert): bv-f32-ref,
bv-f32-set!, bv-f64-ref, and bv-f64-set! take the index as an untagged
u64 value.
* module/language/cps/types.scm (define-bytevector-uaccessors): New
helper, used while migrating bytevectors to take unboxed indexes.
Adapt f32/f64 accessors to use this definition helper.
* libguile/vm-engine.c (BV_FLOAT_REF, BV_FLOAT_SET): The index is
unboxed.
* module/language/cps/compile-bytecode.scm (compile-function):
* module/language/cps/primitives.scm (*branching-primcall-arities*):
* module/language/cps/type-fold.scm (equal?):
* module/language/cps/types.scm (equal?):
* module/language/tree-il/compile-cps.scm (convert): `equal?' is no
longer a branching primcall, because it isn't inline. The
implementation could lead to bad backtraces also, as it didn't save
the IP, and actually could lead to segfaults as it didn't reload the
SP after the return. There is an eqv? fast-path, though.
* module/system/vm/assembler.scm (br-if-equal): Remove interface.
* module/system/vm/disassembler.scm (code-annotation):
(compute-labels): No need to handle br-if-equal.
* module/language/tree-il/compile-cps.scm (convert): Box results of
bv-f32-ref and bv-f64-ref. Unbox the argument to bv-f32-set! and
bv-f64-set!.
* libguile/vm-engine.c (bv-f32-ref, bv-f64-ref): Results are raw.
(bv-f32-set!, bv-f64-set!): Take unboxed arguments.
* module/system/vm/assembler.scm (emit-scm->f64, emit-f64->scm):
Export.
* module/language/cps/compile-bytecode.scm (compile-function):
* module/language/cps/effects-analysis.scm: Add support for scm->f64 and
f64->scm.
* module/language/cps/slot-allocation.scm (compute-var-representations):
Add cases for primops returning raw values.
* module/language/cps/types.scm (bv-f32-ref, bv-f32-set!)
(bv-f64-ref, bv-f64-set!): Deal in &f64 values instead of reals.
* module/scripts/compile.scm (%options): Resurrect -O option and make it
follow GCC, more or less. The default is equivalent to -O2.
* module/language/cps/compile-bytecode.scm (lower-cps):
* module/language/cps/optimize.scm (optimize-higher-order-cps): Move
split-rec to run unconditionally for now, as closure conversion fails
without it.
(define-optimizer): Only verify the result if we are debugging, to
save time.
(cps-default-optimization-options): New exported procedure.
* module/language/tree-il/optimize.scm
(tree-il-default-optimization-options): New exported procedure.
Fixes <http://bugs.gnu.org/21614>.
Reported by tantalum <sph@posteo.eu>.
* module/language/tree-il/compile-cps.scm (convert): Add missing 'cps'
argument to the continuation passed to 'convert-arg'.
* module/language/cps/intmap.scm: Intmaps can now contain any value;
#f does not indicate the absence of a value. Instead we use a unique
private sentinel to mark absent values or branches.
(*absent*, absent?, present?): New helpers.
(new-branch): Initialize empty elements to *absent*.
(clone-branch-with-edit): New helper.
(clone-branch-and-set): Use clone-branch-with-edit.
(writable-branch): Use clone-branch-with-edit
(empty-intmap): Initialize value to *absent*.
(add-level): clone-branch-and-set doesn't take #f as a branch any
more; use new-branch.
(branch-empty?, make-intmap/prune, intmap-add!):
(intmap-add, intmap-remove, intmap-next, intmap-prev):
(intmap-fold, intmap-union, intmap-intersect): Use absent? to detect
absent branches / values.
(intmap-ref): Likewise. Instead of returning #f if the value is not
found, call the optional not-found procedure. By default this will
signal an error.
* module/language/cps/types.scm:
* module/language/cps2/renumber.scm:
* module/language/cps2/simplify.scm: Adapt to intmap-ref signalling an
error by default if the value is not found.
* module/language/tree-il/compile-cps2.scm: Adapt to intmap-add
signalling an error if #f was in the intmap as a value.
* module/language/tree-il/peval.scm (<operand>): Rename "alias-value"
field to "alias", which is now an operand and not an expression.
This allows the operand to capture its environment; before, the
alias was being visited in its use environment instead of its
definition environment.
(peval): Adapt to operand change. Fix construction of rest bindings
as well.
* test-suite/tests/peval.test ("partial evaluation"): New test.
* module/language/tree-il/compile-cps2.scm (convert): Fix bug
in (apply - ...), because the instruction for "-" is "sub", and
"sub" lookup was failing. Caught by numbers.test. Really I would
like to get rid of $prim, somehow.
This is an intermediate step. We'll replace CPS bit by bit. If it
turns out to be a terrible idea we can just revert.
* module/Makefile.am (TREE_IL_LANG_SOURCES): Remove compile-cps.scm.
(CPS_LANG_SOURCES): Remove arities.scm.
* module/language/cps/arities.scm: Remove.
* module/language/tree-il/compile-cps.scm: Remove.
* module/language/tree-il/spec.scm: Remove use of compile-cps.scm.
* module/language/cps/compile-bytecode.scm: Remove use of arities.scm.
Instead, incoming terms are expected to call their continuations
with the correct number of arguments.
* module/language/tree-il/compile-cps2.scm (adapt-arity): New
procedure. This is equivalent to (language cps arities), but as it
is a necessary pass and not an optimization it's more proper to put
it in the converter itself. Unlike with the nested CPS
representation, it's possible to look up continuations without
making a DFG.
(convert): Adapt arities as necessary.
