Currently failing some guardian tests.
* libguile/tags.h: Refactor tagging so that tc3 bits for a pair live in
the SCM value, not in the heap words. Do the same for structs. This
more rational tagging strategy will make native code generation
easier. Note that this means that to check a heap pointer for its
type, you first have to ensure that it has the expected tc3, as not
all the type bits are on the heap.
(SCM_TYP3): Check the SCM tag type, not the bits in the cell.
(SCM_HAS_TYP3): New helper.
(SCM_I_CONSP): Redefine to just check the typ3.
(scm_tcs_cons_imcar, scm_tcs_cons_nimcar, scm_tcs_struct): Remove, as
they are no longer necessary.
* libguile/array-handle.c (scm_i_array_implementation_for_obj): Check
for heap objects before checking type bits, so we don't check pairs.
* libguile/evalext.c (scm_self_evaluating_p):
* libguile/gc.c (scm_i_tag_name):
* libguile/goops.c (scm_class_of)
* libguile/hash.c (scm_hasher):
* libguile/print.c (iprin1): Adapt to tagging changes.
* libguile/gc.c (scm_storage_prehistory): Register all displacements
here. There are the same displacements as before, unfortunately.
* libguile/list.c (SCM_I_CONS):
* libguile/pairs.c (scm_cons):
* libguile/pairs.h (scm_is_pair):
* libguile/vm-engine.h (CONS): Tag pairs with scm_tc3_pair.
* libguile/modules.c (scm_post_boot_init_modules):
* libguile/modules.h (SCM_MODULEP):
* libguile/struct.c (struct_finalizer_trampoline, scm_i_alloc_struct):
(scm_make_vtable_vtable):
* libguile/struct.h (SCM_STRUCTP, SCM_STRUCT_VTABLE_DATA):
(SCM_STRUCT_VTABLE_SLOTS):
* libguile/vm-i-scheme.c (make-struct): Adapt to struct tagging
changes.
* libguile/numbers.h (SCM_I_INUMP):
* module/rnrs/arithmetic/fixnums.scm (fixnum?, inline-fixnum?): Adapt
to the new fixnum tag.
* libguile/numbers.h (SCM_INEXACTP): Make sure of the tc3 before looking
at the cell type.
* libguile/vm.c:
(vm_error):
(vm_error_bad_instruction):
(vm_error_unbound):
(vm_error_unbound_fluid):
(vm_error_not_a_variable):
(vm_error_not_a_thunk):
(vm_error_apply_to_non_list):
(vm_error_kwargs_length_not_even):
(vm_error_kwargs_invalid_keyword):
(vm_error_kwargs_unrecognized_keyword):
(vm_error_too_many_args):
(vm_error_wrong_num_args):
(vm_error_wrong_type_apply):
(vm_error_stack_overflow):
(vm_error_stack_underflow):
(vm_error_improper_list):
(vm_error_not_a_pair):
(vm_error_not_a_bytevector):
(vm_error_not_a_struct):
(vm_error_no_values):
(vm_error_not_enough_values):
(vm_error_continuation_not_rewindable):
(vm_error_bad_wide_string_length):
(vm_error_invalid_address):
(vm_error_object):
(vm_error_free_variable): New internal helpers, implementing VM error
handling.
* libguile/vm-engine.h (VM_ASSERT): New helper macro.
(ASSERT, CHECK_OBJECT, CHECK_FREE_VARIABLE):
(PRE_CHECK_UNDERFLOW, PUSH_LIST): Use the new helper.
* libguile/vm-i-loader.c:
* libguile/vm-i-scheme.c:
* libguile/vm-i-system.c: Use VM_ASSERT and the out-of-line error
handlers.
* libguile/vm-engine.c (vm_engine): Remove inline error handlers, and
remove a couple of local vars. Use VM_ASSERT. Have halt handle the
return itself.
* libguile/boolean.c (scm_nil_p): New function.
* libguile/vm-i-scheme.c (nilp, not_nilp):
* libguile/vm-i-system.c (br_if_nil, br_if_not_nil): New instructions.
Renumber other ops.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Increment.
* module/language/assembly/compile-bytecode.scm (compile-bytecode): Add
support for writing `br-if-nil' and `br-if-not-nil' instructions.
* module/language/assembly/disassemble.scm (code-annotation): Add
`br-if-nil' and `br-if-not-nil' to the list of branch instructions.
* module/language/tree-il/compile-glil.scm: Add `nil?' to
`*primcall-ops*'.
(flatten): Use the new branch instructions for `nil?' conditionals.
* module/language/tree-il/primitives.scm: Add `nil?' to
`*interesting-primitive-names*', `*effect-free-primitives', and
`*effect+exception-free-primitives*'.
* libguile/vm-engine.h (DEAD): New macro, nulls out a value.
* libguile/vm-i-system.c:
* libguile/vm-i-loader.c:
* libguile/vm-i-scheme.c: Use DEAD when variables become dead.
Later we can #ifdef this out, but I want to give the buildbots a try
with this patch to make sure it's correct.
* libguile/foreign.c: Use `alignof_type' instead of `alignof'; the
latter was removed from Gnulib's <alignof.h> in
408e170e3ae81f73fb65686c5834693d89a96594 (Nov. 1 2011).
* libguile/vm-i-scheme.c: Likewise.
* libguile/_scm.h (SCM_OBJCODE_MAJOR_VERSION): Bump the major version,
indicating the first incompatibility between 2.0 and 2.2.
* libguile/vm-i-scheme.c (string-length, string-ref, vector-length): New
instructions.
* module/language/tree-il/compile-glil.scm (*primcall-ops*): Add
primcall ops for the new instructions.
* libguile/vm-engine.c (VM_CHECK_OBJECT, VM_CHECK_FREE_VARIABLES): Set
to 0 for both engines. These are really internal debugging variables,
which don't affect user-visible features, provided that the compiler
is correct of course.
(VM_CHECK_UNDERFLOW): New var, also off by default: whether to check
for stack underflow when popping values.
(vm_engine): Don't declare object_count if we are not checking object
table accesses.
* libguile/vm-engine.h (CACHE_PROGRAM): Don't muck with object_count
if we are not checking object table accesses.
(CHECK_UNDERFLOW, PRE_CHECK_UNDERFLOW): Nop out if we are not checking
underflow.
(POP2, POP3): New macros which check for underflow before popping more
than one value.
* libguile/vm-i-loader.c (load_array):
* libguile/vm-i-scheme.c (set_car, set_cdr, vector_set, slot_set)
(BV_SET_WITH_ENDIANNESS, BV_FIXABLE_INT_SET, BV_INT_SET)
(BV_FLOAT_SET):
* libguile/vm-i-system.c (partial_cont_call, fix_closure, prompt)
(fluid_set): Use POP2 / POP3.
(local_set, long_local_set): Pop to locals instead of using values on
the stack then dropping; allows for underflow to be checked before the
value is accessed.
(BR): Don't NULLSTACK or DROP after the operation.
(br_if, br_if_not, br_if_eq, br_if_not_eq, br_if_null)
(br_if_not_null): Pop to locals before doing the compare and jump.
* libguile/bytevectors.c:
* libguile/goops.c:
* libguile/instructions.c:
* libguile/numbers.c:
* libguile/random.c:
* libguile/read.c:
* libguile/vm-i-scheme.c: Fix a number of assumptions that a long could
hold an inum. This is not the case on platforms whose void* is larger
than their long.
* libguile/numbers.c (scm_i_inum2big): New helper, only implemented for
sizeof(void*) == sizeof(long); produces a compile error on other
platforms. Basically gmp doesn't have a nice interface for converting
between mpz values and intmax_t.
This results in a 17% improvement in the execution time of the "+" and
"-" benchmarks for fixnums.
* libguile/vm-i-scheme.c (ASM_ADD, ASM_SUB)[defined __x86_64__ &&
SCM_GNUC_PREREQ (4, 5)]: New macros.
(add)[defined ASM_ADD]: Use `ASM_ADD' for the fast path.
(sub)[defined ASM_SUB]: Use `ASM_SUB' for the fast path.
* test-suite/tests/numbers.test ("+")["fixnum + fixnum = bignum
(32-bit)", "fixnum + fixnum = bignum (64-bit)", "bignum + fixnum =
fixnum", "wrong type"]: New tests.
("-")["fixnum - fixnum = bignum (32-bit)", "fixnum - fixnum = bignum
(64-bit)", "bignum - fixnum = fixnum", "wrong type"]: New tests.
* test-suite/tests/00-initial-env.test ("goopsless")["+ wrong type
argument"]: Use `with-test-prefix/c&e' instead of `with-test-prefix'.
["- wrong type argument"]: New test prefix.
* libguile/vm-i-scheme.c (INUM_MAX, INUM_MIN): New macros.
(add1, sub1): Add/subtract without untagging the operand. This leads
to a 44% run time improvement compared to the previous
implementation.
* libguile/vm.c: Include <stdint.h>.
* test-suite/tests/numbers.test ("1+", "1-"): Add tests for
MOST-POSITIVE-FIXNUM, resp. MOST-NEGATIVE-FIXNUM, for 32-bit and
34-bit values thereof.
* benchmark-suite/benchmarks/arithmetic.bm: New file.
* benchmark-suite/Makefile.am (SCM_BENCHMARKS): Add it.
* libguile/vm-i-scheme.c (symbol?, vector?): New
instructions. Renumbered the rest.
* libguile/vm-i-system.c: Renumber instructions.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bump.
* module/ice-9/psyntax.scm (binding-type, binding-value): Define using
macros so that we inline to car and cdr opcodes. Oh, for an inliner :)
* module/language/tree-il/compile-glil.scm (*primcall-ops*)
* module/language/tree-il/primitives.scm
(*interesting-primitive-names*, *effect-free-primitives*)
(*effect+exception-free-primitives*): Add symbol? and vector?
inlines.
* libguile/vm-engine.c: Add func_name local, for error reporting.
(vm_error_apply_to_non_list): New error case.
(vm_error_wrong_type_arg): Remove this generic error case.
(vm_error_wrong_type_apply): Remove FUNC_NAME -- no sense in seeing
"vm-debug-engine" in the error report.
(vm_error_not_a_pair, vm_error_not_a_bytevector)
(vm_error_not_a_struct, vm_error_not_a_thunk): Use func_name instead
of FUNC_NAME, so we can indicate what caused the error.
* libguile/vm-i-scheme.c (VM_VALIDATE_CONS, car, cdr, set-car!)
(set-cdr!): Indicate provenance of errors.
(VM_VALIDATE_STRUCT, struct-vtable):
(VM_VALIDATE_BYTEVECTOR, BV_FIXABLE_INT_REF, BV_INT_REF)
(BV_FLOAT_REF, BV_FIXABLE_INT_SET, BV_INT_SET, BV_FLOAT_SET): Same.
* libguile/vm-i-system.c (apply, tail-apply): Use
vm_error_apply_to_non_list.
* libguile/vm-i-scheme.c (ALIGNED_P): New macro.
(BV_FIXABLE_INT_REF, BV_INT_REF, BV_FLOAT_REF, BV_FIXABLE_INT_SET,
BV_INT_SET, BV_FLOAT_SET): Check the alignment of the pointer instead
of checking "i % size == 0". This fixes bus errors on
`sparc64-linux-gnu'.
* libguile/vm.c: Include <alignof.h>.
* libguile/_scm.h (SCM_OBJCODE_MINOR_VERSION): Bump for make-struct
change.
* libguile/struct.c (scm_i_alloc_struct): Use scm_words instead of
scm_gc_malloc to simplify the code and inline the call to GC_MALLOC.
* module/language/tree-il/compile-glil.scm (*primcall-ops*): Compile
make-struct/no-tail to make-struct.
* module/language/tree-il/primitives.scm (define-primitive-expander):
Allow a conditional branch of #f to aboirt inlining.
(make-struct): Expand into make-struct/no-tail in the case that
tail-size is 0.
* libguile/vm-i-scheme.c (make-struct): Adapt to always assume tail-size
is 0. Inline allocation if possible. Don't decrement the SP past live
objects on the stack, which could cause GC to miss references. Use the
NULLSTACK macro.
* libguile/pairs.h (scm_is_null): Nil is also null.
* libguile/vm-i-scheme.c (not, not-not, null?, not-null?):
* libguile/vm-i-system.c (br-if-null, br-if-not-null): Remove some more
nil special cases.
Thanks to Andy for noticing this.
* libguile/vm-engine.h (SYNC_REGISTER, CACHE_REGISTER): Add comment.
* libguile/vm-i-scheme.c (make_struct): Call `SYNC_REGISTER ()' in all
cases since the GC is going to run.
(struct_ref, struct_set): Call `SYNC_REGISTER ()' on the slow path.
(BV_REF_WITH_ENDIANNESS, BV_FIXABLE_INT_REF, BV_INT_REF): Likewise.
(BV_FLOAT_REF): Always `SYNC_REGISTER ()'.
* 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/frames.c, libguile/instructions.c, libguile/objcodes.c,
libguile/programs.c, libguile/throw.c, libguile/vm-i-scheme.c,
libguile/vm.c: Replace uses of discouraged constructs by their
current counterparts.
This should also fix "condition is always true" warnings.
* libguile/vm-i-scheme.c (bv_u32_native_ref, bv_s32_native_ref,
bv_u32_native_set, bv_s32_native_set): Conditionalize use of
`BV_FIXABLE_INT_REF' vs. `BV_INT_REF' based on `SIZEOF_VOID_P'.
* libguile/vm-i-scheme.c (vector-ref, vector-set): Sync registers if we
call out to C.
* module/language/tree-il/compile-glil.scm (flatten-lambda): Add an
extra argument, the self-label, which should be the gensym under which
the procedure is bound in a <fix> expression.
(flatten): If we see a call to a lexical ref to the self-label in a
tail position, rename and goto instead of goto/args, which will tear
down the frame -- or will, in the future. It's a primitive form of
loop detection.
* module/language/tree-il/primitives.scm (zero?): Expand to (= x 0).
* libguile/vm-i-scheme.c: Add add1 and sub1 instructions.
* module/language/tree-il/compile-glil.scm: Compile 1+ and 1- to add1
and sub1.
* module/language/tree-il/primitives.scm (define-primitive-expander):
Add support for `if' statements in the consequent.
(+, -): Compile (- x 1), (+ x 1), and (+ 1 x) to 1- or 1+ as
appropriate.
(1-): Remove this one. Seems we forgot 1+ before, but we weren't
compiling it nicely anyway.
* test-suite/tests/tree-il.test ("void"): Fix expected compilation of (+
(void) 1) to allow for add1.
* libguile/objcodes.c (OBJCODE_COOKIE): Bump.
* libguile/vm-i-loader.c:
* libguile/vm-i-scheme.c:
* libguile/vm-i-system.c: Renumber instructions, so I can have a bit
more space to work.
* libguile/vm-i-scheme.c (BV_FIXABLE_INT_REF, BV_INT_REF):
(BV_FLOAT_REF, BV_FIXABLE_INT_SET, BV_INT_SET, BV_FLOAT_SET): Fix the
bounds check for the last element.
* libguile/vm-i-scheme.c (vector_ref, vector_set, BV_FIXABLE_INT_REF,
BV_INT_REF, BV_FLOAT_REF, BV_FIXABLE_INT_SET, BV_INT_SET,
BV_FLOAT_SET): Explicitly initialize all locals, to make some versions
of GCC happier. Patch by Dale P. Smith <dsmich@roadrunner.com>.
* libguile/instructions.c (scm_instruction_list): Fix a longstanding bug
in this humble function.
* libguile/vm-i-scheme.c (BV_FIXABLE_INT_SET, BV_INT_SET, BV_FLOAT_SET):
Fix some bugs in these macros -- now the bytevector ops work.
* module/language/tree-il/compile-glil.scm (*primcall-ops*): Compile
bytevector calls to VM ops.
* module/language/tree-il/primitives.scm
(*interesting-primitive-names*): Resolve bytevector calls to primitive
calls.
* module/language/tree-il/primitives.scm
(*interesting-primitive-names*): Resolve vector-ref and vector-set!.
* module/language/tree-il/compile-glil.scm (*primcall-ops*): And compile
vector-ref and vector-set! to their opcodes.
* libguile/vm-i-scheme.c (vector-ref, vector-set): New opcodes, placed
before the bytevector ops. The renumbering shouldn't affect anyone,
given that the bytevector ops were not yet used. Fix a few bugs in the
bytevector ops.
* libguile/instructions.h (SCM_VM_NUM_INSTRUCTIONS): Enlarge to 255. Not
sure what performance effects this will have.
* libguile/vm-engine.c: Add new error case, vm_error_not_a_bytevector.
* libguile/vm-engine.h: Don't assign specific registers for i386. Having
added the new VM vector ops, GCC 4.4 is erroring for me now.
* libguile/vm-i-scheme.c: Add bytevector-specific ops to the VM.
We don't actually use them yet, though.
