Avoiding systematic conversion from source vectors to property alists
saves 20% on the final heap size of a process doing:
(compile-file FILE #:optimization-level 1)
where FILE is large.
* module/language/tree-il.scm (tree-il-src/ensure-alist): New procedure
with setter. Export as 'tree-il-src'.
* module/ice-9/psyntax.scm (build-void, build-call)
(build-conditional, build-lexical-reference, build-lexical-assignment)
(build-global-reference, build-global-assignment)
(build-global-definition, build-simple-lambda, build-case-lambda)
(build-lambda-case, build-primcall, build-primref)
(build-data, build-sequence, build-let, build-named-let)
(build-letrec, expand-body): Remove (sourcev->alist src) calls.
* module/ice-9/psyntax-pp.scm: Regenerate.
* module/language/tree-il/analyze.scm (shadowed-toplevel-analysis): Use
'tree-il-src' instead of accessing the 'src' slot directly.
* module/system/vm/assembler.scm (link-debug): Adjust so PC can be
followed by a vector or an alist.
* libguile/srfi-14.i.c: Update from Unicode 14.0.0.
* libguile/unidata_to_charset.pl (compute): Write arrays as static const
data, to avoid polluting the namespace and to avoid adding these to the
GC root set.
* libguile/srfi-14.c (SCM_CODEPOINT_F_IMMUTABLE): New flag.
(scm_charset_is_immutable): New internal predicate.
(SCM_VALIDATE_MUTABLE_CHARSET): New internal validator.
(cs_full_ranges, cs_full_ranges_len): Re-express as separate ranges and
len, because the ranges pointer in scm_t_char_set is mutable.
(scm_char_set_unfold_x, scm_list_to_char_set_x)
(scm_string_to_char_set_x, scm_char_set_filter_x)
(scm_i_ucs_range_to_char_set, scm_char_set_adjoin_x)
(scm_char_set_delete_x, scm_char_set_complement_x)
(scm_char_set_unfold_x, scm_char_set_intersection_x)
(scm_char_set_difference_x, scm_char_set_xor_x): Require mutable
charsets.
(define_charset): Add immutable flag.
(scm_init_srfi_14): Adapt initial charset definitions.
* module/ice-9/psyntax.scm (source-annotation): Only return source
properties from syntax objects.
(source-wrap): Don't look for source properties.
(expand-macro): Rebuild source properties on macro output via
source-wrap, not source properties. Only annotate head of a chain of
pairs.
(strip): Here's the only use of set-source-properties!: when stripping
a syntax object to a datum.
(macroexpand): If the input expression is not a syntax object, eagerly
extract its source properties.
(datum->syntax): Fix case in which source is given as an alist.
* module/ice-9/psyntax-pp.scm: Regenerate.
* prebuilt/i686-pc-linux-gnu/Makefile.am:
* prebuilt/mips-unknown-linux-gnu/Makefile.am:
* prebuilt/x86_64-unknown-linux-gnu/Makefile.am: Only prebuild eval.go,
psyntax-pp.go, and boot-9.go. The builder can handle the rest.
Otherwise the tarball size would explode, given the expansion in
bootstrap.am.
* configure.ac:
* Makefile.am (SUBDIRS): Replace bootstrap/ with stage0, stage1, and
stage2.
* am/bootstrap.am: Include all files and all rules.
* meta/build-env.in (GUILE_AUTO_COMPILE): Always turn off
auto-compilation. Take a GUILE_BOOTSTRAP_STAGE argument, which can be
stage0, stage1, stage2, or unset. Adapt the load-compiled path
accordingly.
* meta/uninstalled-env.in: Include .go files from stage2.
* module/Makefile.am: Rework to use boostrap.am.
* module/system/base/optimize.scm (available-optimizations): Punt the
inlinable-exports machinery to -O2.
* stage0/Makefile.am:
* stage1/Makefile.am:
* stage2/Makefile.am: New files.
* configure.ac: Check if the compiler supports link-time optimization.
If it does, turn it on. Saves around 15% off libguile binary size.
Controllable via --enable-lto configure option.
* module/language/tree-il/peval.scm (peval): The intention was for
resolve-interface to return an interface only if it exists, but actually
it throws if the interface isn't know. Fix to do what we intended to
do.
* module/language/tree-il/inlinable-exports.scm (compute-decoder): Fix
bug in which inlinable exports in a pure module would have a missing
make-struct/no-tail binding.
* libguile/numbers.c (log_of_shifted_double, scm_log10): Avoid complex
extension when the argument is a real nan.
* test-suite/tests/numbers.test: Tests for nans of either sign.
* libguile/numbers.c (scm_abs): As stated. When x is a nan with the sign bit
set, this changes the behavior of (magnitude x) back to what it was in 3.0.7,
to clear that bit.
* libguile/integers.c (scm_integer_abs_z): Just call
scm_integer_negate_z if negation is needed.
(scm_integer_sub_zz): Use mpz_sub instead of cloning bignum.
Calling out to Scheme was a performance regression.
* libguile/integers.h:
* libguile/integers.c (scm_integer_expt_ii, scm_integer_expt_zi): New
internal functions.
* libguile/numbers.c (scm_integer_expt): Go back to C. But, include
fast cases for inums and doubles.
* module/ice-9/boot-9.scm: Revert addition of integer-expt.
* libguile/deprecated.c (make_bignum): Move here from numbers.c, to
support scm_i_long2big etc.
(scm_i_big2dbl):
(scm_i_long2big):
(scm_i_ulong2big):
(scm_i_clonebig):
(scm_i_normbig): Deprecate.
(scm_install_gmp_memory_functions): Deprecate, happily! SCM bignums now
have digits allocated inline with the bignum itself, so they are
completely transparent to the GC already. The price is that if GMP ever
allocates digits via the MPZ API, those digits then have to be copied
back into managed memory. But we avoid having to install finalizers and
we avoid having to muck with GMP's allocator.
* libguile/numbers.c (scm_from_mpz): Use scm_integer_from_mpz.
(scm_init_numbers): Never muck with GMP's allocators.
* doc/ref/guile-invoke.texi (Environment Variables): Remove note about
GUILE_INSTALL_GMP_MEMORY_FUNCTIONS.
* meta/build-env.in: No need to set GUILE_INSTALL_GMP_MEMORY_FUNCTIONS.
