Merge changes in the way that fasl streams are compressed. The new
approach makes compression explicit in the fasl representation, which
means that tricks like uzing zcat on a fasl file will no longer work
(at least not efficiently).
original commit: 167ac7294a2dc400821e4336f0cfc4de621efe97
This first cut is heavy on describing how backends work, because
that's fresh in mind and relevant to ongoing effort.
original commit: 964ce95dc910e6c1825b341bf05492af5903cc82
Commit 88d57da163 incorrectly merged, because the syntax of `mref` has
changed to support unboxed floating-point arithmetic.
original commit: a8eeb536aecc0e0a363aa68d43edafa7d6c9df52
Especially fix problems with `__collect_safe`, but also deallocate
stack space for `&` returns.
original commit: b593b0f14e7a803d8e089f49e266194245504197
Besides adding supportt for `__collect-safe` and other repairs,
introduce a write-write fence with the write barrier, which is
intended to avoid one thread using an object created in another thread
before the object's initializing writes are visible.
original commit: 543bd16739c08e5a8f88c470b52db0f23a27d260
Maintain 16-byte stack alignment (if incoming frames are 16-byte
aligned), because newer versions of gcc count on 16-bit alignment for
some generated SSE instructions.
[This patch is extracted from cisco/ChezScheme#518.
Merging the rest of that change will take longer.]
original commit: 88d57da163ca287ab37534fd858361a035ce7e4a
Besides updating for unboxed floating point, the ppc32 build uses a
return register, and the continuation-attachments implementation was
not right for that mode.
original commit: dd2d01fb26ace819c73f258b9b53739f9dda1d34
Flonum operations like `fltruncate` and `flsin` are implemented by
calling functions from the C library. Unboxing these involves a
generalazation the `foreign-call` intermediate form to handle unboxing
and to work in a non-tail position (especially by telling the register
allocator that caller-saved registers will be trashed). An internal
'atomic convention on a foreign call indicates that no callback into
Scheme is possible, so some setup/teardown (including stashing
callee-saved registers) can be skipped.
original commit: fd89919634d0d5272e046b47bb81bcc66e22a741
Shift addition of boxing as needed into the main loop, infer unboxed
variables and `mref`s, and centralize lifting of the `unboxed-fp`
declaration.
original commit: ed8ca4b6c77bdd436b0dee467a8350a450a44fb3
The comparison was off for 32-bit plaforms, because it didn't allow
fractional increments, The comparison was off for 64-bit platforms,
bbecause it didn't account for round-trip failure when starting from
the largest fixnum.
original commit: 74eb0583ae1b6212fbde459d7486c3d4a0498401
Simplify and normalize backend elements for loading, storing, and
converting floating-point numbers, taking better advantage of
new support for floating-pointer registers.
original commit: 4066af9cf3799392ef785a77da69f7cfff74d2fe
This is a follow-up to 276f8da076, where `(%tc-ref cp)` was supposed
to be preserved by moving it into %cp, but intrinisics for bytevector
arguments can kill %cp. Use a temporary to expose things properly to
the register allocator.
original commit: 3a29db06a452e46e69ebcde524b3b9acb435dec3
This reverts commit aa230ac79bed1efa02779bb7bbcde5c009818b74, so it
can be replaced with a solution that is less clumsy and less fragile.
original commit: 533940fdc6905d810deabb457d7004a031a3ac05
This is a follow-up to 276f8da076, where `(%tc-ref cp)` was preserved
by moving into `%cp`, but I missed that intrinisics for string and
bytevector arguments kill `%cp`.
original commit: aa230ac79bed1efa02779bb7bbcde5c009818b74
Fix nested-loop handling with union find, and fix unboxing check
to properly account for a primitive's arity.
original commit: af2f8784d96f09624b97476fb7b674a77097f9d0
Preinfo recoreds were mutated under the incorrect assumption that they
were specific to a `lambda` form, which is not true, especially after
some inlining.
original commit: 489b4c732d7359b1c7c4558d41d5e6ba3a9abafa
The problem especially affects `enable-object-counts`, where mark mode
can update an old rtd instead of a copy that is already made.
original commit: 31e2df63894d39b005c6b9984745b44409198d02
Avoid allocating a flonum object for floating-opint calculations
that are consumed only by other floating-point caculations.
For this first cut, unboxing applies only to fl+, fl-, fl*, fl/,
flabs, fl<, fl<=, fl=, fl>, fl>=, bytevector-ieee-double-[native-]ref,
and bytevector-ieee-double-[native-]set!. Local variables can be
unboxed in the same way as implicit temporaries, and loop arguments
can be unboxed, but values in a closure and function-call arguments
are always boxed.
arm32 support is mostly in place, but not yet right. ppc32 support is
not yet implemented.
This commit includes a small change that is incompatible with previous
Chez Scheme versions: `(fl= +nan.0)` (and similar for other
comparisons) produces true instead of false.
original commit: 36459e43f10705aa3e383376ca7d54cf2998b7ee
Kent noticed the bug, where `make-nontail` used `ignored` where it
should have used `effect`. Also, Kent points out that `make-nontail`
can skip the `$values` wrapped in `effect` contexts in unsafe mode.
original commit: edd9ba3d656f3bd712e5e235b77225f756397077
