Two related removals, and one optimisation that fell out of the second.
The cursor accessors go. xGetPtr(), xSetPtr() and
xGetAlignedCallTarget() were macros over x86Ptr, kept -- per their own
comment -- against a future change of storage scheme that never came
in fifteen years; xGetAlignedCallTarget() did no alignment at all, it
was the cursor under a misleading name. 23 files now read and assign
x86Ptr directly.
legacy_instructions.h goes with them. Its jump half (J8Rel/J32Rel and
the JE8/JMP32/... macros over them, plus x86SetJ8/J32/J32A) duplicated
what the C89 core already does with xe_fwd_jcc8/32 and xe_fwd_set8/32:
145 call sites across iFPU, iFPUd, iMMI, iR5900Branch, iR3000Atables
and iR3000A convert mechanically. Its SSE half -- eight
SSE*_XMM_to_XMM wrappers -- had exact xe_* equivalents already, same
prefix and opcode.
Four things needed more than a rename, and they are the reason this is
one commit rather than a sed:
* x86SetJ32A padded to a 16-byte boundary with NOPs before patching.
The core had no equivalent, so xe_fwd_set32_aligned is new rather
than the alignment being quietly dropped.
* The branch bookkeeping was typed u32* because the legacy API
returned that; the C89 slot is e_u8*. recDoBranchImm's signature,
recSetBranchEQ/L's returns, _setupBranchTest's function-pointer
parameter (now a Jcc code) and ~20 locals follow.
* JNS8 mapped to a condition the enum does not spell: there is no
Jcc_NotSigned, the member is Jcc_Unsigned = 0x9. Mapped explicitly
with a comment, since it reads like a typo otherwise.
* The SSE wrappers were stored in function-pointer tables
(recComOpXMM_to_XMM in iFPU, recFPUOpXMM_to_XMM in iFPUd). Macros
have no address, so the tables become switches -- and since op is
a constant at every call site, that removes an indirect call from
the recompiler's emit path and lets the emission inline. That is
the optimisation: the tables were costing a call to save nothing.
tests/emitter/newbind.cpp oracled the eight C wrappers; it now oracles
the macros the recompilers actually call, which is the better test
anyway, and the reference tree drops its include of the deleted
header.
Verification: gates 36/36 in both configurations, byte suites green
(95,692 + 112,118 + 27,337, zero divergent -- the newbind figure is
unchanged because the same 8x16x16 cases are still covered, through
the macros now), Tekken and the 2048 ELF clean at 3000 real-BIOS
frames, and all four JIT region hashes identical to the previous
build. 145 jump sites and two dispatch tables rewritten with no
change to a single emitted byte.
Three changes, one intent: bare `make` and bare cmake now build the C89
emitter with nothing to specify, and common/emitter contains only the six
headers the recompilers actually use (c89emit.h, x86emitter.h, x86types.h,
instructions.h, legacy_instructions.h, x86emitter_shim.h).
Default flip. Makefile gains `C89_EMITTER ?= 1` -- set BEFORE the
Makefile.common include, because the source exclusion there reads it. The
first attempt set it next to the CXXFLAGS append at line 839, after the
include: the define applied but the exclusion never fired, and the build
went on linking reference objects it no longer needed. Caught by `make -n`
grep: a bare plan now names zero reference objects, a C89_EMITTER=0 plan
names all eight. CMake's option defaults to ON with the same semantics
(-DC89_EMITTER=OFF for the reference).
Reference headers. internal.h, legacy_internal.h and the thirteen-header
implement/ tree move to tests/emitter/reference beside the sources that
need them. internal.h now includes the implement set itself -- it is the
reference implementation's own root and must be self-sufficient; x86types.h
includes them only under the reference configuration, for the extern
declarations in instructions.h's #else branches. The three G*Type operator
enums move into x86types.h: recompilers name their values, which makes them
API, not implementation. iR5900-32.cpp's MOVQ-workaround include of
internal.h was already reference-gated and is repointed. cmake's installed-
header list is repointed too (the stale entries failed configure).
Legacy jump API. The fourteen rel8/rel32 wrappers become macros over two
static helpers (J8Rel, J32Rel), matching the C89 core's idiom. The helpers
are static -- internal linkage per TU -- so there is no header-linkage
subtlety left to get wrong, where the previous inline definitions carried
one. The speculative conditional mode initially drafted into J32Rel was
removed before commit: only the unconditional form has callers.
The legacy SSE C-API wrappers (SSE_MINSS_XMM_to_XMM and friends) now emit
through E_SSE_RR directly instead of hopping through the C++ shim objects.
Their constants are proven end-to-end in newbind: reference function vs
C89 macro across all register pairs, 2,048 cases within the 110,070 total.
Verified: six suites green, switchover identical, both compile gates 31/31
in both configurations, bare make links at 11,351,104 bytes with zero
reference encoder symbols and boots, cmake configure passes in both
configurations. The cmake three-target build has not completed in this
container (cmake 4 needs -DCMAKE_POLICY_VERSION_MINIMUM=3.5 for the
vendored cpuinfo; GL dev packages were also missing) -- it needs one run on
a normally-provisioned machine. C89_EMITTER=0 static: the gates and suites
compile and byte-verify that configuration; its last full link ran two
commits ago unchanged.
l