GXRuntime: inline the hot paths of the FP gate and paired-single access

[dougchansan]

Profiling a recompiled Mario Kart: Double Dash on Apple Silicon (sample, race scene) put two runtime helpers near the top of self time inside StaticRecompCore::Run:

symbol	self samples	share
ppc_fp_available	869	~7.1%
psq_load_value + psq_store_value	1101	~9.0%

Generated code calls them once per instruction site — 121,874 FP sites and 19,642 paired-single sites on this title.

What changed

ppc_fp_available_inline() puts the MSR[FP] test in the header, leaving only the exception raise out of line. A running game always has MSR[FP] set, so the cost was the call, not the test.

ppc_psq_load_inline() / ppc_psq_store_inline() handle GQR type 0 — plain IEEE singles, no quantisation, no scale, which is what games leave GQR0 at — as two 32-bit accesses and a conversion. Quantised types (4..7), invalid types, and the LSQE illegal-instruction check all fall through to the existing out-of-line functions, so behaviour is unchanged.

Nothing new had to be exposed: f64_value, convert_to_double, convert_to_single_ftz and f64_bits were already static inline in core/types.h, and the memory accessors and PPC_HID2_LSQE were already header-visible.

ppc_fp_available(), ppc_psq_load() and ppc_psq_store() all remain real symbols — the LLVM backend emits calls to them by name.

Measured

Idle Apple M5, alternating A/B with a reversed-order block, largest competing process under 70%:

change	delta	ranges
inline FP gate	+11.8%	do not overlap
inline paired-single	+4.0%	do not overlap

Both beat their profiled share. An out-of-line call is also an optimisation barrier: at every generated site w, gqr_index, indexed and cia are literals, so inlining lets the compiler fold the LSQE test and the w branch away per site. The paired-single module came out slightly smaller despite 19,642 inlined sites, which is that folding showing up.

With PGO on top (+7.5%), MKDD under static recompilation on Metal went 51.1 → 62.5 fps single-player and 40.6 → 50.8 fps split-screen.

Ordering

Pairs with a DolRecomp change that emits the _inline forms. This one should land first — with the emitter updated and these absent, a port fails to compile.

[dougchansan]

Correction to the absolute fps figures in this description. The relative results are unaffected; the absolutes are roughly half what the hardware does.

Every macOS number I recorded before 2026-08-06 was measured with the laptop in Low Power Mode — a 2.02× CPU throttle I only found today, by toggling the setting and watching a previously recorded figure reproduce to four decimal places. So this line:

With PGO on top (+7.5%), MKDD under static recompilation on Metal went 51.1 → 62.5 fps single-player and 40.6 → 50.8 fps split-screen.

should read, unthrottled and on AC:

1P race, Metal          2.0040x   120.1 fps   sd 0.0136
split-screen, Metal     1.7099x   102.5 fps   sd 0.0235

The +11.8% and +4.0% in this PR stand unchanged. They were alternating A/B arms measured inside a single session, so a constant throttle scales both arms and cancels — it cannot manufacture a difference between them. That is precisely why the measurements were interleaved rather than compared against a previous day's number. What the throttle invalidates is the absolute fps, not the deltas.

Nothing in the diff changes. Flagging it only so the numbers in the description aren't read as this patch's ceiling — static recomp on this scene is comfortably past 60 fps, not scraping it.