Match Hitbox::update

[skijer]

Changes

• Extract bsp->radius to a local before the multiply in Hitbox::update.
• Rewrite the scaledPos.y arithmetic as in-place mutations
  (+= hitboxElevation, *= scale.y, += postScaleHitboxElevation)
  instead of sum/summult temporaries.
• Promote kart/KartHitbox.cpp from NonMatching to Equivalent.

What this fixes

Hitbox::update now produces byte-identical instructions to the target.
Previously it diffed in ~12 instructions due to a float register
allocation cascade where postScaleHitboxElevation landed in f4
instead of f5. Extracting bsp->radius to a named local before
the multiply moves the load earlier in the schedule and shifts the
allocation.

[em-eight]

Looks good! Thank you for the contribution.

Can you also delete the MARK_FLOW_CHECK(0x805b7fbc); macro above the function? This is to signal that the function assembly is not exactly the same but it is functionally equivalent, but with your changes that is not needed

Also now since the whole file is matching now, you should be able to change it to Matching in configure.py!

[decomp-dev]

Report for RMCP01 (f5142e5 - 3e782da)

📈 Matched code: 5.76% (+0.00%, +236 bytes)

✅ 1 new match

Unit	Item	Bytes	Before	After
StaticR/kart/KartHitbox	Kart::Hitbox::update(const EGG::Vector3f&, const EGG::Quatf&, const EGG::Vector3f&, float, float)	+4	98.73%	100.00%