- Update to Softloat 3 from 2
- FREM and FMOD now generate the quotient bits in FPSR, required by Apple's SANE to do sin/cos/tan properly.
- FMOVE of a float to a Dx integer register generates the exception status bits, fixing square roots in SANE
- Rewrote how FMOVEM instructions are decoded and executed, fixing issues including skipping too few or too many
opcode bytes and causing serious weird behavior.
- FPU instructions all now have more realistic cycle timings for a 68881.
- All FPU instructions now generate exception bits in FPSR.
3rdparty/softfloat3: Updates [R. Belmont]
- Softfloat3 was always being built for a big-endian host, causing incorrect math on LE x64 and AArch64 machines.
- Fixed up Softfloat3 to build properly as part of MAME and up-ported the Bochs extensions. In latest Bochs, they
were only partially up-ported and Softfloat3 had been hacked up to be more like 2; here they're fixed to work
with stock Softfloat3.
This effectively reverts b380514764 and
c24473ddff, restoring the state at
598cd52272.
Before pushing, please check that what you're about to push is sane.
Check your local commit log and ensure there isn't anything out-of-place
before pushing to mainline. When things like this happen, it wastes
everyone's time. I really don't need this in a week when real work™ is
busting my balls and I'm behind where I want to be with preparing for
MAME release.
* softfloat3: fix msvc build (nw)
Is this how we do these things? I assume we don't touch the library code itself, and have to "fix" any problems via this header instead?
* softfloat3: do we like this better? (nw)
* softfloat3: it gets the hose again (nw)
* softfloat3: sonofa (nw)
This is a pre-usage import so everyone can get things in line for various compilers. GCC/Clang should be OK on most targets as-is;
MSVC may need to edit 3rdparty/softfloat3/build/MAME/build.h to conform. Please report failure/success with various compilers.
