Optimisations.

[Nebuleon]

I attach some commits that optimise some rather common cases in interpreted opcodes:

• floating-point operations setting the rounding mode when that's not needed;
• inter-block branches (J*_OUT and B*_OUT) targetting the physical addresses (0x80000000 inclusive to 0xC0000000 exclusive).

[richard42]

Hi Nebuleon, thanks for this patch. In reviewing the changes, I think that the following instructions still need to retain the rounding mode set function call:

cvt_w_s
cvt_w_d
cvt_l_s
cvt_l_d

These instructions perform a conversion from floating-point to integer, which loses precision. As such, I think that their results will be affected by the current rounding mode. Aside from this, the other changes look good.

[Nebuleon]

cvt_[int]_[float] call into one of sixteen functions depending on the value of the N64 FCR31, each of which in turn calls one of eight C standard library functions (truncf, roundf, floorf, ceilf, trunc, round, floor, and ceil) which set their own native rounding modes as implied by their names.

Because none of the sixteen functions actually needs the rounding mode to already be set, the issue is not that the rounding mode is not needed, it's that the rounding mode is needlessly set twice. I am sorry for the confusion and the lumping into the same commit.

Reference:

mupen64plus-core/src/r4300/fpu.h

Line 162 in f2247c5

M64P_FPU_INLINE void cvt_w_s(float *source,int *dest)

[richard42]

I see, thanks for the explanation.