Re: floating-point loads vs dreamcast

[Marcus Comstedt <marcus%mc.pp.se@localhost>]

Some systematic test results:


fmov @Rm,FRn

SZ PR      n=0000                    n=0001
0  0       32bit->FR0                32bit->FR1
0  1       32bit->FR0                32bit->FR1
1  0       64bit->DR0(FR0&FR1)       64bit->XD0(XF0&XF1)
1  1       64bit->DR0(FR0&FR1)       64bit->XD0(XF0&XF1)


fmov FRm,FRn

SZ PR      m=0000,n=0010   m=0000,n=0011   m=0001,n=0010   m=0001,n=0011
0  0       FR0->FR2        FR0->FR3        FR1->FR2        FR1->FR3
0  1       FR0->FR2        FR0->FR3        FR1->FR2        FR1->FR3
1  0       DR0->DR1        DR0->XD1        XD0->DR1        XD0->XD1
1  1       DR0->DR1        DR0->XD1        XD0->DR1        XD0->XD1


So basically page 271 is completely correct, but page 275 is wrong in
that it should say "SZ = 1" instead of "PR = 1".


Sorry for the confusion


  // Marcus