Re: floating-point loads vs dreamcast
Marcus Comstedt <[email protected]> Fri, 13 May 2011 23:38:52 +0200
| Newsgroups | gmane.os.netbsd.ports.dreamcast |
|---|---|
| Message-ID | <[email protected]> |
der Mouse <[email protected]> writes: [...] > Any idea what's going on here? As I read the hardware docs, > fmov @rN,drM certainly ought to do a 64-bit double-precision load. Only if you set SZ to 1. PR controls the size of arithmetic operations, SZ controls the size of loads and stores. fmov @rN,drM and fmov.s @rN,frM has exactly the same encoding (1111nnnnmmmm1000) (only that m needs to be even in the 64 bit case), SZ controls which one you get. See section 10.37 in the Programming Manual. // Marcus