Floating Point
On the SPARC, double-precision operations use floating point registers in even-numbered pairs.
| fabss | A,D | | A | &rarr D | ||
| (single, works for double with fmovs) | ||||
| faddd | A,B,D | A + B &rarr D | ||
| fcmpd | A,B | compare A - B | ||
| fdivd | A,B,D | A / B &rarr D | ||
| fdtoi | A,B | convert double to integer | ||
| fitod | A,B | convert integer to double | ||
| fmovs | A,D | A &rarr D (single: must do two moves) | ||
| fmuld | A,B,D | A * B &rarr D | ||
| fnegs | A,D | - A &rarr D | ||
| (single, works for double with fmovs) | ||||
| fsqrtd | A,D | &radic A &rarr D | ||
| fsubd | A,B,D | A - B &rarr D | ||
A routine is provided to generate the instruction sequence for a float operation.