On 12/21/15 23:01, Richard Henderson wrote: > On 12/20/2015 07:30 AM, Chen Gang wrote: >> And we have to still check TILEGX_F_CALC_CVT, for they are really two >> different format: TILEGX_F_CALC_CVT has no HBIT, but TILEGX_F_CALC_NCVT >> has HBIT (which we need process it specially). > > The both do, in that you re-normalize to produce that HBIT. > That's the whole point. >
Oh, yes. But all together, we want to normalize the float value in fsingle_pack2, so we can not use float64_to_float32(), it assumes the input is already normalized (if we can let the input normalized, we will return directly). Thanks. -- Chen Gang (陈刚) Open, share, and attitude like air, water, and life which God blessed
