On 3/21/15 01:03, Richard Henderson wrote: > On 03/18/2015 07:04 PM, Chen Gang wrote: >> For me, I am not quite sure about it, the related functional description >> is: >> >> rf[Dest] = signExtend32 ((int32_t) rf[SrcA] * (int32_t) rf[SrcB]); >> >> Do you mean it is equal to: >> >> rf[Dest] = signExtend32 (rf[SrcA] * rf[SrcB]); > > Yes, of course. > > Think about how multiplication works. Bit N of the input cannot affect any > output bit in [N-1,0]. Thus the first sign-extend cannot affect the output. >
OK, thanks. I shall skip the first type cast in patch v8. -- Chen Gang Open, share, and attitude like air, water, and life which God blessed