Hi David, On 6/12/19 7:48 AM, Brenken, David (EFS-GH2) wrote:
Thank you for your hint. Would anyone mind, if keep my insert implementation anyway? If I compare the pseudo code of the instruction manual to the insert implementation it looks nearly the same whereas it seems kind of difficult to directly map the pseudo code of the instruction manual to the general deposit instruction.
I think deposit is better, as tcg can emit the native deposit instruction of the host processor (which it does for x86). gen_insert() on the other hand needs 8 tcg ops.
Regarding the pseudo code, take a look at tcg/README. You'll see that the semantics of insert and deposit are the same. Thus, I think Richards fix is better.
Cheers, Bastian
