On 2018-06-16, at 12:06:48, Farley, Peter x23353 wrote: > And I got that instruction specification wrong. VFAEB is the extended > mnemonic form for VFAE with M4 set to 0 for byte length, so the "inverted" > instruction would be VFAEB V1,V1,V0,9 (adding 8 for mask bit zero to the M5 > operand to invert the results). > I stand corrected. Thanks.
This architecture has grown beyond human ken. Let the compiler do it, and hope the compiler author gets it right. -- gil
