hi.

1. All instructions with at least 1 operand have the ModR/M byte?
2. reg/op field in ModR/M refers to the first operand (ex: ADD op1, op2)?
3. r/m field to the second one? (or vice-versa?) (or neither?)
4. Whats the difference of use an opcode extention (ex. ADD (80)) or not (ex. ADD (03 or 05))?
5. For example, why does INC instruction in the format 40+rw, does not appear in Appendix A or B, in those encoding tables?

This is my first real contact with the manual, so i'm still confuse.

Teehee,

1. No. Many arithmetic/logical instructions have special encodings for acc, imm operands. Other notable examples are inc/dec reg and instructions with implicit operands, like in / out.
   
2. No. Bit d (often bit 1) in opcode indicates whether destination is reg (d==1) or r/m (d==0).
   
3. See above.
   
4. add instruction opcode with first byte 80 corresponds to add r/m8, imm8, with 03 — to add reg, r/m, with 05 — to add acc, imm. I'd say they're completely different.
   
5. How's that? inc is on pages A-10 and B-13, dec — on A-11 and B-12.
   

hi baldr,

oh.. my Foxit finder didn't find it! i see it now! thanks!
and thanks for reply!

Hello.


Hows that? i didn't understand.

One of the solutions to branch prediction penalties was to allow the program to explicitly specify the likely path. This was supported on very few processor models because (it's a dumb idea and) the pipeline was shorted and an improved predictor makes it superfluous.

One prefix means branch is likely taken and other not-taken.