I'm implementing binary translation and have to deal with sequences of NOPs (0x90) with length about 16 opcodes. Is it better for performance to place JMP (to the end) at start of such sequences?
The Intel Architecture Software developer's guide, volume 2B (instructions N-Z) contains the following table (pg 4-12) about NOP:
Table 4-9. Recommended Multi-Byte Sequence of NOP Instruction
Length Assembly Byte Sequence ================================================================================= 2 bytes 66 NOP 66 90H 3 bytes NOP DWORD ptr [EAX] 0F 1F 00H 4 bytes NOP DWORD ptr [EAX + 00H] 0F 1F 40 00H 5 bytes NOP DWORD ptr [EAX + EAX*1 + 00H] 0F 1F 44 00 00H 6 bytes 66 NOP DWORD ptr [EAX + EAX*1 + 00H] 66 0F 1F 44 00 00H 7 bytes NOP DWORD ptr [EAX + 00000000H] 0F 1F 80 00 00 00 00H 8 bytes NOP DWORD ptr [EAX + EAX*1 + 00000000H] 0F 1F 84 00 00 00 00 00H 9 bytes 66 NOP DWORD ptr [EAX + EAX*1 + 00000000H] 66 0F 1F 84 00 00 00 00 00H
This allows you to construct "padding NOP" of certain sizes. With two of those, you can bridge 16 Bytes, although I second the suggestion to check the optimization guides (for the CPU you're targeting) whether a JMP is faster than two such NOPs.
If the NOPs are to align the stream, then they have more value than just being a NO OP. if your concerned with pure speed, see Agner Fog's Optimization Manuals Vol. 4.
being a binary translation I would start by translating (them into equivalent nops on the target system). Once things are working then optimize out dead code. At the same time since this string of instructions caught your eye, try to understand what they were there for, perhaps waiting on hardware to do something, and make sure that your translated system functions the same.
来源：https://stackoverflow.com/questions/6776385/what-is-faster-jmp-or-string-of-nops