Index
> Main > Branch Likely Taken or Branch Unlikely Taken: mnemonic? |
| What should be FASM's mnemonics for Branch Likely Taken and Branch Unlikely Taken? | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Total Votes : 15 |
| Author |
|
|
rugxulo 11 Jun 2006, 19:16
Okay, let's get a feel for what people would prefer to use (although, obviously, the implementation is fully at Privalov's discretion).
|
|||
11 Jun 2006, 19:16 |
|
|
Tomasz Grysztar 11 Jun 2006, 19:26
I'd prefer to stick with CS and DS (unless Intel/AMD come up with some standard), taking into consideration that other segment prefixes may also have some applications (see http://board.flatassembler.net/topic.php?p=39873#39873).
|
|||
|
11 Jun 2006, 19:26 |
|
|
LocoDelAssembly 11 Jun 2006, 20:18
I agree with Tomasz, specially because you can always use EQU to add the mnemonics.
AMD has this feature? which cores? |
|||
|
11 Jun 2006, 20:18 |
|
|
LocoDelAssembly 12 Jun 2006, 14:42
If you have some code that demostrates the existence of the feature post it and I'll test it for you. I have a Venice core
Regards |
|||
|
12 Jun 2006, 14:42 |
|
|
MazeGen 13 Jun 2006, 12:39
What about some modification of the instruction mnemonic itself, e.g., lje (likely taken jump if equal), ujb (unlikely taken jump if below), ajnbe (alternating jump if not above or equal)? It is ugly, but not uglier than lt je and ut jb.
|
|||
|
13 Jun 2006, 12:39 |
|
|
LocoDelAssembly 13 Jun 2006, 13:01
Code: irps instr, ja jae jb jbe jc jcxz jecxz je jg jge jl jle jna jnae jnb jnbe jnc jne jng jnge jnl jnle jno jnp jns jnz jo jp jpe jpo js jz { macro l#instr offset\{ds instr offset\} macro u#instr offset\{cs instr offset\} ; Agner Fog's discovery macro a#instr offset\{fs instr offset\} } Nice idea MazeGen, I still think that Tomasz is right in wait for a standard. However you can use macros in the mean time 
Regards [edit]match directive removed and added Agner Fog's discovery[/edit] Last edited by LocoDelAssembly on 13 Jun 2006, 13:13; edited 1 time in total |
|||
|
13 Jun 2006, 13:01 |
|
|
Tomasz Grysztar 13 Jun 2006, 13:03
The "match" is redundant in your macro.
 |
|||
|
13 Jun 2006, 13:03 |
|
|
LocoDelAssembly 13 Jun 2006, 13:17
Corrected, I used match because without it the source didn't pass the preprossessing stage, however I had not this problem again :S I don't know which was the first problem but anyway, it works fine know
Regards |
|||
|
13 Jun 2006, 13:17 |
|
|
Reverend 13 Jun 2006, 22:05
This feature is documented
Intel manual "Volume 2A: Instruction Set Reference A-M" has a sub-chapter "Instruction prefixes" and there are mentioned the branch hints prefixes. EDIT: Ah, I misunderstood Tomasz's post  . I thought the branch taken/not taken hints are the undocumented prefixes, and showed a reference to such info. But the trick with FS is in fact undocumented. |
|||
|
13 Jun 2006, 22:05 |
|
< Last Thread | Next Thread > |
Forum Rules:
|