flat assembler
Message board for the users of flat assembler.
Index
> Compiler Internals > Support MOVBE Goto page Previous 1, 2 |
Author |
|
bitRAKE 07 Jul 2009, 20:35
|
|||
07 Jul 2009, 20:35 |
|
r22 08 Jul 2009, 20:49
Code: movbe eax,dword[string_a] movbe ebx,dword[string_b] Is more compact, but is it faster than... Code: mov eax,dword[string_a] mov ebx,dword[string_b] bswap eax bswap ebx In any case the string compare is a good example of usage for the MOVBE REG,MEM (although it wouldn't be useful for unicode). Can anyone think of a similar snippet/example/algo for MOVBE MEM,REG? The equivalent MOV / BSWAP / MOV would seem to benefit from this new opcode but I can't think of a situation where it would be done in iteration and where SIMD wouldn't be better served for the task. Does Intel have a rationalization for MOVBE? Maybe it was purely for the sake of completeness or did they find a bottle neck in a large algorithm that the Intel Compiler couldn't handle well. At this point in the x86 instruction set it's funny that BSWAP is getting a "service pack" |
|||
08 Jul 2009, 20:49 |
|
Goto page Previous 1, 2 < Last Thread | Next Thread > |
Forum Rules:
|
Copyright © 1999-2024, Tomasz Grysztar. Also on GitHub, YouTube.
Website powered by rwasa.