how to disable the multiplication optimization?

[Yixin Cao]

GCC always avoids multiplication instructions as much as possible, so a *= 10; will be translated into:

leal (%eax, %eax, 4), %eax
add %eax, %eax

I do know that it is due to historic reason, for that old CPU spent 30+ clock cycles to finish a multiplication
instruction. But it is never true now, while 3 clock cycles will suffice, so it brings the side affect of bigger object
files without any gain.
It might do it default for the consideration of compatibility, but my problem is how to disable it, for my code will
never run on a box been made a decade ago.

Yixin Cao

 

[Ian Collins]

GCC always avoids multiplication instructions as much as possible, so a
*= 10; will be translated into:

leal (%eax, %eax, 4), %eax
add %eax, %eax

That's a compiler specific feature, bast to ask on gcc list.

 

[Keith Thompson]

GCC always avoids multiplication instructions as much as possible,
so a *= 10; will be translated into: [snip]
It might do it default for the consideration of compatibility, but my
problem is how to disable it, for my code will never run on a box been
made a decade ago.

If reading the gcc documentation doesn't answer this, try gnu.gcc.help.

 

[Yixin Cao]

Following is what Intel C Compiler generates:
9: 8d 0c 89 lea (%ecx,%ecx,4),%ecx
c: 8d 04 48 lea (%eax,%ecx,2),%eax

So, it might not a be the gcc only setting.

Anyway, thanks, I have sent this to gnu.gcc.help to try to find an answer.

GCC always avoids multiplication instructions as much as possible,
so a *= 10; will be translated into: [snip]
It might do it default for the consideration of compatibility, but my
problem is how to disable it, for my code will never run on a box been
made a decade ago.

If reading the gcc documentation doesn't answer this, try gnu.gcc.help.

 

[christian.bau]

GCC always avoids multiplication instructions as much as possible, so a *= 10; will be translated into:

leal (%eax, %eax, 4), %eax
add %eax, %eax

I do know that it is due to historic reason, for that old CPU spent 30+ clock cycles to finish a multiplication
instruction. But it is never true now, while 3 clock cycles will suffice, so it brings the side affect of bigger object
files without any gain.
It might do it default for the consideration of compatibility, but my problem is how to disable it, for my code will
never run on a box been made a decade ago.

Yixin Cao

1. It's still likely to be faster.
2. Did you actually check how many bytes of code are generated in each
case?
3. What about running on the box made a decade from now?