Re: -mno-sse and -mno-sse2 broken on opteron?

[Steve Kargl <sgk at troutmask dot apl dot washington dot edu>]

On Thu, Jan 06, 2005 at 10:55:23AM -0800, Richard Henderson wrote:
> On Thu, Jan 06, 2005 at 07:41:52AM -0800, Steve Kargl wrote:
> > gcc.info implies that one can use -mno-sse{2} with x86_64.
> > If this is not true, then gcc.info needs to be fix and these
> > options should be disabled/removed for x86_64 targets.
> 
> It is true.  Indeed, one *wants* to use -mno-sse when compiling
> kernel code, so that you're sure that the compiler won't clobber
> these registers without your knowledge.
> 
> It's just that, for user-level code, -mno-sse isn't that useful
> for x86_64.
> 

>From gcc.info

`-mmmx'
`-mno-mmx'

`-msse'
`-mno-sse'

`-msse2'
`-mno-sse2'

`-msse3'
`-mno-sse3'

`-m3dnow'
`-mno-3dnow'
     These switches enable or disable the use of built-in functions
     that allow direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow
     extensions of the instruction set.

     *Note X86 Built-in Functions::, for details of the functions
     enabled and disabled by these switches.

     To have SSE/SSE2 instructions generated automatically from
     floating-point code, see `-mfpmath=sse'.


These options are to "enable or disable the use of built-in functions
     that allow direct access to the MMX, SSE, SSE2, SSE3 and 3Dnow
     extensions of the instruction set"

I tried to DISABLE the use of SSE and SSE2.  It was not possible.

If the options are only useful for integer-only code, then it
should be stated.  Here I assume that your statement "one *wants*
to use -mno-sse when compiling kernel code" means compiling 
kernel code where FP is verboten.  


-- 
Steve