Re: [gentoo-amd64] CFLAGS for Core2?

[Paul de Vrieze]

On Wednesday 20 September 2006 22:25, Jack Lloyd wrote:
> Tomorrow my new Core 2 workstation should be arriving, and reading
> through the 2006.1 amd64 install notes and so forth, the recomended
> CFLAGS for EM64T CPUs is -march=nocona. However as that is tuned for
> 64-bit enabled Netburst rather than Conroe, I'm wondering if anyone
> had thoughts on good flags. In particular as (to the extent that I
> know anything about microarchitecture), Conroe seems closer to a K8
> than a P4, especially with regards to pipeline length and execution
> resources, I'm not sure that P4/Netburst tuning is the right thing to
> do. But the GCC docs say that -march=k8 enables 3dnow, which Intel
> chips don't have. So, right now the "seems likely" options would
> be:
>
> -march=k8 -mno-3dnow -msse3
> -march=nocona
> -march=pentium-m -m64 [saw this suggested in the forums, seems like a bad
> idea]
>
> Or can I get away with just something like
>
> -mtune=k8 -msse3   (will this get me 64-bit code?)
>
> To be clear, my main goals here are a) enable generation of all
> instructions the CPU has, b) keep GCC from generating 3dnow,etc
> instructions, and c) get instruction scheduling that is at least
> moderately decent for the uarch I'm using.
>
> Since SSE/SSE2, -mfpmath=sse, and argument passing via registers are
> default in 64-bit mode, I suspect the only other options I might try
> are -momit-leaf-frame-pointer, -frename-registers, and
> -finline-functions.
>
Be aware that -march and -mtune can be used together. This means that you 
use -march=nocona to specify that such code must be generated, but you can 
try different options for -mtune that only influences code scheduling and 
which of alternate code options is chosen.

Paul

-- 
Paul de Vrieze
Gentoo Developer
Mail: pauldv@g.o
Homepage: http://www.devrieze.net