[Mesa-dev] [PATCH 1/2] mesa/x86: add SSE4.1 runtime detection

[Ian Romanick]

NAK.

common_x86 is only used for 32-bit builds, and there is nothing
equivalent for 64-bit.  As a result, this will disable the optimization
completely for 64-bit.

Gallium has a bunch of CPU detection code, so if we need to update the
CPU detection code in core Mesa, we should move the Gallium code to a
higher level in the tree to be shared.  I believe this has been
discussed before.

On 05/02/2014 11:11 AM, Matthew Atwood wrote:
> From: Matt Atwood <matthew.s.atwood at intel.com>
> 
> add in bit to _mesa_x86_features for sse4.1, along with macros to query
> ---
>  src/mesa/x86/common_x86.c          | 3 +++
>  src/mesa/x86/common_x86_features.h | 3 +++
>  2 files changed, 6 insertions(+)
> 
> diff --git a/src/mesa/x86/common_x86.c b/src/mesa/x86/common_x86.c
> index 3c1adc9..87fa68a 100644
> --- a/src/mesa/x86/common_x86.c
> +++ b/src/mesa/x86/common_x86.c
> @@ -238,6 +238,7 @@ _mesa_get_x86_features(void)
>  
>         /* get cpu features */
>         cpu_features = _mesa_x86_cpuid_edx(1);
> +       cpu_features_ecx = _mesa_x86_cpuid_ecx(1);
>  
>         if (cpu_features & X86_CPU_FPU)
>  	   _mesa_x86_cpu_features |= X86_FEATURE_FPU;
> @@ -254,6 +255,8 @@ _mesa_get_x86_features(void)
>  	   _mesa_x86_cpu_features |= X86_FEATURE_XMM;
>         if (cpu_features & X86_CPU_XMM2)
>  	   _mesa_x86_cpu_features |= X86_FEATURE_XMM2;
> +       if (cpu_features & x86_CPU_SSE4_1)
> +	   _mesa_x86_features |= X86_FEATURE_SSE4_1;
>  #endif
>  
>         /* query extended cpu features */
> diff --git a/src/mesa/x86/common_x86_features.h b/src/mesa/x86/common_x86_features.h
> index 8625484..1e4bbfc 100644
> --- a/src/mesa/x86/common_x86_features.h
> +++ b/src/mesa/x86/common_x86_features.h
> @@ -43,6 +43,7 @@
>  #define X86_FEATURE_XMM2	(1<<6)
>  #define X86_FEATURE_3DNOWEXT	(1<<7)
>  #define X86_FEATURE_3DNOW	(1<<8)
> +#define X86_FEATURE_SSE4_1	(1<<9)
>  
>  /* standard X86 CPU features */
>  #define X86_CPU_FPU		(1<<0)
> @@ -50,6 +51,7 @@
>  #define X86_CPU_MMX		(1<<23)
>  #define X86_CPU_XMM		(1<<25)
>  #define X86_CPU_XMM2		(1<<26)
> +#define X86_CPU_SSE4_1		(1<<19)
>  
>  /* extended X86 CPU features */
>  #define X86_CPUEXT_MMX_EXT	(1<<22)
> @@ -62,6 +64,7 @@
>  #define cpu_has_xmm2		(_mesa_x86_cpu_features & X86_FEATURE_XMM2)
>  #define cpu_has_3dnow		(_mesa_x86_cpu_features & X86_FEATURE_3DNOW)
>  #define cpu_has_3dnowext	(_mesa_x86_cpu_features & X86_FEATURE_3DNOWEXT)
> +#define cpu_has_sse4_1		(_mesa_x86_cpu_features & X86_FEATURE_SSE4_1)
>  
>  #endif
>  
>