x86: Update Intel Atom processor family optimization

Message ID 20200516140814.151673-1-hjl.tools@gmail.com
State New
Headers show
Series
  • x86: Update Intel Atom processor family optimization
Related show

Commit Message

H.J. Lu via Libc-alpha May 16, 2020, 2:08 p.m.
Enable Intel Silvermont optimization for Intel Goldmont Plus.  Detect more
Intel Airmont processors.  Optimize Intel Tremont like Intel Silvermont
with rep string instructions.
---
 sysdeps/x86/cpu-features.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

-- 
2.26.2

Comments

H.J. Lu via Libc-alpha May 19, 2020, 12:21 p.m. | #1
On Sat, May 16, 2020 at 7:08 AM H.J. Lu <hjl.tools@gmail.com> wrote:
>

> Enable Intel Silvermont optimization for Intel Goldmont Plus.  Detect more

> Intel Airmont processors.  Optimize Intel Tremont like Intel Silvermont

> with rep string instructions.

> ---

>  sysdeps/x86/cpu-features.c | 20 +++++++++++++++++++-

>  1 file changed, 19 insertions(+), 1 deletion(-)

>

> diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c

> index bfb415f05a..b1e4c2d1aa 100644

> --- a/sysdeps/x86/cpu-features.c

> +++ b/sysdeps/x86/cpu-features.c

> @@ -348,18 +348,23 @@ init_cpu_features (struct cpu_features *cpu_features)

>             case 0x57:

>               /* Knights Landing.  Enable Silvermont optimizations.  */

>

> +           case 0x7a:

> +             /* Unaligned load versions are faster than SSSE3

> +                on Goldmont Plus.  */

> +

>             case 0x5c:

>             case 0x5f:

>               /* Unaligned load versions are faster than SSSE3

>                  on Goldmont.  */

>

>             case 0x4c:

> +           case 0x5a:

> +           case 0x75:

>               /* Airmont is a die shrink of Silvermont.  */

>

>             case 0x37:

>             case 0x4a:

>             case 0x4d:

> -           case 0x5a:

>             case 0x5d:

>               /* Unaligned load versions are faster than SSSE3

>                  on Silvermont.  */

> @@ -370,6 +375,19 @@ init_cpu_features (struct cpu_features *cpu_features)

>                     | bit_arch_Slow_SSE4_2);

>               break;

>

> +           case 0x86:

> +           case 0x96:

> +           case 0x9c:

> +             /* Enable rep string instructions, unaligned load, unaligned

> +                copy, pminub and avoid SSE 4.2 on Tremont.  */

> +             cpu_features->feature[index_arch_Fast_Rep_String]

> +               |= (bit_arch_Fast_Rep_String

> +                   | bit_arch_Fast_Unaligned_Load

> +                   | bit_arch_Fast_Unaligned_Copy

> +                   | bit_arch_Prefer_PMINUB_for_stringop

> +                   | bit_arch_Slow_SSE4_2);

> +             break;

> +

>             default:

>               /* Unknown family 0x06 processors.  Assuming this is one

>                  of Core i3/i5/i7 processors if AVX is available.  */

> --

> 2.26.2

>


I am checking it in.

-- 
H.J.

Patch

diff --git a/sysdeps/x86/cpu-features.c b/sysdeps/x86/cpu-features.c
index bfb415f05a..b1e4c2d1aa 100644
--- a/sysdeps/x86/cpu-features.c
+++ b/sysdeps/x86/cpu-features.c
@@ -348,18 +348,23 @@  init_cpu_features (struct cpu_features *cpu_features)
 	    case 0x57:
 	      /* Knights Landing.  Enable Silvermont optimizations.  */
 
+	    case 0x7a:
+	      /* Unaligned load versions are faster than SSSE3
+		 on Goldmont Plus.  */
+
 	    case 0x5c:
 	    case 0x5f:
 	      /* Unaligned load versions are faster than SSSE3
 		 on Goldmont.  */
 
 	    case 0x4c:
+	    case 0x5a:
+	    case 0x75:
 	      /* Airmont is a die shrink of Silvermont.  */
 
 	    case 0x37:
 	    case 0x4a:
 	    case 0x4d:
-	    case 0x5a:
 	    case 0x5d:
 	      /* Unaligned load versions are faster than SSSE3
 		 on Silvermont.  */
@@ -370,6 +375,19 @@  init_cpu_features (struct cpu_features *cpu_features)
 		    | bit_arch_Slow_SSE4_2);
 	      break;
 
+	    case 0x86:
+	    case 0x96:
+	    case 0x9c:
+	      /* Enable rep string instructions, unaligned load, unaligned
+	         copy, pminub and avoid SSE 4.2 on Tremont.  */
+	      cpu_features->feature[index_arch_Fast_Rep_String]
+		|= (bit_arch_Fast_Rep_String
+		    | bit_arch_Fast_Unaligned_Load
+		    | bit_arch_Fast_Unaligned_Copy
+		    | bit_arch_Prefer_PMINUB_for_stringop
+		    | bit_arch_Slow_SSE4_2);
+	      break;
+
 	    default:
 	      /* Unknown family 0x06 processors.  Assuming this is one
 		 of Core i3/i5/i7 processors if AVX is available.  */