Re: [RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

Re: [RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

[Alan Cox]

On Llu, 2003-04-14 at 23:27, Dave Jones wrote:
> On Tue, Apr 15, 2003 at 12:21:10AM +0200, Adrian Bunk wrote:
>  > If my patch is wrong and this is a RTFM please give me a hint where to 
>  > find the "M".
>  > 
>  > The AMD K6-II and K6-III do support 3DNow!
> 
> The 3dnow memory copies aren't a win on anything
> earlier than an Athlon iirc.

Under 1% on a K6. The processor is so horribly memory bound the
actual copy function is borderline irrelevant. If its not in the
K6-III cache its not arriving in a hurry

[RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

[Adrian Bunk]

If my patch is wrong and this is a RTFM please give me a hint where to 
find the "M".

The AMD K6-II and K6-III do support 3DNow!

The patch below adds a new option for the AMD K6-II and K6-III and 
enables CONFIG_X86_USE_3DNOW.

I'm currently running 2.5.67 with this patch applied on my K6-II.

Questions:
Is it really correct to enable CONFIG_X86_USE_3DNOW?
Is the -march=k6-2 correct?

TIA
Adrian


--- linux-2.5.67/arch/i386/Kconfig.old	2003-04-14 22:17:56.000000000 +0200
+++ linux-2.5.67/arch/i386/Kconfig	2003-04-14 22:21:07.000000000 +0200
@@ -128,7 +128,8 @@
 	  - "Pentium-II" for the Intel Pentium II or pre-Coppermine Celeron.
 	  - "Pentium-III" for the Intel Pentium III or Coppermine Celeron.
 	  - "Pentium-4" for the Intel Pentium 4 or P4-based Celeron.
-	  - "K6" for the AMD K6, K6-II and K6-III (aka K6-3D).
+	  - "K6" for the AMD K6.
+	  - "K6_2" for the AMD K6-II and K6-III (aka K6-3D).
 	  - "Athlon" for the AMD K7 family (Athlon/Duron/Thunderbird).
 	  - "Crusoe" for the Transmeta Crusoe series.
 	  - "Winchip-C6" for original IDT Winchip.
@@ -200,12 +201,19 @@
 	  optimizations.
 
 config MK6
-	bool "K6/K6-II/K6-III"
+	bool "K6"
 	help
-	  Select this for an AMD K6-family processor.  Enables use of
-	  some extended instructions, and passes appropriate optimization
+	  Select this for an AMD K6 processor.  Enables use of some
+	  extended instructions, and passes appropriate optimization
 	  flags to GCC.
 
+config MK6_2
+	bool "K6-II/K6-III"
+	help
+	  Select this for an AMD K6-II and K6-III processors.  Enables
+	  use of some extended instructions, and passes appropriate
+	  optimization flags to GCC.
+
 config MK7
 	bool "Athlon/Duron/K7"
 	help
@@ -289,7 +297,7 @@
 config X86_L1_CACHE_SHIFT
 	int
 	default "4" if MELAN || M486 || M386
-	default "5" if MWINCHIP3D || MWINCHIP2 || MWINCHIPC6 || MCRUSOE || MCYRIXIII || MK6 || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || M586 || MVIAC3_2
+	default "5" if MWINCHIP3D || MWINCHIP2 || MWINCHIPC6 || MCRUSOE || MCYRIXIII || MK6 || MK6_2 || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || M586 || MVIAC3_2
 	default "6" if MK7 || MK8
 	default "7" if MPENTIUM4
 
@@ -335,7 +343,7 @@
 
 config X86_ALIGNMENT_16
 	bool
-	depends on MWINCHIP3D || MWINCHIP2 || MWINCHIPC6 || MCYRIXIII || MELAN || MK6 || M586MMX || M586TSC || M586 || M486 || MVIAC3_2
+	depends on MWINCHIP3D || MWINCHIP2 || MWINCHIPC6 || MCYRIXIII || MELAN || MK6 || MK6_2 || M586MMX || M586TSC || M586 || M486 || MVIAC3_2
 	default y
 
 config X86_GOOD_APIC
@@ -350,12 +358,12 @@
 
 config X86_USE_PPRO_CHECKSUM
 	bool
-	depends on MWINCHIP3D || MWINCHIP2 || MWINCHIPC6 || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMIII || MPENTIUMII || M686 || MK8 || MVIAC3_2
+	depends on MWINCHIP3D || MWINCHIP2 || MWINCHIPC6 || MCYRIXIII || MK7 || MK6 || MK6_2 || MPENTIUM4 || MPENTIUMIII || MPENTIUMII || M686 || MK8 || MVIAC3_2
 	default y
 
 config X86_USE_3DNOW
 	bool
-	depends on MCYRIXIII || MK7
+	depends on MCYRIXIII || MK6_2 || MK7
 	default y
 
 config X86_OOSTORE
@@ -478,7 +486,7 @@
 
 config X86_TSC
 	bool
-	depends on (MWINCHIP3D || MWINCHIP2 || MCRUSOE || MCYRIXIII || MK7 || MK6 || MPENTIUM4 || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || MK8 || MVIAC3_2) && !X86_NUMAQ
+	depends on (MWINCHIP3D || MWINCHIP2 || MCRUSOE || MCYRIXIII || MK7 || MK6 || MK6_2 || MPENTIUM4 || MPENTIUMIII || MPENTIUMII || M686 || M586MMX || M586TSC || MK8 || MVIAC3_2) && !X86_NUMAQ
 	default y
 
 config X86_MCE
--- linux-2.5.67/arch/i386/Makefile.old	2003-04-14 22:23:34.000000000 +0200
+++ linux-2.5.67/arch/i386/Makefile	2003-04-14 22:25:01.000000000 +0200
@@ -39,6 +39,7 @@
 cflags-$(CONFIG_MPENTIUMIII)	+= $(call check_gcc,-march=pentium3,-march=i686)
 cflags-$(CONFIG_MPENTIUM4)	+= $(call check_gcc,-march=pentium4,-march=i686)
 cflags-$(CONFIG_MK6)		+= $(call check_gcc,-march=k6,-march=i586)
+cflags-$(CONFIG_MK6_2)		+= $(call check_gcc,-march=k6-2,-march=i586)
 cflags-$(CONFIG_MK7)		+= $(call check_gcc,-march=athlon,-march=i686 $(align)-functions=4)
 cflags-$(CONFIG_MK8)		+= $(call check_gcc,-march=k8,$(call check_gcc,-march=athlon,-march=i686 $(align)-functions=4))
 cflags-$(CONFIG_MCRUSOE)	+= -march=i686 $(align)-functions=0 $(align)-jumps=0 $(align)-loops=0
--- linux-2.5.67/include/asm-i386/module.h.old	2003-04-14 22:25:59.000000000 +0200
+++ linux-2.5.67/include/asm-i386/module.h	2003-04-14 22:26:34.000000000 +0200
@@ -30,6 +30,8 @@
 #define MODULE_PROC_FAMILY "PENTIUM4 "
 #elif defined CONFIG_MK6
 #define MODULE_PROC_FAMILY "K6 "
+#elif defined CONFIG_MK6_2
+#define MODULE_PROC_FAMILY "K6II "
 #elif defined CONFIG_MK7
 #define MODULE_PROC_FAMILY "K7 "
 #elif defined CONFIG_MK8

Re: [RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

[Dave Jones]

On Tue, Apr 15, 2003 at 12:21:10AM +0200, Adrian Bunk wrote:
 > If my patch is wrong and this is a RTFM please give me a hint where to 
 > find the "M".
 > 
 > The AMD K6-II and K6-III do support 3DNow!

The 3dnow memory copies aren't a win on anything
earlier than an Athlon iirc.

		Dave

Re: [RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

[Robert Love]

On Mon, 2003-04-14 at 18:21, Adrian Bunk wrote:

> If my patch is wrong and this is a RTFM please give me a hint where to 
> find the "M".

Patch looks right...

> Questions:
> Is it really correct to enable CONFIG_X86_USE_3DNOW?

If you are right that the K6-2 and K6-3D support 3DNow!, then yes.  At
least its "correct" but it may not be optimal: I seem to recall 3DNow
memory copies were not worthwhile on anything before an Athlon.  Double
check that, though.

> Is the -march=k6-2 correct?

Yes.  Even if the above is true, splitting the K6 out like this is
useful for the extra -march here.  It certainly does not hurt (picking
the original K6 will give proper support for the whole family, in
needed).

	Robert Love

Re: [RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

[Paweł Gołaszewski]

On Tue, 14 Apr 2003, Robert Love wrote:
> > Questions: Is it really correct to enable CONFIG_X86_USE_3DNOW?
> If you are right that the K6-2 and K6-3D support 3DNow!, then yes.

Yes, it does.

-- 
pozdr.  Paweł Gołaszewski        
---------------------------------
worth to see: http://www.againsttcpa.com/
CPU not found - software emulation...

Re: [RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

[Adrian Bunk]

On Mon, Apr 14, 2003 at 11:06:28PM +0100, Alan Cox wrote:
> On Llu, 2003-04-14 at 23:27, Dave Jones wrote:
> > On Tue, Apr 15, 2003 at 12:21:10AM +0200, Adrian Bunk wrote:
> >  > If my patch is wrong and this is a RTFM please give me a hint where to 
> >  > find the "M".
> >  > 
> >  > The AMD K6-II and K6-III do support 3DNow!
> > 
> > The 3dnow memory copies aren't a win on anything
> > earlier than an Athlon iirc.
> 
> Under 1% on a K6. The processor is so horribly memory bound the
> actual copy function is borderline irrelevant. If its not in the
> K6-III cache its not arriving in a hurry

Ah thanks, this is the information I didn't know.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: [RFC][2.5 patch] K6-II/K6-II: enable X86_USE_3DNOW

[Adrian Bunk]

On Mon, Apr 14, 2003 at 06:36:20PM -0400, Robert Love wrote:
> On Mon, 2003-04-14 at 18:21, Adrian Bunk wrote:
> 
> > If my patch is wrong and this is a RTFM please give me a hint where to 
> > find the "M".
> 
> Patch looks right...
> 
> > Questions:
> > Is it really correct to enable CONFIG_X86_USE_3DNOW?
> 
> If you are right that the K6-2 and K6-3D support 3DNow!, then yes.  At
> least its "correct" but it may not be optimal: I seem to recall 3DNow
> memory copies were not worthwhile on anything before an Athlon.  Double
> check that, though.

It does, but Alan already explained why it doesn't make a difference.

> > Is the -march=k6-2 correct?
> 
> Yes.  Even if the above is true, splitting the K6 out like this is
> useful for the extra -march here.  It certainly does not hurt (picking
> the original K6 will give proper support for the whole family, in
> needed).

>From a quick look of the gcc-3.2.2 sources it seems that except for the 
3DNow! support gcc doesn't differenciate between K6 and K6-II, so this 
change alone doesn't seem to make it worth adding a new processor type.

> 	Robert Love

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed