Skip to contentSkip to conversation

Resend: [PATCH] oprofile support for Power 5++

DISCUSSION Mike Wolf 5 participants 7 messages Jul 10, 2007 – Jul 11, 2007

lore
Share & tools
lore

Mike Wolf

author
#0 lore raw

The patch has not been included and there have been no comments so I'm resending.

This patch adds a new oprofile cpu type for Power 5 revision 3 chips. The new name is ppc64/power5++ and is used so that the performance counters can be set up correctly.

-------- linux-2.6.18.ppc64.orig/arch/powerpc/kernel/cputable.c 2006-09-19 22:42:06.000000000 -0500
+++ linux-2.6.18.ppc64/arch/powerpc/kernel/cputable.c 2007-06-11 12:29:47.000000000 -0500
@@ -236,6 +236,21 @@

		.oprofile_mmcra_sipr	= MMCRA_SIPR,
		.platform		= "power5",
	},

+ { /* Power5++ */
+ .pvr_mask = 0xffffff00,
+ .pvr_value = 0x003b0300,
+ .cpu_name = "POWER5+ (gs)",
+ .cpu_features = CPU_FTRS_POWER5,
+ .cpu_user_features = COMMON_USER_POWER5_PLUS,
+ .icache_bsize = 128,
+ .dcache_bsize = 128,
+ .num_pmcs = 6,
+ .oprofile_cpu_type = "ppc64/power5++",
+ .oprofile_type = PPC_OPROFILE_POWER4,
+ .oprofile_mmcra_sihv = MMCRA_SIHV,
+ .oprofile_mmcra_sipr = MMCRA_SIPR,
+ .platform = "power5+",
+ },

	{	/* Power5 GS */
		.pvr_mask		= 0xffff0000,
		.pvr_value		= 0x003b0000,
#1 lore raw
Quoted 7 lines

The patch has not been included and there have been no comments so I'm resending.

This patch adds a new oprofile cpu type for Power 5 revision 3 chips. The new name is ppc64/power5++ and is used so that the performance counters can be set up correctly.

Does it make more sense to call this "ppc64/power5+rev3"?

Mikey

Quoted 5 lines

--------
linux-2.6.18.ppc64.orig/arch/powerpc/kernel/cputable.c 2006-09-19 22:4

2:06.000000000 -0500

Quoted 1 line

+++ linux-2.6.18.ppc64/arch/powerpc/kernel/cputable.c 2007-06-11 12:29:47.000

000000 -0500

Quoted 28 lines
(quoted diff) +15 -0
@@ -236,6 +236,21 @@
 		.oprofile_mmcra_sipr	= MMCRA_SIPR,
 		.platform		= "power5",
 	},
+	{	/* Power5++ */
+		.pvr_mask		= 0xffffff00,
+		.pvr_value		= 0x003b0300,
+		.cpu_name		= "POWER5+ (gs)",
+		.cpu_features		= CPU_FTRS_POWER5,
+		.cpu_user_features	= COMMON_USER_POWER5_PLUS,
+		.icache_bsize		= 128,
+		.dcache_bsize		= 128,
+		.num_pmcs		= 6,
+		.oprofile_cpu_type	= "ppc64/power5++",
+		.oprofile_type		= PPC_OPROFILE_POWER4,
+		.oprofile_mmcra_sihv	= MMCRA_SIHV,
+		.oprofile_mmcra_sipr	= MMCRA_SIPR,
+		.platform		= "power5+",
+	},
 	{	/* Power5 GS */
 		.pvr_mask		= 0xffff0000,
 		.pvr_value		= 0x003b0000,

_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-dev

author
#2 lore raw
Quoted 10 lines · Michael Neuling
Quoted 7 lines

The patch has not been included and there have been no comments so I'm resending.

This patch adds a new oprofile cpu type for Power 5 revision 3 chips. The new name is ppc64/power5++ and is used so that the performance counters can be set up correctly.

Does it make more sense to call this "ppc64/power5+rev3"?

This is a change to support new counter setup for oprofile. It may be the same if there is a revision 4 or 5 etc. So since the internal name was ++ I followed that convention.

Quoted 43 lines

Mikey

Quoted 5 lines

--------
linux-2.6.18.ppc64.orig/arch/powerpc/kernel/cputable.c 2006-09-19 22:4

2:06.000000000 -0500

Quoted 2 lines

+++ linux-2.6.18.ppc64/arch/powerpc/kernel/cputable.c 2007-06-11 12:29:47.000

000000 -0500

Quoted 29 lines
(quoted diff) +15 -0
@@ -236,6 +236,21 @@
 		.oprofile_mmcra_sipr	= MMCRA_SIPR,
 		.platform		= "power5",
 	},
+	{	/* Power5++ */
+		.pvr_mask		= 0xffffff00,
+		.pvr_value		= 0x003b0300,
+		.cpu_name		= "POWER5+ (gs)",
+		.cpu_features		= CPU_FTRS_POWER5,
+		.cpu_user_features	= COMMON_USER_POWER5_PLUS,
+		.icache_bsize		= 128,
+		.dcache_bsize		= 128,
+		.num_pmcs		= 6,
+		.oprofile_cpu_type	= "ppc64/power5++",
+		.oprofile_type		= PPC_OPROFILE_POWER4,
+		.oprofile_mmcra_sihv	= MMCRA_SIHV,
+		.oprofile_mmcra_sipr	= MMCRA_SIPR,
+		.platform		= "power5+",
+	},
 	{	/* Power5 GS */
 		.pvr_mask		= 0xffff0000,
 		.pvr_value		= 0x003b0000,

_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-dev

#3 lore raw
Quoted 6 lines
Quoted 3 lines

Does it make more sense to call this "ppc64/power5+rev3"?

This is a change to support new counter setup for oprofile. It may be the same if there is a revision 4 or 5 etc. So since the internal name was ++ I followed that convention.

I'm not too fussed, but if rev 4 comes out, the counters change and we need power5+++, it's gonna look pretty silly :-)

Mikey

#4 lore raw
Quoted 6 lines · Michael Neuling
Quoted 6 lines
Quoted 3 lines

Does it make more sense to call this "ppc64/power5+rev3"?

This is a change to support new counter setup for oprofile. It may be the same if there is a revision 4 or 5 etc. So since the internal name was ++ I followed that convention.

There is an oprofile change already made, or in-flight, or at least submitted, that will be looking for the "power5++" string, so if the string changes here, it'll need to be changed there too.

( http://article.gmane.org/gmane.linux.oprofile/4670 )

Quoted 3 lines

I'm not too fussed, but if rev 4 comes out, the counters change and we need power5+++, it's gonna look pretty silly :-)

silly, but would be funny to see. :-)

Quoted 6 lines

Mikey
_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-dev

#5 lore raw
Quoted 7 lines · Will Schmidt
Quoted 5 lines · Michael Neuling
Quoted 5 lines
Quoted 2 lines

Does it make more sense to call this "ppc64/power5+rev3"?

This is a change to support new counter setup for oprofile. It may be the same if there is a revision 4 or 5 etc. So since the internal name was ++

I have no idea if there will be a revision 4, etc, but I'm assuming the behavior would be the same as rev 3. So I'm not in favor of changing this string to include "rev3" in it. But whatever decision is made, I hope it can be made soon. The OProfile community is testing a release candidate right now that includes the userspace portion of this issue which checks for the string "ppc64/power5++". We hope to GA this next oprofile release very soon.

Thanks.
-Maynard

Quoted 24 lines
Quoted 1 line
Quoted 1 line

I followed that convention.

There is an oprofile change already made, or in-flight, or at least submitted, that will be looking for the "power5++" string, so if the string changes here, it'll need to be changed there too.

( http://article.gmane.org/gmane.linux.oprofile/4670 )

Quoted 2 lines

I'm not too fussed, but if rev 4 comes out, the counters change and we need power5+++, it's gonna look pretty silly :-)

silly, but would be funny to see. :-)

Quoted 5 lines

Mikey
_______________________________________________
Linuxppc-dev mailing list
[email protected]
https://ozlabs.org/mailman/listinfo/linuxppc-dev

#6 lore raw
Quoted 11 lines · Maynard Johnson
Quoted 7 lines · Will Schmidt
Quoted 5 lines · Michael Neuling
Quoted 5 lines
Quoted 2 lines

Does it make more sense to call this "ppc64/power5+rev3"?

This is a change to support new counter setup for oprofile. It may be the same if there is a revision 4 or 5 etc. So since the internal name was ++

I have no idea if there will be a revision 4, etc, but I'm assuming the behavior would be the same as rev 3. So I'm not in favor of changing

The way the cputable patch is now, the rev 4 would match the base revision anyway. Maybe it makes more sense to make PVR xxxx01xx and xxxx02xx explicitly match the old power5+, and make everything else match power5++?

I guess it all depends on the chance of IBM doing another major rev of power5. Given it's current phase of product maturity I suppose it's not all that likely. Doing it this way saves yet another cputable entry as well, since it would otherwise mean two added entries instead of one.

-Olof