All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH V3 1/1] perf, Add support for Xeon-Phi Knights Corner PMU
@ 2012-09-26 18:12 Vince Weaver
  2012-10-05  7:24 ` [tip:perf/urgent] perf/x86: Add support for Intel " tip-bot for Vince Weaver
  0 siblings, 1 reply; 2+ messages in thread
From: Vince Weaver @ 2012-09-26 18:12 UTC (permalink / raw)
  To: linux-kernel
  Cc: Peter Zijlstra, Cyrill Gorcunov, Paul Mackerras, Ingo Molnar,
	Arnaldo Carvalho de Melo, eranian, Meadows, Lawrence F

Hello

Version 3 of the patch simply renames all occurances of PHI to KNC
as Xeon Phi is a marketing term that will probably map to many
chips but Knights Corner (KNC) is the codename for this specific model.

The following patch adds perf_event support for the Xeon-Phi PMU,
as documented in the "Intel Xeon Phi Coprocessor (codename: Knights 
Corner) Performance Monitoring Units" manual.

Even though it is a co-processor, a Phi runs a full Linux environment
and can support performance counters.

This is just barebones support, it does not add support for
interesting new features such as the SPFLT intruction that
allows starting/stopping events without entering the kernel.

The PMU internally is just like that of an original Pentium, but 
a "P6-like" MSR interface is provided.  The interface is different enough
from a real P6 that it's not easy (or practical) to re-use the code in 
perf_event_p6.c

Acked-by: Lawrence F Meadows <lawrence.f.meadows@intel.com>
Acked-by: Cyrill Gorcunov <gorcunov@openvz.org>
Signed-off-by: Vince Weaver <vincent.weaver@maine.edu>

diff -Nur linux-3.6-rc6/arch/x86/include/asm/msr-index.h linux-3.6-rc6-mic/arch/x86/include/asm/msr-index.h
--- linux-3.6-rc6/arch/x86/include/asm/msr-index.h	2012-09-16 17:58:51.000000000 -0400
+++ linux-3.6-rc6-mic/arch/x86/include/asm/msr-index.h	2012-09-26 12:44:45.409898989 -0400
@@ -121,6 +121,11 @@
 #define MSR_P6_EVNTSEL0			0x00000186
 #define MSR_P6_EVNTSEL1			0x00000187
 
+#define MSR_KNC_PERFCTR0               0x00000020
+#define MSR_KNC_PERFCTR1               0x00000021
+#define MSR_KNC_EVNTSEL0               0x00000028
+#define MSR_KNC_EVNTSEL1               0x00000029
+
 /* AMD64 MSRs. Not complete. See the architecture manual for a more
    complete list. */
 
diff -Nur linux-3.6-rc6/arch/x86/kernel/cpu/Makefile linux-3.6-rc6-mic/arch/x86/kernel/cpu/Makefile
--- linux-3.6-rc6/arch/x86/kernel/cpu/Makefile	2012-09-16 17:58:51.000000000 -0400
+++ linux-3.6-rc6-mic/arch/x86/kernel/cpu/Makefile	2012-09-26 12:41:24.129899156 -0400
@@ -32,7 +32,7 @@
 
 ifdef CONFIG_PERF_EVENTS
 obj-$(CONFIG_CPU_SUP_AMD)		+= perf_event_amd.o
-obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_p6.o perf_event_p4.o
+obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_p6.o perf_event_knc.o perf_event_p4.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_intel_lbr.o perf_event_intel_ds.o perf_event_intel.o
 obj-$(CONFIG_CPU_SUP_INTEL)		+= perf_event_intel_uncore.o
 endif
diff -Nur linux-3.6-rc6/arch/x86/kernel/cpu/perfctr-watchdog.c linux-3.6-rc6-mic/arch/x86/kernel/cpu/perfctr-watchdog.c
--- linux-3.6-rc6/arch/x86/kernel/cpu/perfctr-watchdog.c	2012-09-16 17:58:51.000000000 -0400
+++ linux-3.6-rc6-mic/arch/x86/kernel/cpu/perfctr-watchdog.c	2012-09-26 12:45:52.109898934 -0400
@@ -56,6 +56,8 @@
 		switch (boot_cpu_data.x86) {
 		case 6:
 			return msr - MSR_P6_PERFCTR0;
+		case 11:
+			return msr - MSR_KNC_PERFCTR0;
 		case 15:
 			return msr - MSR_P4_BPU_PERFCTR0;
 		}
@@ -82,6 +84,8 @@
 		switch (boot_cpu_data.x86) {
 		case 6:
 			return msr - MSR_P6_EVNTSEL0;
+		case 11:
+			return msr - MSR_KNC_EVNTSEL0;
 		case 15:
 			return msr - MSR_P4_BSU_ESCR0;
 		}
diff -Nur linux-3.6-rc6/arch/x86/kernel/cpu/perf_event.h linux-3.6-rc6-mic/arch/x86/kernel/cpu/perf_event.h
--- linux-3.6-rc6/arch/x86/kernel/cpu/perf_event.h	2012-09-16 17:58:51.000000000 -0400
+++ linux-3.6-rc6-mic/arch/x86/kernel/cpu/perf_event.h	2012-09-26 12:41:43.585899140 -0400
@@ -624,6 +624,8 @@
 
 int p6_pmu_init(void);
 
+int knc_pmu_init(void);
+
 #else /* CONFIG_CPU_SUP_INTEL */
 
 static inline void reserve_ds_buffers(void)
diff -Nur linux-3.6-rc6/arch/x86/kernel/cpu/perf_event_intel.c linux-3.6-rc6-mic/arch/x86/kernel/cpu/perf_event_intel.c
--- linux-3.6-rc6/arch/x86/kernel/cpu/perf_event_intel.c	2012-09-16 17:58:51.000000000 -0400
+++ linux-3.6-rc6-mic/arch/x86/kernel/cpu/perf_event_intel.c	2012-09-26 12:41:54.161899131 -0400
@@ -1906,6 +1906,8 @@
 		switch (boot_cpu_data.x86) {
 		case 0x6:
 			return p6_pmu_init();
+		case 0xb:
+			return knc_pmu_init();
 		case 0xf:
 			return p4_pmu_init();
 		}
diff -Nur linux-3.6-rc6/arch/x86/kernel/cpu/perf_event_knc.c linux-3.6-rc6-mic/arch/x86/kernel/cpu/perf_event_knc.c
--- linux-3.6-rc6/arch/x86/kernel/cpu/perf_event_knc.c	1969-12-31 19:00:00.000000000 -0500
+++ linux-3.6-rc6-mic/arch/x86/kernel/cpu/perf_event_knc.c	2012-09-26 12:43:04.713899073 -0400
@@ -0,0 +1,248 @@
+/* Driver for Intel Xeon Phi "Knights Corner" PMU */
+
+#include <linux/perf_event.h>
+#include <linux/types.h>
+
+#include "perf_event.h"
+
+static const u64 knc_perfmon_event_map[] =
+{
+  [PERF_COUNT_HW_CPU_CYCLES]		= 0x002a,
+  [PERF_COUNT_HW_INSTRUCTIONS]		= 0x0016,
+  [PERF_COUNT_HW_CACHE_REFERENCES]	= 0x0028,
+  [PERF_COUNT_HW_CACHE_MISSES]		= 0x0029,
+  [PERF_COUNT_HW_BRANCH_INSTRUCTIONS]	= 0x0012,
+  [PERF_COUNT_HW_BRANCH_MISSES]		= 0x002b,
+};
+
+static __initconst u64 knc_hw_cache_event_ids
+				[PERF_COUNT_HW_CACHE_MAX]
+				[PERF_COUNT_HW_CACHE_OP_MAX]
+				[PERF_COUNT_HW_CACHE_RESULT_MAX] =
+{
+ [ C(L1D) ] = {
+	[ C(OP_READ) ] = {
+		/* On Xeon Phi event "0" is a valid DATA_READ          */
+		/*   (L1 Data Cache Reads) Instruction.                */
+		/* We code this as ARCH_PERFMON_EVENTSEL_INT as this   */
+		/* bit will always be set in x86_pmu_hw_config().      */
+		[ C(RESULT_ACCESS) ] = ARCH_PERFMON_EVENTSEL_INT,
+						/* DATA_READ           */
+		[ C(RESULT_MISS)   ] = 0x0003,	/* DATA_READ_MISS      */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0001,	/* DATA_WRITE          */
+		[ C(RESULT_MISS)   ] = 0x0004,	/* DATA_WRITE_MISS     */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0011,	/* L1_DATA_PF1         */
+		[ C(RESULT_MISS)   ] = 0x001c,	/* L1_DATA_PF1_MISS    */
+	},
+ },
+ [ C(L1I ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x000c,	/* CODE_READ          */
+		[ C(RESULT_MISS)   ] = 0x000e,	/* CODE_CACHE_MISS    */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0,
+		[ C(RESULT_MISS)   ] = 0x0,
+	},
+ },
+ [ C(LL  ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0,
+		[ C(RESULT_MISS)   ] = 0x10cb,	/* L2_READ_MISS */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x10cc,	/* L2_WRITE_HIT */
+		[ C(RESULT_MISS)   ] = 0,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x10fc,	/* L2_DATA_PF2      */
+		[ C(RESULT_MISS)   ] = 0x10fe,	/* L2_DATA_PF2_MISS */
+	},
+ },
+ [ C(DTLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = ARCH_PERFMON_EVENTSEL_INT,
+						/* DATA_READ */
+						/* see note on L1 OP_READ */
+		[ C(RESULT_MISS)   ] = 0x0002,	/* DATA_PAGE_WALK */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0001,	/* DATA_WRITE */
+		[ C(RESULT_MISS)   ] = 0x0002,	/* DATA_PAGE_WALK */
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0,
+		[ C(RESULT_MISS)   ] = 0x0,
+	},
+ },
+ [ C(ITLB) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x000c,	/* CODE_READ */
+		[ C(RESULT_MISS)   ] = 0x000d,	/* CODE_PAGE_WALK */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+ [ C(BPU ) ] = {
+	[ C(OP_READ) ] = {
+		[ C(RESULT_ACCESS) ] = 0x0012,	/* BRANCHES */
+		[ C(RESULT_MISS)   ] = 0x002b,	/* BRANCHES_MISPREDICTED */
+	},
+	[ C(OP_WRITE) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+	[ C(OP_PREFETCH) ] = {
+		[ C(RESULT_ACCESS) ] = -1,
+		[ C(RESULT_MISS)   ] = -1,
+	},
+ },
+};
+
+
+static u64 knc_pmu_event_map(int hw_event)
+{
+	return knc_perfmon_event_map[hw_event];
+}
+
+static struct event_constraint knc_event_constraints[] =
+{
+	INTEL_EVENT_CONSTRAINT(0xc3, 0x1),	/* HWP_L2HIT */
+	INTEL_EVENT_CONSTRAINT(0xc4, 0x1),	/* HWP_L2MISS */
+	INTEL_EVENT_CONSTRAINT(0xc8, 0x1),	/* L2_READ_HIT_E */
+	INTEL_EVENT_CONSTRAINT(0xc9, 0x1),	/* L2_READ_HIT_M */
+	INTEL_EVENT_CONSTRAINT(0xca, 0x1),	/* L2_READ_HIT_S */
+	INTEL_EVENT_CONSTRAINT(0xcb, 0x1),	/* L2_READ_MISS */
+	INTEL_EVENT_CONSTRAINT(0xcc, 0x1),	/* L2_WRITE_HIT */
+	INTEL_EVENT_CONSTRAINT(0xce, 0x1),	/* L2_STRONGLY_ORDERED_STREAMING_VSTORES_MISS */
+	INTEL_EVENT_CONSTRAINT(0xcf, 0x1),	/* L2_WEAKLY_ORDERED_STREAMING_VSTORE_MISS */
+	INTEL_EVENT_CONSTRAINT(0xd7, 0x1),	/* L2_VICTIM_REQ_WITH_DATA */
+	INTEL_EVENT_CONSTRAINT(0xe3, 0x1),	/* SNP_HITM_BUNIT */
+	INTEL_EVENT_CONSTRAINT(0xe6, 0x1),	/* SNP_HIT_L2 */
+	INTEL_EVENT_CONSTRAINT(0xe7, 0x1),	/* SNP_HITM_L2 */
+	INTEL_EVENT_CONSTRAINT(0xf1, 0x1),	/* L2_DATA_READ_MISS_CACHE_FILL */
+	INTEL_EVENT_CONSTRAINT(0xf2, 0x1),	/* L2_DATA_WRITE_MISS_CACHE_FILL */
+	INTEL_EVENT_CONSTRAINT(0xf6, 0x1),	/* L2_DATA_READ_MISS_MEM_FILL */
+	INTEL_EVENT_CONSTRAINT(0xf7, 0x1),	/* L2_DATA_WRITE_MISS_MEM_FILL */
+	INTEL_EVENT_CONSTRAINT(0xfc, 0x1),	/* L2_DATA_PF2 */
+	INTEL_EVENT_CONSTRAINT(0xfd, 0x1),	/* L2_DATA_PF2_DROP */
+	INTEL_EVENT_CONSTRAINT(0xfe, 0x1),	/* L2_DATA_PF2_MISS */
+	INTEL_EVENT_CONSTRAINT(0xff, 0x1),	/* L2_DATA_HIT_INFLIGHT_PF2 */
+	EVENT_CONSTRAINT_END
+};
+
+#define MSR_KNC_IA32_PERF_GLOBAL_STATUS		0x0000002d
+#define MSR_KNC_IA32_PERF_GLOBAL_OVF_CONTROL	0x0000002e
+#define MSR_KNC_IA32_PERF_GLOBAL_CTRL		0x0000002f
+
+#define KNC_ENABLE_COUNTER0			0x00000001
+#define KNC_ENABLE_COUNTER1			0x00000002
+
+static void knc_pmu_disable_all(void)
+{
+	u64 val;
+
+	rdmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val);
+	val &= ~(KNC_ENABLE_COUNTER0|KNC_ENABLE_COUNTER1);
+	wrmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val);
+}
+
+static void knc_pmu_enable_all(int added)
+{
+	u64 val;
+
+	rdmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val);
+	val |= (KNC_ENABLE_COUNTER0|KNC_ENABLE_COUNTER1);
+	wrmsrl(MSR_KNC_IA32_PERF_GLOBAL_CTRL, val);
+}
+
+static inline void
+knc_pmu_disable_event(struct perf_event *event)
+{
+	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
+	struct hw_perf_event *hwc = &event->hw;
+	u64 val;
+
+	val = hwc->config;
+	if (cpuc->enabled)
+		val &= ~ARCH_PERFMON_EVENTSEL_ENABLE;
+
+	(void)wrmsrl_safe(hwc->config_base + hwc->idx, val);
+}
+
+static void knc_pmu_enable_event(struct perf_event *event)
+{
+	struct cpu_hw_events *cpuc = &__get_cpu_var(cpu_hw_events);
+	struct hw_perf_event *hwc = &event->hw;
+	u64 val;
+
+	val = hwc->config;
+	if (cpuc->enabled)
+		val |= ARCH_PERFMON_EVENTSEL_ENABLE;
+
+	(void)wrmsrl_safe(hwc->config_base + hwc->idx, val);
+}
+
+PMU_FORMAT_ATTR(event,	"config:0-7"	);
+PMU_FORMAT_ATTR(umask,	"config:8-15"	);
+PMU_FORMAT_ATTR(edge,	"config:18"	);
+PMU_FORMAT_ATTR(inv,	"config:23"	);
+PMU_FORMAT_ATTR(cmask,	"config:24-31"	);
+
+static struct attribute *intel_knc_formats_attr[] = {
+	&format_attr_event.attr,
+	&format_attr_umask.attr,
+	&format_attr_edge.attr,
+	&format_attr_inv.attr,
+	&format_attr_cmask.attr,
+	NULL,
+};
+
+static __initconst struct x86_pmu knc_pmu = {
+	.name			= "knc",
+	.handle_irq		= x86_pmu_handle_irq,
+	.disable_all		= knc_pmu_disable_all,
+	.enable_all		= knc_pmu_enable_all,
+	.enable			= knc_pmu_enable_event,
+	.disable		= knc_pmu_disable_event,
+	.hw_config		= x86_pmu_hw_config,
+	.schedule_events	= x86_schedule_events,
+	.eventsel		= MSR_KNC_EVNTSEL0,
+	.perfctr		= MSR_KNC_PERFCTR0,
+	.event_map		= knc_pmu_event_map,
+	.max_events             = ARRAY_SIZE(knc_perfmon_event_map),
+	.apic			= 1,
+	.max_period		= (1ULL << 31) - 1,
+	.version		= 0,
+	.num_counters		= 2,
+	/* in theory 40 bits, early silicon is buggy though */
+	.cntval_bits		= 32,
+	.cntval_mask		= (1ULL << 32) - 1,
+	.get_event_constraints	= x86_get_event_constraints,
+	.event_constraints	= knc_event_constraints,
+	.format_attrs		= intel_knc_formats_attr,
+};
+
+__init int knc_pmu_init(void)
+{
+	x86_pmu = knc_pmu;
+
+	memcpy(hw_cache_event_ids, knc_hw_cache_event_ids, 
+		sizeof(hw_cache_event_ids));
+
+	return 0;
+}

^ permalink raw reply	[flat|nested] 2+ messages in thread
end of thread, other threads:[~2012-10-05  7:25 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2012-09-26 18:12 [PATCH V3 1/1] perf, Add support for Xeon-Phi Knights Corner PMU Vince Weaver
2012-10-05  7:24 ` [tip:perf/urgent] perf/x86: Add support for Intel " tip-bot for Vince Weaver

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.