[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <20100510152708.GS5104@lenovo>
Date: Mon, 10 May 2010 19:27:08 +0400
From: Cyrill Gorcunov <gorcunov@...il.com>
To: Lin Ming <ming.m.lin@...el.com>
Cc: LKML <linux-kernel@...r.kernel.org>,
Peter Zijlstra <peterz@...radead.org>,
Ingo Molnar <mingo@...e.hu>,
Steven Rostedt <rostedt@...dmis.org>,
Frederic Weisbecker <fweisbec@...il.com>
Subject: [rfc] x86,perf: P4 PMU -- use hash for p4_get_escr_idx
Hi Ming,
could you give this patch a shot if possible? Compile
tested only. I would appreciate review and complains
as well :) I know you're busy with other perf task
so there is no hurry. Just to share the patch as early
as possible.
Have CC'ed a number of people involved in P4 as well ;)
-- Cyrill
---
x86,perf: P4 PMU -- use hash for p4_get_escr_idx
Linear search over all p4 MSRs should be fine if only
we would not use it in events scheduling routine which
is pretty time crititcal. Lets use hashes. It should speed
scheduling up significantly.
CC: Peter Zijlstra <peterz@...radead.org>
CC: Ingo Molnar <mingo@...e.hu>
CC: Steven Rostedt <rostedt@...dmis.org>
CC: Frederic Weisbecker <fweisbec@...il.com>
CC: Lin Ming <ming.m.lin@...el.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@...nvz.org>
---
arch/x86/kernel/cpu/perf_event_p4.c | 123 +++++++++++++++++++-----------------
1 file changed, 67 insertions(+), 56 deletions(-)
Index: linux-2.6.git/arch/x86/kernel/cpu/perf_event_p4.c
=====================================================================
--- linux-2.6.git.orig/arch/x86/kernel/cpu/perf_event_p4.c
+++ linux-2.6.git/arch/x86/kernel/cpu/perf_event_p4.c
@@ -668,66 +668,77 @@ static void p4_pmu_swap_config_ts(struct
}
}
-/* ESCRs are not sequential in memory so we need a map */
-static const unsigned int p4_escr_map[ARCH_P4_TOTAL_ESCR] = {
- MSR_P4_ALF_ESCR0, /* 0 */
- MSR_P4_ALF_ESCR1, /* 1 */
- MSR_P4_BPU_ESCR0, /* 2 */
- MSR_P4_BPU_ESCR1, /* 3 */
- MSR_P4_BSU_ESCR0, /* 4 */
- MSR_P4_BSU_ESCR1, /* 5 */
- MSR_P4_CRU_ESCR0, /* 6 */
- MSR_P4_CRU_ESCR1, /* 7 */
- MSR_P4_CRU_ESCR2, /* 8 */
- MSR_P4_CRU_ESCR3, /* 9 */
- MSR_P4_CRU_ESCR4, /* 10 */
- MSR_P4_CRU_ESCR5, /* 11 */
- MSR_P4_DAC_ESCR0, /* 12 */
- MSR_P4_DAC_ESCR1, /* 13 */
- MSR_P4_FIRM_ESCR0, /* 14 */
- MSR_P4_FIRM_ESCR1, /* 15 */
- MSR_P4_FLAME_ESCR0, /* 16 */
- MSR_P4_FLAME_ESCR1, /* 17 */
- MSR_P4_FSB_ESCR0, /* 18 */
- MSR_P4_FSB_ESCR1, /* 19 */
- MSR_P4_IQ_ESCR0, /* 20 */
- MSR_P4_IQ_ESCR1, /* 21 */
- MSR_P4_IS_ESCR0, /* 22 */
- MSR_P4_IS_ESCR1, /* 23 */
- MSR_P4_ITLB_ESCR0, /* 24 */
- MSR_P4_ITLB_ESCR1, /* 25 */
- MSR_P4_IX_ESCR0, /* 26 */
- MSR_P4_IX_ESCR1, /* 27 */
- MSR_P4_MOB_ESCR0, /* 28 */
- MSR_P4_MOB_ESCR1, /* 29 */
- MSR_P4_MS_ESCR0, /* 30 */
- MSR_P4_MS_ESCR1, /* 31 */
- MSR_P4_PMH_ESCR0, /* 32 */
- MSR_P4_PMH_ESCR1, /* 33 */
- MSR_P4_RAT_ESCR0, /* 34 */
- MSR_P4_RAT_ESCR1, /* 35 */
- MSR_P4_SAAT_ESCR0, /* 36 */
- MSR_P4_SAAT_ESCR1, /* 37 */
- MSR_P4_SSU_ESCR0, /* 38 */
- MSR_P4_SSU_ESCR1, /* 39 */
- MSR_P4_TBPU_ESCR0, /* 40 */
- MSR_P4_TBPU_ESCR1, /* 41 */
- MSR_P4_TC_ESCR0, /* 42 */
- MSR_P4_TC_ESCR1, /* 43 */
- MSR_P4_U2L_ESCR0, /* 44 */
- MSR_P4_U2L_ESCR1, /* 45 */
+/*
+ * ESCR address hashing is tricky, ESCRs are not sequential
+ * in memory but all starts from MSR_P4_BSU_ESCR0 (0x03e0) and
+ * the metric between any ESCRs is laid in range [0xa0,0xe1]
+ *
+ * so we make ~70% filled hashtable
+ */
+
+#define P4_ESCR_MSR_BASE 0x000003a0
+#define P4_ESCR_MSR_MAX 0x000003e1
+#define P4_ESCR_MSR_TABLE_SIZE (P4_ESCR_MSR_MAX - P4_ESCR_MSR_BASE + 1)
+#define P4_ESCR_MSR_IDX(msr) (msr - P4_ESCR_MSR_BASE)
+#define P4_ESCR_MSR_TABLE_ENTRY(msr) [P4_ESCR_MSR_IDX(msr)] = msr
+
+static const unsigned int p4_escr_table[P4_ESCR_MSR_TABLE_SIZE] = {
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_ALF_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_ALF_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_BPU_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_BPU_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_BSU_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_BSU_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_CRU_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_CRU_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_CRU_ESCR2),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_CRU_ESCR3),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_CRU_ESCR4),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_CRU_ESCR5),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_DAC_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_DAC_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_FIRM_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_FIRM_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_FLAME_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_FLAME_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_FSB_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_FSB_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_IQ_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_IQ_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_IS_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_IS_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_ITLB_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_ITLB_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_IX_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_IX_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_MOB_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_MOB_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_MS_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_MS_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_PMH_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_PMH_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_RAT_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_RAT_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_SAAT_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_SAAT_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_SSU_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_SSU_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_TBPU_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_TBPU_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_TC_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_TC_ESCR1),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_U2L_ESCR0),
+ P4_ESCR_MSR_TABLE_ENTRY(MSR_P4_U2L_ESCR1),
};
static int p4_get_escr_idx(unsigned int addr)
{
- unsigned int i;
+ unsigned int idx = P4_ESCR_MSR_IDX(addr);
- for (i = 0; i < ARRAY_SIZE(p4_escr_map); i++) {
- if (addr == p4_escr_map[i])
- return i;
- }
+ BUG_ON(idx >= P4_ESCR_MSR_TABLE_SIZE);
+ BUG_ON(!p4_escr_table[idx]);
- return -1;
+ return idx;
}
static int p4_next_cntr(int thread, unsigned long *used_mask,
@@ -747,7 +758,7 @@ static int p4_next_cntr(int thread, unsi
static int p4_pmu_schedule_events(struct cpu_hw_events *cpuc, int n, int *assign)
{
unsigned long used_mask[BITS_TO_LONGS(X86_PMC_IDX_MAX)];
- unsigned long escr_mask[BITS_TO_LONGS(ARCH_P4_TOTAL_ESCR)];
+ unsigned long escr_mask[BITS_TO_LONGS(P4_ESCR_MSR_TABLE_SIZE)];
int cpu = raw_smp_processor_id();
struct hw_perf_event *hwc;
struct p4_event_bind *bind;
@@ -755,7 +766,7 @@ static int p4_pmu_schedule_events(struct
int cntr_idx, escr_idx;
bitmap_zero(used_mask, X86_PMC_IDX_MAX);
- bitmap_zero(escr_mask, ARCH_P4_TOTAL_ESCR);
+ bitmap_zero(escr_mask, P4_ESCR_MSR_TABLE_SIZE);
for (i = 0, num = n; i < n; i++, num--) {
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists