summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authormdr78 <mdr@ashroe.eu>2021-03-19 19:03:54 +0000
committerDamjan Marion <dmarion@me.com>2021-04-27 09:22:35 +0000
commit8e1384f7bf2e806e0d65d07af38da269ef1b8338 (patch)
tree4fd4720a58cfeec80550388d37c208b8befc02ac
parent3f923d2d465aadec8d70fa19e0559030751c2c6c (diff)
perfmon: top down level 1 support
Adding perfmon node TMAM support on ICX. Type: improvement Signed-off-by: Ray Kinsella <mdr@ashroe.eu> Change-Id: I48a9a9ff6a72efc28eaf0cb11ef39fb62cebb126
-rw-r--r--src/plugins/perfmon/CMakeLists.txt1
-rw-r--r--src/plugins/perfmon/cli.c7
-rw-r--r--src/plugins/perfmon/dispatch_wrapper.c89
-rw-r--r--src/plugins/perfmon/intel/bundle/topdown_metrics.c115
-rw-r--r--src/plugins/perfmon/intel/core.c12
-rw-r--r--src/plugins/perfmon/intel/core.h4
-rw-r--r--src/plugins/perfmon/perfmon.c14
-rw-r--r--src/plugins/perfmon/perfmon.h30
-rw-r--r--src/vppinfra/cpu.h1
9 files changed, 248 insertions, 25 deletions
diff --git a/src/plugins/perfmon/CMakeLists.txt b/src/plugins/perfmon/CMakeLists.txt
index c6113aa5097..af0bd3c3f51 100644
--- a/src/plugins/perfmon/CMakeLists.txt
+++ b/src/plugins/perfmon/CMakeLists.txt
@@ -30,4 +30,5 @@ add_vpp_plugin(perfmon
intel/bundle/cache_hit_miss.c
intel/bundle/branch_mispred.c
intel/bundle/power_license.c
+ intel/bundle/topdown_metrics.c
)
diff --git a/src/plugins/perfmon/cli.c b/src/plugins/perfmon/cli.c
index 39e0319603a..f6678772d84 100644
--- a/src/plugins/perfmon/cli.c
+++ b/src/plugins/perfmon/cli.c
@@ -128,7 +128,8 @@ show_perfmon_bundle_command_fn (vlib_main_t *vm, unformat_input_t *input,
vlib_cli_output (vm, "%U\n", format_perfmon_bundle, 0, 0);
for (int i = 0; i < vec_len (vb); i++)
- vlib_cli_output (vm, "%U\n", format_perfmon_bundle, vb[i], verbose);
+ if (!vb[i]->cpu_supports || vb[i]->cpu_supports ())
+ vlib_cli_output (vm, "%U\n", format_perfmon_bundle, vb[i], verbose);
vec_free (vb);
return 0;
@@ -290,7 +291,8 @@ show_perfmon_stats_command_fn (vlib_main_t *vm, unformat_input_t *input,
n_instances = vec_len (it->instances);
vec_validate (readings, n_instances - 1);
- for (int i = 0; i < n_instances; i++)
+ /*Only perform read() for THREAD or SYSTEM bundles*/
+ for (int i = 0; i < n_instances && b->type != PERFMON_BUNDLE_TYPE_NODE; i++)
{
in = vec_elt_at_index (it->instances, i);
r = vec_elt_at_index (readings, i);
@@ -340,6 +342,7 @@ show_perfmon_stats_command_fn (vlib_main_t *vm, unformat_input_t *input,
table_set_cell_align (t, col, -1, TTAA_RIGHT);
table_set_cell_fg_color (t, col, -1, TTAC_CYAN);
clib_memcpy_fast (&ns, tr->node_stats + j, sizeof (ns));
+
for (int j = 0; j < n_row; j++)
table_format_cell (t, col, j, "%U", b->format_fn, &ns, j);
}
diff --git a/src/plugins/perfmon/dispatch_wrapper.c b/src/plugins/perfmon/dispatch_wrapper.c
index 4ae9c7707ef..fe0a449df99 100644
--- a/src/plugins/perfmon/dispatch_wrapper.c
+++ b/src/plugins/perfmon/dispatch_wrapper.c
@@ -13,6 +13,7 @@
* limitations under the License.
*/
+#include "vppinfra/string.h"
#include <vnet/vnet.h>
#include <vlibapi/api.h>
@@ -49,24 +50,32 @@ perfmon_read_pmcs (u64 *counters, int *pmc_index, u8 n_counters)
}
static_always_inline int
-perfmon_calc_pmc_index (perfmon_thread_runtime_t *tr, u8 i)
+perfmon_calc_mmap_offset (perfmon_thread_runtime_t *tr, u8 i)
{
return (int) (tr->mmap_pages[i]->index + tr->mmap_pages[i]->offset);
}
+static_always_inline int
+perfmon_metric_index (perfmon_bundle_t *b, u8 i)
+{
+ return (int) (b->metrics[i]);
+}
+
uword
-perfmon_dispatch_wrapper (vlib_main_t *vm, vlib_node_runtime_t *node,
- vlib_frame_t *frame)
+perfmon_dispatch_wrapper_mmap (vlib_main_t *vm, vlib_node_runtime_t *node,
+ vlib_frame_t *frame)
{
perfmon_main_t *pm = &perfmon_main;
perfmon_thread_runtime_t *rt =
vec_elt_at_index (pm->thread_runtimes, vm->thread_index);
perfmon_node_stats_t *s =
vec_elt_at_index (rt->node_stats, node->node_index);
+
u8 n_events = rt->n_events;
- int pmc_index[PERF_MAX_EVENTS];
+
u64 before[PERF_MAX_EVENTS];
u64 after[PERF_MAX_EVENTS];
+ int pmc_index[PERF_MAX_EVENTS];
uword rv;
clib_prefetch_load (s);
@@ -75,33 +84,87 @@ perfmon_dispatch_wrapper (vlib_main_t *vm, vlib_node_runtime_t *node,
{
default:
case 7:
- pmc_index[6] = perfmon_calc_pmc_index (rt, 6);
+ pmc_index[6] = perfmon_calc_mmap_offset (rt, 6);
case 6:
- pmc_index[5] = perfmon_calc_pmc_index (rt, 5);
+ pmc_index[5] = perfmon_calc_mmap_offset (rt, 5);
case 5:
- pmc_index[4] = perfmon_calc_pmc_index (rt, 4);
+ pmc_index[4] = perfmon_calc_mmap_offset (rt, 4);
case 4:
- pmc_index[3] = perfmon_calc_pmc_index (rt, 3);
+ pmc_index[3] = perfmon_calc_mmap_offset (rt, 3);
case 3:
- pmc_index[2] = perfmon_calc_pmc_index (rt, 2);
+ pmc_index[2] = perfmon_calc_mmap_offset (rt, 2);
case 2:
- pmc_index[1] = perfmon_calc_pmc_index (rt, 1);
+ pmc_index[1] = perfmon_calc_mmap_offset (rt, 1);
case 1:
- pmc_index[0] = perfmon_calc_pmc_index (rt, 0);
+ pmc_index[0] = perfmon_calc_mmap_offset (rt, 0);
break;
}
- perfmon_read_pmcs (before, pmc_index, n_events);
+ perfmon_read_pmcs (&before[0], pmc_index, n_events);
rv = node->function (vm, node, frame);
- perfmon_read_pmcs (after, pmc_index, n_events);
+ perfmon_read_pmcs (&after[0], pmc_index, n_events);
if (rv == 0)
return rv;
s->n_calls += 1;
s->n_packets += rv;
+
for (int i = 0; i < n_events; i++)
s->value[i] += after[i] - before[i];
return rv;
}
+
+uword
+perfmon_dispatch_wrapper_metrics (vlib_main_t *vm, vlib_node_runtime_t *node,
+ vlib_frame_t *frame)
+{
+ perfmon_main_t *pm = &perfmon_main;
+ perfmon_thread_runtime_t *rt =
+ vec_elt_at_index (pm->thread_runtimes, vm->thread_index);
+ perfmon_node_stats_t *s =
+ vec_elt_at_index (rt->node_stats, node->node_index);
+
+ u8 n_events = rt->n_events;
+
+ u64 before[PERF_MAX_EVENTS];
+ int pmc_index[PERF_MAX_EVENTS];
+ uword rv;
+
+ clib_prefetch_load (s);
+
+ switch (n_events)
+ {
+ default:
+ case 7:
+ pmc_index[6] = perfmon_metric_index (rt->bundle, 6);
+ case 6:
+ pmc_index[5] = perfmon_metric_index (rt->bundle, 5);
+ case 5:
+ pmc_index[4] = perfmon_metric_index (rt->bundle, 4);
+ case 4:
+ pmc_index[3] = perfmon_metric_index (rt->bundle, 3);
+ case 3:
+ pmc_index[2] = perfmon_metric_index (rt->bundle, 2);
+ case 2:
+ pmc_index[1] = perfmon_metric_index (rt->bundle, 1);
+ case 1:
+ pmc_index[0] = perfmon_metric_index (rt->bundle, 0);
+ break;
+ }
+
+ perfmon_read_pmcs (&before[0], pmc_index, n_events);
+ rv = node->function (vm, node, frame);
+
+ clib_memcpy_fast (&s->t[0].value[0], &before, sizeof (before));
+ perfmon_read_pmcs (&s->t[1].value[0], pmc_index, n_events);
+
+ if (rv == 0)
+ return rv;
+
+ s->n_calls += 1;
+ s->n_packets += rv;
+
+ return rv;
+}
diff --git a/src/plugins/perfmon/intel/bundle/topdown_metrics.c b/src/plugins/perfmon/intel/bundle/topdown_metrics.c
new file mode 100644
index 00000000000..386f3843bc3
--- /dev/null
+++ b/src/plugins/perfmon/intel/bundle/topdown_metrics.c
@@ -0,0 +1,115 @@
+/*
+ * Copyright (c) 2021 Intel and/or its affiliates.
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at:
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <vnet/vnet.h>
+#include <perfmon/perfmon.h>
+#include <perfmon/intel/core.h>
+
+#define GET_METRIC(m, i) (((m) >> (i * 8)) & 0xff)
+#define GET_RATIO(m, i) (((m) >> (i * 32)) & 0xffffffff)
+#define RDPMC_FIXED_SLOTS (1 << 30) /* fixed slots */
+#define RDPMC_L1_METRICS (1 << 29) /* l1 metric counters */
+
+#define FIXED_COUNTER_SLOTS 3
+#define METRIC_COUNTER_TOPDOWN_L1 0
+
+typedef enum
+{
+ TOPDOWN_E_METRIC_RETIRING = 0,
+ TOPDOWN_E_METRIC_BAD_SPEC,
+ TOPDOWN_E_METRIC_FE_BOUND,
+ TOPDOWN_E_METRIC_BE_BOUND,
+} topdown_lvl1_counters_t;
+
+enum
+{
+ TOPDOWN_SLOTS = 0,
+ TOPDOWN_METRICS,
+} topdown_lvl1_metrics_t;
+
+static_always_inline f32
+topdown_lvl1_parse_row (perfmon_node_stats_t *ns, topdown_lvl1_counters_t e)
+{
+ f64 slots_t0 =
+ ns->t[0].value[TOPDOWN_SLOTS] *
+ ((f64) GET_METRIC (ns->t[0].value[TOPDOWN_METRICS], e) / 0xff);
+ f64 slots_t1 =
+ ns->t[1].value[TOPDOWN_SLOTS] *
+ ((f64) GET_METRIC (ns->t[1].value[TOPDOWN_METRICS], e) / 0xff);
+ u64 slots_delta =
+ ns->t[1].value[TOPDOWN_SLOTS] - ns->t[0].value[TOPDOWN_SLOTS];
+
+ slots_t1 = slots_t1 - slots_t0;
+
+ return (slots_t1 / slots_delta) * 100;
+}
+
+static u8 *
+format_topdown_lvl1 (u8 *s, va_list *args)
+{
+ perfmon_node_stats_t *st = va_arg (*args, perfmon_node_stats_t *);
+ u64 row = va_arg (*args, int);
+
+ switch (row)
+ {
+ case 0:
+ s = format (s, "%f",
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_BAD_SPEC) +
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_RETIRING));
+ break;
+ case 1:
+ s = format (s, "%f",
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_BE_BOUND) +
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_FE_BOUND));
+ break;
+ case 2:
+ s = format (s, "%f",
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_RETIRING));
+ break;
+ case 3:
+ s = format (s, "%f",
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_BAD_SPEC));
+ break;
+ case 4:
+ s = format (s, "%f",
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_FE_BOUND));
+ break;
+ case 5:
+ s = format (s, "%f",
+ topdown_lvl1_parse_row (st, TOPDOWN_E_METRIC_BE_BOUND));
+ break;
+ }
+ return s;
+}
+
+PERFMON_REGISTER_BUNDLE (topdown_lvl1) = {
+ .name = "topdown-level1",
+ .description = "Top-down Microarchitecture Analysis Level 1",
+ .source = "intel-core",
+ .type = PERFMON_BUNDLE_TYPE_NODE,
+ .offset_type = PERFMON_OFFSET_TYPE_METRICS,
+ .events[0] = INTEL_CORE_E_TOPDOWN_SLOTS,
+ .events[1] = INTEL_CORE_E_TOPDOWN_L1_METRICS,
+ .metrics[0] = RDPMC_FIXED_SLOTS | FIXED_COUNTER_SLOTS,
+ .metrics[1] = RDPMC_L1_METRICS | METRIC_COUNTER_TOPDOWN_L1,
+ .n_events = 2,
+ .cpu_supports = clib_cpu_supports_avx512_bitalg,
+ .format_fn = format_topdown_lvl1,
+ .column_headers = PERFMON_STRINGS ("% NS", "% ST", "% NS.RT", "% NS.BS",
+ "% ST.FE", "% ST.BE"),
+ .footer = "Not Stalled (NS),STalled (ST),\n"
+ " Retiring (RT), Bad Speculation (BS),\n"
+ " FrontEnd bound (FE), BackEnd bound (BE)",
+};
diff --git a/src/plugins/perfmon/intel/core.c b/src/plugins/perfmon/intel/core.c
index b4267e8170c..cef6f32d7e6 100644
--- a/src/plugins/perfmon/intel/core.c
+++ b/src/plugins/perfmon/intel/core.c
@@ -20,12 +20,12 @@
static perfmon_event_t events[] = {
#define _(event, umask, edge, any, inv, cmask, n, suffix, desc) \
- [INTEL_CORE_E_##n##_##suffix] = { \
- .type = PERF_TYPE_RAW, \
- .config = PERF_INTEL_CODE (event, umask, edge, any, inv, cmask), \
- .name = #n "." #suffix, \
- .description = desc, \
- },
+ [INTEL_CORE_E_##n##_##suffix] = { .type = PERF_TYPE_RAW, \
+ .config = PERF_INTEL_CODE ( \
+ event, umask, edge, any, inv, cmask), \
+ .name = #n "." #suffix, \
+ .description = desc, \
+ .exclude_kernel = 1 },
foreach_perf_intel_core_event
#undef _
diff --git a/src/plugins/perfmon/intel/core.h b/src/plugins/perfmon/intel/core.h
index 01945d2c023..cd5c31ba260 100644
--- a/src/plugins/perfmon/intel/core.h
+++ b/src/plugins/perfmon/intel/core.h
@@ -27,6 +27,10 @@
"Core cycles when the thread is not in halt state") \
_ (0x00, 0x03, 0, 0, 0, 0x00, CPU_CLK_UNHALTED, REF_TSC, \
"Reference cycles when the core is not in halt state.") \
+ _ (0x00, 0x04, 0, 0, 0, 0x00, TOPDOWN, SLOTS, \
+ "TMA slots available for an unhalted logical processor.") \
+ _ (0x00, 0x80, 0, 0, 0, 0x00, TOPDOWN, L1_METRICS, \
+ "TMA slots metrics for an unhalted logical processor.") \
_ (0x03, 0x02, 0, 0, 0, 0x00, LD_BLOCKS, STORE_FORWARD, \
"Loads blocked due to overlapping with a preceding store that cannot be" \
" forwarded.") \
diff --git a/src/plugins/perfmon/perfmon.c b/src/plugins/perfmon/perfmon.c
index f9402f8f8e9..46c8cf9ca04 100644
--- a/src/plugins/perfmon/perfmon.c
+++ b/src/plugins/perfmon/perfmon.c
@@ -193,6 +193,7 @@ perfmon_set (vlib_main_t *vm, perfmon_bundle_t *b)
{
perfmon_thread_runtime_t *rt;
rt = vec_elt_at_index (pm->thread_runtimes, i);
+ rt->bundle = b;
rt->n_events = b->n_events;
rt->n_nodes = n_nodes;
vec_validate_aligned (rt->node_stats, n_nodes - 1,
@@ -235,11 +236,20 @@ perfmon_start (vlib_main_t *vm, perfmon_bundle_t *b)
return clib_error_return_unix (0, "ioctl(PERF_EVENT_IOC_ENABLE)");
}
}
- if (pm->active_bundle->type == PERFMON_BUNDLE_TYPE_NODE)
+ if (b->type == PERFMON_BUNDLE_TYPE_NODE)
{
+
+ vlib_node_function_t *funcs[PERFMON_OFFSET_TYPE_MAX];
+#define _(type, pfunc) funcs[type] = pfunc;
+
+ foreach_permon_offset_type
+#undef _
+
+ ASSERT (funcs[b->offset_type]);
+
for (int i = 0; i < vlib_get_n_threads (); i++)
vlib_node_set_dispatch_wrapper (vlib_get_main_by_index (i),
- perfmon_dispatch_wrapper);
+ funcs[b->offset_type]);
}
pm->sample_time = vlib_time_now (vm);
diff --git a/src/plugins/perfmon/perfmon.h b/src/plugins/perfmon/perfmon.h
index f24a23bd974..3d0bc0608f9 100644
--- a/src/plugins/perfmon/perfmon.h
+++ b/src/plugins/perfmon/perfmon.h
@@ -20,6 +20,7 @@
#include <vppinfra/clib.h>
#include <vppinfra/format.h>
#include <vppinfra/error.h>
+#include <vppinfra/cpu.h>
#include <vlib/vlib.h>
#define PERF_MAX_EVENTS 7 /* 3 fixed and 4 programmable */
@@ -32,6 +33,13 @@ typedef enum
PERFMON_BUNDLE_TYPE_SYSTEM,
} perfmon_bundle_type_t;
+typedef enum
+{
+ PERFMON_OFFSET_TYPE_MMAP,
+ PERFMON_OFFSET_TYPE_METRICS,
+ PERFMON_OFFSET_TYPE_MAX,
+} perfmon_offset_type_t;
+
typedef struct
{
u32 type_from_instance : 1;
@@ -61,7 +69,12 @@ typedef struct
} perfmon_instance_type_t;
struct perfmon_source;
-vlib_node_function_t perfmon_dispatch_wrapper;
+vlib_node_function_t perfmon_dispatch_wrapper_mmap;
+vlib_node_function_t perfmon_dispatch_wrapper_metrics;
+
+#define foreach_permon_offset_type \
+ _ (PERFMON_OFFSET_TYPE_MMAP, perfmon_dispatch_wrapper_mmap) \
+ _ (PERFMON_OFFSET_TYPE_METRICS, perfmon_dispatch_wrapper_metrics)
typedef clib_error_t *(perfmon_source_init_fn_t) (vlib_main_t *vm,
struct perfmon_source *);
@@ -78,8 +91,10 @@ typedef struct perfmon_source
} perfmon_source_t;
struct perfmon_bundle;
+
typedef clib_error_t *(perfmon_bundle_init_fn_t) (vlib_main_t *vm,
struct perfmon_bundle *);
+
typedef struct perfmon_bundle
{
char *name;
@@ -87,7 +102,9 @@ typedef struct perfmon_bundle
char *source;
char *footer;
perfmon_bundle_type_t type;
+ perfmon_offset_type_t offset_type;
u32 events[PERF_MAX_EVENTS];
+ u32 metrics[PERF_MAX_EVENTS];
u32 n_events;
perfmon_bundle_init_fn_t *init_fn;
@@ -95,6 +112,7 @@ typedef struct perfmon_bundle
char **column_headers;
char **raw_column_headers;
format_function_t *format_fn;
+ clib_cpu_supports_func_t cpu_supports;
/* do not set manually */
perfmon_source_t *src;
@@ -114,7 +132,14 @@ typedef struct
CLIB_CACHE_LINE_ALIGN_MARK (cacheline0);
u64 n_calls;
u64 n_packets;
- u64 value[PERF_MAX_EVENTS];
+ union
+ {
+ struct
+ {
+ u64 value[PERF_MAX_EVENTS];
+ } t[2];
+ u64 value[PERF_MAX_EVENTS * 2];
+ };
} perfmon_node_stats_t;
typedef struct
@@ -122,6 +147,7 @@ typedef struct
u8 n_events;
u16 n_nodes;
perfmon_node_stats_t *node_stats;
+ perfmon_bundle_t *bundle;
struct perf_event_mmap_page *mmap_pages[PERF_MAX_EVENTS];
} perfmon_thread_runtime_t;
diff --git a/src/vppinfra/cpu.h b/src/vppinfra/cpu.h
index 6925d584e52..8d2465ba4d6 100644
--- a/src/vppinfra/cpu.h
+++ b/src/vppinfra/cpu.h
@@ -185,6 +185,7 @@ clib_get_cpuid (const u32 lev, u32 * eax, u32 * ebx, u32 * ecx, u32 * edx)
return 1;
}
+typedef int (*clib_cpu_supports_func_t) ();
#define _(flag, func, reg, bit) \
static inline int \