1 Using TopDown metrics in user space
2 -----------------------------------
4 Intel CPUs (since Sandy Bridge and Silvermont) support a TopDown
5 methology to break down CPU pipeline execution into 4 bottlenecks:
6 frontend bound, backend bound, bad speculation, retiring.
8 For more details on Topdown see [1][5]
10 Traditionally this was implemented by events in generic counters
11 and specific formulas to compute the bottlenecks.
13 perf stat --topdown implements this.
15 Full Top Down includes more levels that can break down the
16 bottlenecks further. This is not directly implemented in perf,
17 but available in other tools that can run on top of perf,
18 such as toplev[2] or vtune[3]
20 New Topdown features in Ice Lake
21 ===============================
23 With Ice Lake CPUs the TopDown metrics are directly available as
24 fixed counters and do not require generic counters. This allows
25 to collect TopDown always in addition to other events.
27 % perf stat -a --topdown -I1000
28 # time retiring bad speculation frontend bound backend bound
29 1.001281330 23.0% 15.3% 29.6% 32.1%
30 2.003009005 5.0% 6.8% 46.6% 41.6%
31 3.004646182 6.7% 6.7% 46.0% 40.6%
32 4.006326375 5.0% 6.4% 47.6% 41.0%
33 5.007991804 5.1% 6.3% 46.3% 42.3%
34 6.009626773 6.2% 7.1% 47.3% 39.3%
35 7.011296356 4.7% 6.7% 46.2% 42.4%
36 8.012951831 4.7% 6.7% 47.5% 41.1%
39 This also enables measuring TopDown per thread/process instead
42 Using TopDown through RDPMC in applications on Ice Lake
43 ======================================================
45 For more fine grained measurements it can be useful to
46 access the new directly from user space. This is more complicated,
47 but drastically lowers overhead.
49 On Ice Lake, there is a new fixed counter 3: SLOTS, which reports
50 "pipeline SLOTS" (cycles multiplied by core issue width) and a
51 metric register that reports slots ratios for the different bottleneck
54 The metrics counter is CPU model specific and is not available on older
60 Library functions to do the functionality described below
61 is also available in libjevents [4]
63 The application opens a group with fixed counter 3 (SLOTS) and any
64 metric event, and allow user programs to read the performance counters.
66 Fixed counter 3 is mapped to a pseudo event event=0x00, umask=04,
67 so the perf_event_attr structure should be initialized with
68 { .config = 0x0400, .type = PERF_TYPE_RAW }
69 The metric events are mapped to the pseudo event event=0x00, umask=0x8X.
70 For example, the perf_event_attr structure can be initialized with
71 { .config = 0x8000, .type = PERF_TYPE_RAW } for Retiring metric event
72 The Fixed counter 3 must be the leader of the group.
74 #include <linux/perf_event.h>
75 #include <sys/syscall.h>
78 /* Provide own perf_event_open stub because glibc doesn't */
80 int perf_event_open(struct perf_event_attr *attr, pid_t pid,
81 int cpu, int group_fd, unsigned long flags)
83 return syscall(__NR_perf_event_open, attr, pid, cpu, group_fd, flags);
86 /* Open slots counter file descriptor for current task. */
87 struct perf_event_attr slots = {
88 .type = PERF_TYPE_RAW,
89 .size = sizeof(struct perf_event_attr),
94 int slots_fd = perf_event_open(&slots, 0, -1, -1, 0);
99 * Open metrics event file descriptor for current task.
100 * Set slots event as the leader of the group.
102 struct perf_event_attr metrics = {
103 .type = PERF_TYPE_RAW,
104 .size = sizeof(struct perf_event_attr),
109 int metrics_fd = perf_event_open(&metrics, 0, -1, slots_fd, 0);
114 The RDPMC instruction (or _rdpmc compiler intrinsic) can now be used
115 to read slots and the topdown metrics at different points of the program:
118 #include <x86intrin.h>
120 #define RDPMC_FIXED (1 << 30) /* return fixed counters */
121 #define RDPMC_METRIC (1 << 29) /* return metric counters */
123 #define FIXED_COUNTER_SLOTS 3
124 #define METRIC_COUNTER_TOPDOWN_L1 0
126 static inline uint64_t read_slots(void)
128 return _rdpmc(RDPMC_FIXED | FIXED_COUNTER_SLOTS);
131 static inline uint64_t read_metrics(void)
133 return _rdpmc(RDPMC_METRIC | METRIC_COUNTER_TOPDOWN_L1);
136 Then the program can be instrumented to read these metrics at different
139 It's not a good idea to do this with too short code regions,
140 as the parallelism and overlap in the CPU program execution will
141 cause too much measurement inaccuracy. For example instrumenting
142 individual basic blocks is definitely too fine grained.
144 Decoding metrics values
145 =======================
147 The value reported by read_metrics() contains four 8 bit fields
148 that represent a scaled ratio that represent the Level 1 bottleneck.
149 All four fields add up to 0xff (= 100%)
151 The binary ratios in the metric value can be converted to float ratios:
153 #define GET_METRIC(m, i) (((m) >> (i*8)) & 0xff)
155 #define TOPDOWN_RETIRING(val) ((float)GET_METRIC(val, 0) / 0xff)
156 #define TOPDOWN_BAD_SPEC(val) ((float)GET_METRIC(val, 1) / 0xff)
157 #define TOPDOWN_FE_BOUND(val) ((float)GET_METRIC(val, 2) / 0xff)
158 #define TOPDOWN_BE_BOUND(val) ((float)GET_METRIC(val, 3) / 0xff)
160 and then converted to percent for printing.
162 The ratios in the metric accumulate for the time when the counter
163 is enabled. For measuring programs it is often useful to measure
164 specific sections. For this it is needed to deltas on metrics.
166 This can be done by scaling the metrics with the slots counter
167 read at the same time.
169 Then it's possible to take deltas of these slots counts
170 measured at different points, and determine the metrics
171 for that time period.
173 slots_a = read_slots();
174 metric_a = read_metrics();
176 ... larger code region ...
178 slots_b = read_slots()
179 metric_b = read_metrics()
181 # compute scaled metrics for measurement a
182 retiring_slots_a = GET_METRIC(metric_a, 0) * slots_a
183 bad_spec_slots_a = GET_METRIC(metric_a, 1) * slots_a
184 fe_bound_slots_a = GET_METRIC(metric_a, 2) * slots_a
185 be_bound_slots_a = GET_METRIC(metric_a, 3) * slots_a
187 # compute delta scaled metrics between b and a
188 retiring_slots = GET_METRIC(metric_b, 0) * slots_b - retiring_slots_a
189 bad_spec_slots = GET_METRIC(metric_b, 1) * slots_b - bad_spec_slots_a
190 fe_bound_slots = GET_METRIC(metric_b, 2) * slots_b - fe_bound_slots_a
191 be_bound_slots = GET_METRIC(metric_b, 3) * slots_b - be_bound_slots_a
193 Later the individual ratios for the measurement period can be recreated
196 slots_delta = slots_b - slots_a
197 retiring_ratio = (float)retiring_slots / slots_delta
198 bad_spec_ratio = (float)bad_spec_slots / slots_delta
199 fe_bound_ratio = (float)fe_bound_slots / slots_delta
200 be_bound_ratio = (float)be_bound_slots / slota_delta
202 printf("Retiring %.2f%% Bad Speculation %.2f%% FE Bound %.2f%% BE Bound %.2f%%\n",
203 retiring_ratio * 100.,
204 bad_spec_ratio * 100.,
205 fe_bound_ratio * 100.,
206 be_bound_ratio * 100.);
208 Resetting metrics counters
209 ==========================
211 Since the individual metrics are only 8bit they lose precision for
212 short regions over time because the number of cycles covered by each
213 fraction bit shrinks. So the counters need to be reset regularly.
215 When using the kernel perf API the kernel resets on every read.
216 So as long as the reading is at reasonable intervals (every few
217 seconds) the precision is good.
219 When using perf stat it is recommended to always use the -I option,
220 with no longer interval than a few seconds
222 perf stat -I 1000 --topdown ...
224 For user programs using RDPMC directly the counter can
225 be reset explicitly using ioctl:
227 ioctl(perf_fd, PERF_EVENT_IOC_RESET, 0);
229 This "opens" a new measurement period.
231 A program using RDPMC for TopDown should schedule such a reset
232 regularly, as in every few seconds.
237 Four pseudo TopDown metric events are exposed for the end-users,
238 topdown-retiring, topdown-bad-spec, topdown-fe-bound and topdown-be-bound.
239 They can be used to collect the TopDown value under the following
241 - All the TopDown metric events must be in a group with the SLOTS event.
242 - The SLOTS event must be the leader of the group.
243 - The PERF_FORMAT_GROUP flag must be applied for each TopDown metric
246 The SLOTS event and the TopDown metric events can be counting members of
247 a sampling read group. Since the SLOTS event must be the leader of a TopDown
248 group, the second event of the group is the sampling event.
249 For example, perf record -e '{slots, $sampling_event, topdown-retiring}:S'
252 [1] https://software.intel.com/en-us/top-down-microarchitecture-analysis-method-win
253 [2] https://github.com/andikleen/pmu-tools/wiki/toplev-manual
254 [3] https://software.intel.com/en-us/intel-vtune-amplifier-xe
255 [4] https://github.com/andikleen/pmu-tools/tree/master/jevents
256 [5] https://sites.google.com/site/analysismethods/yasin-pubs