summaryrefslogtreecommitdiffstats
path: root/src/plugins/perfmon/perfmon_periodic.c
AgeCommit message (Collapse)AuthorFilesLines
2020-04-22misc: asan: do not poison memory after munmap()Benoît Ganne1-2/+0
It is a bad idea to poison memory after munmap because the address space can be reused (eg. for global data of dlopen()ed object) and ASan model allows access by default. Moreover, access to a stale address space will fault. Type: fix Change-Id: I356de422f255447d9d50a3a71fb0c2eaa790d731 Signed-off-by: Benoît Ganne <bganne@cisco.com>
2019-12-17perfmon: fix per-worker data initializationBenoît Ganne1-3/+14
When perfmon_init is called at initialization time worker threads are not created yet and vec_len(vlib_mains) returns 1. Initialize per-worker data when the number of workers is known, when enabling data collection instead. Type: fix Change-Id: I36887cc7b2a3e88d9728d3cd7262d9b1c968dd3c Signed-off-by: Benoît Ganne <bganne@cisco.com>
2019-05-24Add callback multiplex supportDave Barach1-15/+47
Change-Id: Iddeb3a1b0e20706e72ec8f74dabc60b342f003ba Signed-off-by: Dave Barach <dave@barachs.net>
2019-05-01Add node, frame to vlib main loop perf analysis callback argumentsDave Barach1-1/+3
Change-Id: Iaa5cd89791b0dfdb56a75009c564581d10696d83 Signed-off-by: Dave Barach <dave@barachs.net>
2019-03-26perfmon: fix pmc hw indices out-dated when multiple pmcSu Wang1-3/+15
When adding two or more events using a single "set pmc", the pmc hardware indices might be out-dated due to kernel reschdeduling the perf_event hardware counters. E.g. set pmc cpu-cycles cache-misses Solution: Open and enable all the events first, then aquire the indices from the kernel. Change-Id: I6913a871ab169e3b2855ac6159f527a1fca343e9 Signed-off-by: Su Wang <su.z.wang@ericsson.com>
2019-01-27Fix issue with cpu_id and numa_code captured too earlyDamjan Marion1-1/+1
Change-Id: I79b213b34c6071d14acf1922f89037a4a5a36c45 Signed-off-by: Damjan Marion <damarion@cisco.com>
2019-01-27perfmon: collect data on selected thread(s)Dave Barach1-15/+58
Add missing pre-input node runtime fork and refork code. unix-epoll-input runs on all threads; each instance needs its own runtime stats. Change-Id: I16b02e42d0c95f863161176c4bb9f9917bef809d Signed-off-by: Dave Barach <dave@barachs.net>
2019-01-24perfmon: enable pmc event before reading rdpmc indexDamjan Marion1-6/+6
Change-Id: I9b0a101e5d78c10257e3c5d8f5573c3eb29bfdef Signed-off-by: Damjan Marion <damarion@cisco.com>
2019-01-24perfmon plugin: 2-way parallel stat collectionDave Barach1-117/+178
As a FUD reduction measure, this patch implements 2-way parallel counter collection. Synthetic stat component counter pairs run at the same time. Running two counters (of any kind) at the same time naturally reduces the aggregate time required by an approximate factor-of-2, depending on whether an even or odd number of stats have been requested. I don't completely buy the argument that computing synthetic stats such as instructions-per-clock will be inaccurate if component counter values are collected sequentially. Given uniform traffic pattern, it must make no difference. As the collection interval increases, the difference between serial and parallel component counter collection will approach zero, see also the Central Limit theorem. Change-Id: I36ebdcf125e8882cca8a1929ec58f17fba1ad8f1 Signed-off-by: Dave Barach <dave@barachs.net>
2018-11-14Remove c-11 memcpy checks from perf-critical codeDave Barach1-1/+1
Change-Id: Id4f37f5d4a03160572954a416efa1ef9b3d79ad1 Signed-off-by: Dave Barach <dave@barachs.net>
2018-10-22X86_64 perf counter pluginDave Barach1-0/+433
Change-Id: Ie5a00c15ee9536cc61afab57f6cadc1aa1972f3c Signed-off-by: Dave Barach <dave@barachs.net>