perf/x86: Add memory profiling via PEBS Load Latency
authorStephane Eranian <eranian@google.com>
Thu, 24 Jan 2013 15:10:32 +0000 (16:10 +0100)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Mon, 1 Apr 2013 15:16:31 +0000 (12:16 -0300)
commitf20093eef5f7843a25adfc0512617d4b1ff1aa6e
tree1e1d008f98adab4477e3803ed24f3f2a22b34aaf
parentd6be9ad6c960f43800a6f118932bc8a5a4eadcd1
perf/x86: Add memory profiling via PEBS Load Latency

This patch adds support for memory profiling using the
PEBS Load Latency facility.

Load accesses are sampled by HW and the instruction
address, data address, load latency, data source, tlb,
locked information can be saved in the sampling buffer
if using the PERF_SAMPLE_COST (for latency),
PERF_SAMPLE_ADDR, PERF_SAMPLE_DATA_SRC types.

To enable PEBS Load Latency, users have to use the
model specific event:

 - on NHM/WSM: MEM_INST_RETIRED:LATENCY_ABOVE_THRESHOLD
 - on SNB/IVB: MEM_TRANS_RETIRED:LATENCY_ABOVE_THRESHOLD

To make things easier, this patch also exports a generic
alias via sysfs: mem-loads. It export the right event
encoding based on the host CPU and can be used directly
by the perf tool.

Loosely based on Intel's Lin Ming patch posted on LKML
in July 2011.

Signed-off-by: Stephane Eranian <eranian@google.com>
Cc: peterz@infradead.org
Cc: ak@linux.intel.com
Cc: acme@redhat.com
Cc: jolsa@redhat.com
Cc: namhyung.kim@lge.com
Link: http://lkml.kernel.org/r/1359040242-8269-9-git-send-email-eranian@google.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
arch/x86/include/uapi/asm/msr-index.h
arch/x86/kernel/cpu/perf_event.c
arch/x86/kernel/cpu/perf_event.h
arch/x86/kernel/cpu/perf_event_intel.c
arch/x86/kernel/cpu/perf_event_intel_ds.c