[GIT]: libpfm4 branch, master, updated. 01c6dcedc9330442bac2834ddcb7ea6bfc525e9d
Status: Beta
Brought to you by:
seranian
From: seranian <ser...@us...> - 2010-11-19 15:17:43
|
This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "libpfm4". The branch, master has been updated via 01c6dcedc9330442bac2834ddcb7ea6bfc525e9d (commit) from d7520a7a6564272efd3be79c045fe411c288ed71 (commit) Those revisions listed above that are new to this repository have not appeared on any other notification email; so we list those revisions in full, below. - Log ----------------------------------------------------------------- commit 01c6dcedc9330442bac2834ddcb7ea6bfc525e9d Author: Stephane Eranian <er...@gm...> Date: Fri Nov 19 16:16:57 2010 +0100 add Intel Nehalem/Westmere OFFCORE_RESPONSE event support The OFFCORE_RESPONSE is a special event available on Intel Nehalem and Westmere based processor. It is documented in Intel Vol3b Section 30.6.1.3. The event requires an extra MSR to be programmed in order to count properly. The extra MSR contains the actual umasks for the event. OFFCORE_RESPONSE is a very useful event for people interested in measuring the memory subsystem. Nehalem offers once instance of the event: OFFCORE_RESPONSE_0. Westmere offers two instances which can be measured concurrently: OFFCORE_RESPONSE_0 and OFFCORE_RESPONSE_1. The umasks are divided in two groups: request, response. The request filters the type of memory access you want to measure, whereas the response filters the type of location where the cacheline is serviced. User must provide at least on umask in each group. The library checks for this constraint. The support for the event is provided for both the RAW, i.e., pure HW mode, and also for perf_events, thru the dedicated entry point: pfm_get_perf_event_encoding(). The support for OFFCORE_RESPONSE in perf_events should appear in 2.6.37. A patch was recently posted on LKML by Andi Kleen. The man pages for Intel Nehalem and Westmere have been updated accordingly. Signed-off-by: Stephane Eranian <er...@gm...> ----------------------------------------------------------------------- Summary of changes: docs/man3/libpfm_intel_nhm.3 | 17 +++ docs/man3/libpfm_intel_wsm.3 | 18 +++ lib/events/intel_nhm_events.h | 157 +++++++++++----------------- lib/events/intel_wsm_events.h | 229 ++++++++++++++++++++++++++++++++++++++++- lib/pfmlib_intel_x86.c | 61 +++++++----- lib/pfmlib_intel_x86_priv.h | 1 + lib/pfmlib_perf_event.c | 15 +++- lib/pfmlib_priv.h | 1 + 8 files changed, 376 insertions(+), 123 deletions(-) hooks/post-receive -- libpfm4 |