Skip to content

powerpc: P10 hardware counter (PMU/performance counters/perf:) Enhancements [FEAT]

Desnes Nunes requested to merge desnesn/centos-stream-9:rh2048779 into main

powerpc: P10 hardware counter (PMU/performance counters/perf:) Enhancements [FEAT]

BUGZILLA

Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=2048779

UPSTREAM STATUS

Upstream Status: Patches have been accepted on kernel/git/powerpc/linux.git

CONFLICTS

A normal conflict on patch 0026 with no functional change.

BUILD INFORMATION

Build Info: http://brewweb.engineering.redhat.com/brew/taskinfo?taskID=44402042

TESTING

The following perf tests verified this feature:

# uname -a
Linux ltcden9-lp3.aus.stglabs.ibm.com 5.14.0-70.PATCHED_rh2048779.el9.ppc64le #1 SMP Wed Mar 9 21:44:26 EST 2022 ppc64le ppc64le ppc64le GNU/Linux

# perf -v
perf version 5.14.0-70.PATCHED_rh2048779.el9.ppc64le

# perf record -I?
available registers: r0 r1 r2 r3 r4 r5 r6 r7 r8 r9 r10 r11 r12 r13 r14 r15 r16 r17 r18 r19 r20 r21 r22 r23 r24 r25 r26 r27 r28 r29 r30 r31 nip msr orig_r3 ctr link xer ccr softe trap dar dsisr sier mmcra mmcr0 mmcr1 mmcr2 mmcr3 sier2 sier3 pmc1 pmc2 pmc3 pmc4 pmc5 pmc6 sdar siar

# perf mem record -t load dd if=/dev/zero of=/dev/null bs=4k count=1M
# perf mem report --sort="local_weight,local_ins_lat,local_p_stage_cyc" --stdio |head -n 20
# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 2K of event 'cpu/mem-loads/'
# Total weight : 73231
# Sort order   : local_weight,local_ins_lat,local_p_stage_cyc
#
# Overhead       Samples  Local Weight  Finish Cyc     Dispatch Cyc 
# ........  ............  ............  .............  .............
#
     5.03%           368  10            7              1
     4.87%            11  324           7              1
     4.81%            11  320           7              1
     4.48%            10  328           7              1
     3.73%           210  13            7              1
     3.70%           226  12            7              1
     3.29%            86  28            7              2
     3.06%           204  11            7              1

# perf mem report --sort="weight,ins_lat,p_stage_cyc" --stdio |head -n 20
# To display the perf.data header info, please use --header/--header-only options.
#
#
# Total Lost Samples: 0
#
# Samples: 2K of event 'cpu/mem-loads/'
# Total weight : 73231
# Sort order   : weight,ins_lat,p_stage_cyc
#
# Overhead       Samples  Weight        Global Finish_cyc  Global Dispatch_cyc
# ........  ............  ............  .................  ...................
#
     5.03%           368  3680          2576               368
     4.87%            11  3564          77                 11
     4.81%            11  3520          77                 11
     4.48%            10  3280          70                 10
     3.73%           210  2730          1470               210
     3.70%           226  2712          1582               226
     3.29%            86  2408          602                172
     3.06%           204  2244          1428               204

DESCRIPTION

This FEAT request brings in a couple of patches which provide multiple enhancements for POWER10 hardware counters support.

Signed-off-by: Desnes A. Nunes do Rosario drosario@redhat.com

Edited by Desnes Nunes

Merge request reports