Re: [PATCH v1 2/2] perf parse-events: Corrections to topdown sorting

[James Clark]

On 05/03/2025 8:37 am, Ian Rogers wrote:
> In the case of '{instructions,slots},faults,topdown-retiring' the
> first event that must be grouped, slots, is ignored causing the
> topdown-retiring event not to be adjacent to the group it needs to be
> inserted into. Don't ignore the group members when computing the
> force_grouped_index.
>
> Make the force_grouped_index be for the leader of the group it is
> within and always use it first rather than a group leader index so
> that topdown events may be sorted from one group into another.
>
> Reported-by: Dapeng Mi <dapeng1.mi@xxxxxxxxxxxxxxx>
> Closes: https://lore.kernel.org/lkml/20250224083306.71813-2-dapeng1.mi@xxxxxxxxxxxxxxx/
> Signed-off-by: Ian Rogers <irogers@xxxxxxxxxx>

Testing on Arm seems ok, but presumably this doesn't change anything 
there because arch_evsel__must_be_in_group() is always false.

On x86 I ran into the topdown metrics not opening on cpu_core at all, so 
I'm not sure if I'm able to test that the original issue is fixed on my 
machine. From looking at the link the issue is that the ungrouped 
topdown event is "<not supported>", but I always see that regardless of 
grouping despite perf list saying it exists:

 $ perf list --unit cpu_core | grep -i topdown
  topdown-bad-spec OR cpu_core/topdown-bad-spec/     [Kernel PMU event]
  topdown-be-bound OR cpu_core/topdown-be-bound/     [Kernel PMU event]
  topdown-br-mispredict OR cpu_core/topdown-br-mispredict/[Kernel PMU 
event]
  topdown-fe-bound OR cpu_core/topdown-fe-bound/     [Kernel PMU event]
  topdown-fetch-lat OR cpu_core/topdown-fetch-lat/   [Kernel PMU event]
  topdown-heavy-ops OR cpu_core/topdown-heavy-ops/   [Kernel PMU event]
  topdown-mem-bound OR cpu_core/topdown-mem-bound/   [Kernel PMU event]
  topdown-retiring OR cpu_core/topdown-retiring/     [Kernel PMU event]
  topdown.backend_bound_slots
  topdown.bad_spec_slots
  topdown.br_mispredict_slots
  topdown.memory_bound_slots
       [TOPDOWN.MEMORY_BOUND_SLOTS. Unit: cpu_core]


 $ sudo perf stat -e topdown-retiring -- true
 Performance counter stats for 'true':
     <not counted>   cpu_atom/topdown-retiring/           (0.00%)
   <not supported>   cpu_core/topdown-retiring/


 $ sudo perf stat -e topdown-retiring -vvv -- true
Control descriptor is not initialized
Opening: topdown-retiring
------------------------------------------------------------
perf_event_attr:
  type                             10 (cpu_atom)
  size                             136
  config                           0xc2 (topdown-retiring)
  sample_type                      IDENTIFIER
  read_format                      TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
  disabled                         1
  inherit                          1
  enable_on_exec                   1
------------------------------------------------------------
sys_perf_event_open: pid 151404  cpu -1  group_fd -1  flags 0x8 = 3
Opening: topdown-retiring
------------------------------------------------------------
perf_event_attr:
  type                             4 (cpu_core)
  size                             136
  config                           0x8000 (topdown-retiring)
  sample_type                      IDENTIFIER
  read_format                      TOTAL_TIME_ENABLED|TOTAL_TIME_RUNNING
  disabled                         1
  inherit                          1
  enable_on_exec                   1
------------------------------------------------------------
sys_perf_event_open: pid 151404  cpu -1  group_fd -1  flags 0x8
sys_perf_event_open failed, error -22
switching off exclude_guest for PMU cpu_core
Using PERF_SAMPLE_READ / :S modifier is not compatible with inherit, 
falling back to no-inherit.
Warning:
topdown-retiring event is not supported by the kernel.