Perf record runs with wildly different output sizes.

gistfile1.txt

bram@ubuntu64:~/apps/BloxTrace$ /usr/bin/perf record --freq 2000 --call-graph dwarf ./bench segment
Loaded 455 primitives.
Testing ray against 455 primitives.
[ perf record: Woken up 289 times to write data ]
[nvidia] with build id d9b6653c0d39e1618ccd50318de6da8df74820a1 not found, continuing without symbols
[ perf record: Captured and wrote 72.134 MB perf.data (8979 samples) ]

bram@ubuntu64:~/apps/BloxTrace$ /usr/bin/perf evlist -v
cycles:pp: size: 112, { sample_period, sample_freq }: 2000, sample_type: IP|TID|TIME|CALLCHAIN|PERIOD|REGS_USER|STACK_USER, disabled: 1, inherit: 1, mmap: 1, comm: 1, freq: 1, enable_on_exec: 1, task: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, exclude_callchain_user: 1, mmap2: 1, comm_exec: 1, sample_regs_user: 0xff0fff, sample_stack_user: 8192

bram@ubuntu64:~/apps/BloxTrace$ /usr/bin/perf record --freq 2000 --call-graph dwarf ./bench segment
Loaded 455 primitives.
Testing ray against 455 primitives.
[ perf record: Woken up 1 times to write data ]
[ perf record: Captured and wrote 0.254 MB perf.data (30 samples) ]

bram@ubuntu64:~/apps/BloxTrace$ /usr/bin/perf evlist -v
cycles:pp: size: 112, { sample_period, sample_freq }: 2000, sample_type: IP|TID|TIME|CALLCHAIN|PERIOD|REGS_USER|STACK_USER, disabled: 1, inherit: 1, mmap: 1, comm: 1, freq: 1, enable_on_exec: 1, task: 1, precise_ip: 2, sample_id_all: 1, exclude_guest: 1, exclude_callchain_user: 1, mmap2: 1, comm_exec: 1, sample_regs_user: 0xff0fff, sample_stack_user: 8192

bram@ubuntu64:~/apps/BloxTrace$ /usr/bin/perf --version
perf version 4.4.90

$ uname -a
Linux ubuntu64 4.4.0-98-generic #121-Ubuntu SMP Tue Oct 10 14:24:03 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

NOTE: These short/long outputs happens if generating the default cpu-cycles:pp events.
When explicitly using -e cpucycles the output is always large.
So it seems like an issue with the :pp suffix.