Yes, it makes sence. I also run perf top 15-20 secs later after fio. Samples: 138K of event 'cycles:ppp', Event count (approx.): 95410433826 Overhead Shared Object Symbol 93.41% libc-2.22.so [.] __memcpy_avx_unaligned 2.36% fio [.] get_io_u 0.65% fio [.] ramp_time_over 0.63% fio [.] io_u_sync_complete 0.34% fio [.] td_io_queue 0.33% [unknown] [k] 0x00007f53f8a82b06 0.33% fio [.] __fio_gettime __memcpy_avx_unaligned /lib64/libc-2.22.so Huh? No selection. Report to linux-kernel@xxxxxxxxxxxxxxx It's a tremendous results. But I'm still thinking is there any chance for further I/O perf boost. Anton -----Original Message----- From: Jeff Furlong [mailto:jeff.furlong@xxxxxxx] Sent: Monday, December 04, 2017 9:18 PM To: Jens Axboe <axboe@xxxxxxxxx>; Gavriliuk, Anton (HPS Ukraine) <anton.gavriliuk@xxxxxxx>; Robert Elliott (Persistent Memory) <elliott@xxxxxxx>; Rebecca Cran <rebecca@xxxxxxxxxxxx>; Sitsofe Wheeler <sitsofe@xxxxxxxxx> Cc: fio@xxxxxxxxxxxxxxx; Kani, Toshimitsu <toshi.kani@xxxxxxx> Subject: RE: fio 3.2 >That seems so odd, it makes no sense. --no-children added to perf report? >I think this is skewed because the code is so hot for startup. Perhaps you want to add --ramp_time=5s to remove startup jitter from host/device? Regards, Jeff ��.n��������+%������w��{.n�������^n�r������&��z�ޗ�zf���h���~����������_��+v���)ߣ�
