FS#61824 - perf stat wrong unit scaling. (Fixed upstream, package needs updating)
Attached to Project:
Community Packages
Opened by Peter Cordes (pcordes) - Thursday, 21 February 2019, 04:35 GMT
Last edited by Balló György (City-busz) - Tuesday, 19 March 2019, 06:02 GMT
Opened by Peter Cordes (pcordes) - Thursday, 21 February 2019, 04:35 GMT
Last edited by Balló György (City-busz) - Tuesday, 19 March 2019, 06:02 GMT
|
Details
Description: perf stat scales "secondary" units twice. This
is an upstream bug that was introduced somewhat recently,
and fixed by a patch from 2018 Dec 17
https://lore.kernel.org/patchwork/patch/1025968/ Commit-ID: 57ddf09173c1e7d0511ead8924675c7198e56545 that only changes 2 lines in the front-end. Additional info: * package version(s): perf 4.20-2, linux 4.20.3.arch1-1 Steps to reproduce: perf stat awk 'BEGIN{for(i=0;i<10000000;i++){}}' Performance counter stats for 'awk BEGIN{for(i=0;i<10000000;i++){}}': 357.49 msec task-clock # 0.993 CPUs utilized 3 context-switches # 8.403 M/sec 1 cpu-migrations # 2.801 M/sec 173 page-faults # 484.594 M/sec 1,383,853,299 cycles # 3876339.773 GHz 3,466,214,435 instructions # 2.50 insn per cycle 931,171,255 branches # 2608322843.137 M/sec 53,122 branch-misses # 0.01% of all branches 0.359906618 seconds time elapsed 0.357623000 seconds user 0.000000000 seconds sys My average CPU clock speed was 3.87 GHz, not 3876339.773 GHz. Obviously most of the other /sec units are wrong, too. |
This task depends upon
Closed by Balló György (City-busz)
Tuesday, 19 March 2019, 06:02 GMT
Reason for closing: Fixed
Additional comments about closing: perf 5.0-1
Tuesday, 19 March 2019, 06:02 GMT
Reason for closing: Fixed
Additional comments about closing: perf 5.0-1