Description of problem: Because the dynamic speed allocation technology of the Nehalem processors allows the processors to run @ a faster than reported speed by disabling the not in use processors the cpuscaling test reports a failure where the SUT shows a +12% gain above in performance than expected (current limit is 10%). How reproducible: Every time, run hts-5.2-20 on RHEL5.3 on a Nehalem with Turbo enabled and it'll fail; sometimes under the -xen kernel there are enough other things going on that the turbo boost won't occur. Additional info: Current solution is to add 8% speed increase to the expected results and then to the 10% margin of error from there. Intel PR claims up2 10% speed bump, but there are other things going on inthe system other than cpu clock, so we've and others have averaged around 7-8% effective increase. Because the margin of error is 10% this should also cover the case where Turbo didn't help though it'll only have a 2% slower window to hit.
Created attachment 326431 [details] cpuscaling.py patch to change expected perfomance for ida This test changes expected perfomance by 8% when the "ida" cpu flag is present. Also, infomation was added to error messages.
Tested on gcase-greencity.usersys.redhat.com as below: - For hts-5.2.20.el5, `hts run -t cpuscaling` failed. # hts print plan Test Plan: ---------------------------------------------------------------- usb pnp_PNP0a03 cdrom scd0 storage_model_DVD_SHD_16S1S cdrom scd1 storage_model_iHAS120___6 network eth1 net_00_13_20_f7_56_bd network eth0 net_00_13_20_f7_56_bc memory video pci_10de_421 cpuscaling acpi_CPU0 core storage sda storage_serial_SATA_Hitachi_HDS7210_GTE205P8GW8R5L - Upgraded to hts-5.3-15.el5, `hts run -t cpuscaling -v 0` passed, `hts run -t cpuscaling` passed with 16 cpuscaling. # hts print plan /var/hts Test Plan: ---------------------------------------------------------------- usb pnp_PNP0a03 dvd scd0 storage_model_DVD_SHD_16S1S dvd scd1 storage_model_iHAS120___6 cdrom scd0 storage_model_DVD_SHD_16S1S cdrom scd1 storage_model_iHAS120___6 network eth1 net_00_13_20_f7_56_bd network eth0 net_00_13_20_f7_56_bc memory video pci_10de_421 cpuscaling 0 acpi_CPU0 cpuscaling 1 acpi_CPU1 cpuscaling 2 acpi_CPU2 cpuscaling 3 acpi_CPU3 cpuscaling 4 acpi_CPU4 cpuscaling 5 acpi_CPU5 cpuscaling 6 acpi_CPU6 cpuscaling 7 acpi_CPU7 cpuscaling 8 acpi_CPU8 cpuscaling 9 acpi_CPU9 cpuscaling 10 acpi_CPUA cpuscaling 11 acpi_CPUB cpuscaling 12 acpi_CPUC cpuscaling 13 acpi_CPUD cpuscaling 14 acpi_CPUE cpuscaling 15 acpi_CPUF core storage sda storage_serial_SATA_Hitachi_HDS7210_GTE205P8GW8R5L profiler info Verified the bug.
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHBA-2009-0047.html