Comment 91 for bug 177895

Hi,

It's good to hear that things are more promising now.

The overall IRQ delta over 10 seconds (from your last results of interrupts_normal_xorg.log) look more promising - IRQ 0 is definitely not being saturated now, but it does appear that you system is busy somewhere causing ~7K rescheduling interrupts per second.

           CPU0 CPU1
  0: 2597 2582 IO-APIC-edge timer
  1: 0 0 IO-APIC-edge i8042
  8: 4 3 IO-APIC-edge rtc
  9: 18 16 IO-APIC-fasteoi acpi
 12: 0 0 IO-APIC-edge i8042
 14: 0 0 IO-APIC-edge libata
 15: 0 0 IO-APIC-edge libata
 16: 35 25 IO-APIC-fasteoi uhci_hcd:usb5, nvidia
 17: 0 0 IO-APIC-fasteoi uhci_hcd:usb1
 18: 0 0 IO-APIC-fasteoi uhci_hcd:usb2
 19: 4 4 IO-APIC-fasteoi ehci_hcd:usb3
 21: 1 1 IO-APIC-fasteoi ehci_hcd:usb4
 22: 0 0 IO-APIC-fasteoi uhci_hcd:usb6, ohci1394, HDA Intel
 23: 0 0 IO-APIC-fasteoi uhci_hcd:usb7
215: 384 410 PCI-MSI-edge iwl4965
216: 14 16 PCI-MSI-edge eth0
217: 3 2 PCI-MSI-edge ahci
NMI: 0 0 Non-maskable interrupts
LOC: 2110 1800 Local timer interrupts
RES: 68628 75310 Rescheduling interrupts
CAL: 0 0 function call interrupts
TLB: 0 6 TLB shootdowns
TRM: 0 0 Thermal event interrupts
SPU: 0 0 Spurious interrupts
ERR: 0
MIS: 0

Running vmstat 1 for ~15-30 seconds will give an idea of how busy the system is - primarily the context switches/second will be interesting to see.

Colin