LTTng 0.51, improved timestamping
Hi,
LTTng 0.51 for 2.6.27.4 does the following :
It improves timestamping on architectures with non-synchronized TSCs in LTTng 0.51. I also now use per-cpu timers to do some time-related stuff I previously used IPIs for. It involves being more closely tied to the CPU hotplug callbacks.
I also replaced x86 tsc_sync.c test by LTTng architecture independent test. (this will be posted separately on LKML for 2.6.28-rc3 tomorrow along with the "tracer clock" patchset).
The event trap_entry, syscall_entry, irq_entry, softirq entry, exit and raise now use smaller fields (2 bytes), which should be enough to encode the IDs for all architectures. This will principally make a difference with non-aligned tracing for most of those events.
Mathieu
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
disable tsc synth when ! trace clock ?
tsc synth get/put ?
amd64 :
- single cpu -> tsc_is_sync : no force on unstable tsc ?
- hotplug out all cpus, recheck
Quoted 29 lines · * Mathieu Desnoyers ([email protected]) wrote:
Hi,
LTTng 0.51 for 2.6.27.4 does the following :
It improves timestamping on architectures with non-synchronized TSCs in LTTng 0.51. I also now use per-cpu timers to do some time-related stuff I previously used IPIs for. It involves being more closely tied to the CPU hotplug callbacks.
I also replaced x86
tsc_sync.ctest by LTTng architecture independent test. (this will be posted separately on LKML for 2.6.28-rc3 tomorrow along with the "tracer clock" patchset).The event
trap_entry,syscall_entry,irq_entry, softirq entry, exit and raise now use smaller fields (2 bytes), which should be enough to encode the IDs for all architectures. This will principally make a difference with non-aligned tracing for most of those events.Mathieu
Signature · 7 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68 _______________________________________________ ltt-dev mailing list [email protected] http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Please ignore this last message. As some of you may have noticed, this is actually the result of a brainstorming while being half-asleep. Those things will be carefully looked at this morning. ;)
Mathieu
Quoted 43 lines · * Mathieu Desnoyers ([email protected]) wrote:
disable tsc synth when ! trace clock ?
tsc synth get/put ?amd64 :
- single cpu ->tsc_is_sync: no force on unstable tsc ?
- hotplug out all cpus, recheckQuoted 29 lines · * Mathieu Desnoyers ([email protected]) wrote:
Hi,
LTTng 0.51 for 2.6.27.4 does the following :
It improves timestamping on architectures with non-synchronized TSCs in LTTng 0.51. I also now use per-cpu timers to do some time-related stuff I previously used IPIs for. It involves being more closely tied to the CPU hotplug callbacks.
I also replaced x86
tsc_sync.ctest by LTTng architecture independent test. (this will be posted separately on LKML for 2.6.28-rc3 tomorrow along with the "tracer clock" patchset).The event
trap_entry,syscall_entry,irq_entry, softirq entry, exit and raise now use smaller fields (2 bytes), which should be enough to encode the IDs for all architectures. This will principally make a difference with non-aligned tracing for most of those events.Mathieu
Signature · 7 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68 _______________________________________________ ltt-dev mailing list [email protected] http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-devSignature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68