You are right, my bad!
I fixed that now in the 3.5 branch in CVS.
Version was bumped to 3.52 so we have a trace.
Pull from CVS and let me know if this now works, then I'll make
a new release.
Thanks for tracking this one down.
On Tue, Sep 2, 2008 at 11:29 PM, William Cohen <wcohen@...> wrote:
> stephane eranian wrote:
>> Hello everyone,
>> Following that discussion we've had last week on this list about a
>> libpfm bug which
>> caused corruption in the Intel Core event table, Will provided a fix.
>> Because the bug
>> was serious, I have decided to make an intermediate release,
>> libpfm-3.51. It only contains
>> that Pentium II fix compared to v3.5.
>> I think the bug affected people running in 32-bit mode only.
>> Please give it a try in your setup.
>> You can download the version at the usual place:
> Hi Stephane,
> The libpfm-3.51 doesn't look quite right the following diff with the cvs
> repository shows an extra .pmu_detect initialization in the code to the
> wrong function:
> $ diff -u libpfm-3.51/lib/pfmlib_i386_p6.c
> < deleted other unrelated changes >
> @@ -572,7 +575,6 @@
> .get_event_counters = pfm_i386_p6_get_event_counters,
> .dispatch_events = pfm_i386_p6_dispatch_events,
> .pmu_detect = pfm_i386_p6_detect_pii,
> - .pmu_detect = pfm_i386_p6_init_pii,
> .pmu_init = pfm_i386_p6_init_pii,
> .get_impl_pmcs = pfm_i386_p6_get_impl_perfsel,
> .get_impl_pmds = pfm_i386_p6_get_impl_perfctr,