Tony,
On Wed, Jul 2, 2008 at 8:13 PM, Luck, Tony <[EMAIL PROTECTED]> wrote:
>> Except that the data passed for one or the other may not be the same.
>> However, it is true that as of today the PMC information is a subset of
>> the PMD's.
>
> True the PMDs are generally scalar values while the PMCs are often
> bitmasks of feature bits ... but so long as they both fit into the
> same scalar type (u64?) then it shouldn't really matter. Are there
> PMUs where this isn't true ... i.e. is there an architecture where
> you could use u32 for PMC but need u64 for PMD?
>
Perfmon always presents PMC and PMD registers as 64 bits regardless
of what the actual registers implement.
I was more talking about the content of pfarg_pmc vs. pfarg_pmd.
One is a subset of the other.
>> Are you also suggestion we merge the two syscalls to write the registers?
>
> Yes. Just a variation on your suggestion of:
>
> pfm_write_regs(int fd, int type, void *arg, size_t arg_sz)
>
> Drop the "type" argument and have "arg" point to an array of
> {regnum, data} pairs where negative regnum values refer to
> PMC and positive ones to PMD.
>
The issue is that when we add sampling support + event sets, new fields
are introduced, some apply only to PMDs.
-------------------------------------------------------------------------
Sponsored by: SourceForge.net Community Choice Awards: VOTE NOW!
Studies have shown that voting for your favorite open source project,
along with a healthy diet, reduces your potential for chronic lameness
and boredom. Vote Now at http://www.sourceforge.net/community/cca08
_______________________________________________
perfmon2-devel mailing list
perfmon2-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/perfmon2-devel
