Re: [dpdk-dev] next releases

[Cyril Chemparathy <cchemparathy@tilera.com>]

Hi Thomas,

On 8/25/2014 10:15 AM, Thomas Monjalon wrote:
> Hello all,
>
> I am back from holidays; thanks for all the
> patches/reviews/comments done during last weeks.
>
> I'd like to have a version 1.7.1, ideally at the end of this week.
>
> For the coming days,
>     - first priority is to integrate bug fixes
>     - some changes which do not imply API could be part of 1.7.1
>     - please, do not send more features until 1.8.0-rc1
>     - features that have been been *properly* reviewed or acked before
>       end of august will be integrated in 1.8.0-rc1
>     - all pending features which do not have any review will be postponed
>       after 1.8.0-rc1
>     - then rc2 will integrate new features if *properly* reviewed at that time
>
> I'd like to have some cleanups in version 1.8. Examples:
>     - get rid of doxygen warnings
>     - check if compile time options can be moved to run time
>     - rename some options (CONFIG_RTE_LIBRTE_*_PMD -> CONFIG_RTE_LIBRTE_PMD_*)
>     - merge common code between linux and bsd implementions
>     - check secondary process rights
>     - remove drivers lists from code for easy integration of new drivers
>     - use rte_eth_dev_atomic_read_link_status in drivers
>     - use librte_cfgfile instead of examples/qos_sched
>     - add sysfs functions as eal services
>     - replace printf calls by rte functions
>     - use new assert macros for unit tests
>     - remove kni traces from bsd
>     - remove bare metal traces
>     - compress test_lpm*_routes.h

Any thoughts on consolidating/cleaning up the timer interfaces?

Usage across rte_rdtsc(), rte_get_tsc_cycles(), and 
rte_get_timer_cycles() could use some rationalization, I think.

It looks like most code should use rte_get_timer_cycles() and generally 
honor user specified timer source selection.  The relatively few places 
that have a good cause to pin down on TSC should probably use 
rte_get_tsc_cycles() instead of rde_rdtsc().  On the other hand, if 
rte_rdtsc() is meant for direct use, why do we need the 
rte_get_tsc_cycles() wrapper?

Also, I'm not quite clear on the intended usage of rte_rdtsc_precise().  
I can't find uses of this function on master, and it is not quite clear 
to me if the intent is to replace rte_rdtsc() in some (or all?) places 
in the code.  Any insights on this?

Thanks
-- Cyril.