On Thu, Aug 17, 2017 at 12:42:19AM -0400, Christopher Li wrote: > Hi Luc, Hi, > Thank you so much for the patch. > > Here is the kernel full allmodconfig running ssa-mini-v1 on my new > compile server. On the new server the timing is very consistent. > The variance of run to run is under 0.1 second. > > ssa-mini-v1: > 1205.82user 463.65system 1:16.97elapsed 2168%CPU (0avgtext+0avgdata > 536216maxresident)k > 0inputs+12824outputs (0major+132162996minor)pagefaults 0swaps > 1206.35user 463.27system 1:16.95elapsed 2169%CPU (0avgtext+0avgdata > 536296maxresident)k > 0inputs+12824outputs (0major+132156314minor)pagefaults 0swaps > > > This is the rc5 as base line. > 1173.42user 453.56system 1:15.21elapsed 2163%CPU (0avgtext+0avgdata > 238072maxresident)k > 0inputs+12784outputs (0major+128858147minor)pagefaults 0swaps > 1172.86user 453.53system 1:15.14elapsed 2164%CPU (0avgtext+0avgdata > 238076maxresident)k > 0inputs+12784outputs (0major+128858804minor)pagefaults 0swaps > > So the ssa-mini-v1 is about 2% slower than the current rc5. I haven't measured myself, but I guess that is because this series also includes the removal of simplify_loads(). I wonder what would be these numbers if you measured: 1) r5 + removal of simplify_loads() 2) r5 + sssa-mini-v1 (which includes the removal). > BTW, how do you want your patches to be merged? Assume there is > some feed back. Do you want to come up with V2 V3 or have them merged > to master then come up with fix up on master? If there is some reasonable needed changes I can make a -v2 But I think that most changes can be done later as additional patches once it's already in master. > diff -ruN linux-checker/rc5/drivers/base/firmware_class.sp linux-checker/ssa/drivers/base/firmware_class.sp > --- linux-checker/rc5/drivers/base/firmware_class.sp 2017-08-17 > +++ linux-checker/ssa/drivers/base/firmware_class.sp 2017-08-17 > @@ -1 +1 @@ > -drivers/base/firmware_class.c:395:9: warning: context imbalance in fw_free_buf' - wrong count at exit > +drivers/base/firmware_class.c:391:13: warning: context imbalance in 'fw_free_buf' - wrong count at exit There are often a little bunch of such changes, I've learnt to not worry about them. Basically you have the same warning but at a different line:colonm, something even a different file (source file / include file). The cause is that the BBs are packed or merged differently and the context checking use the position of the BB and no the postion of a specific token. It's annoying, I would like to not have that, but nothing to worry about. > diff -ruN linux-checker/rc5/drivers/block/drbd/drbd_actlog.sp linux-checker/ssa/drivers/block/drbd/drbd_actlog.sp > --- linux-checker/rc5/drivers/block/drbd/drbd_actlog.sp 2017-08-17 > +++ linux-checker/ssa/drivers/block/drbd/drbd_actlog.sp 2017-08-17 > @@ -26,3 +26,4 @@ > drivers/block/drbd/drbd_int.h:1773:14: error: incompatible types in comparison expression (different address spaces) > drivers/block/drbd/drbd_actlog.c:469:44: error: incompatible types in comparison expression (different address spaces) > drivers/block/drbd/drbd_actlog.c:173:16: warning: context imbalance in '_drbd_md_sync_page_io' - different lock contexts for basic block > +drivers/block/drbd/drbd_actlog.c:1244:24: warning: context imbalance in 'drbd_rs_del_all' - different lock contexts for basic block This one is different, it's part of the 8 or so new 'bad context' I was talking about. I'll investigate about. I have also one such warning that was present in -rc5 and absent in sssa-mini-v1. -- Luc -- To unsubscribe from this list: send the line "unsubscribe linux-sparse" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
