From mboxrd@z Thu Jan 1 00:00:00 1970 From: Alexey Brodkin Date: Wed, 26 Oct 2016 10:35:35 +0000 Subject: [Buildroot] [PATCH] config: add Synopsys nSIM and nSIM OSCI platforms In-Reply-To: <20161025220122.0ae10f17@free-electrons.com> References: <1470394586-6091-1-git-send-email-vzakhar@synopsys.com> <20160805140115.071073db@free-electrons.com> <1470399797.3430.75.camel@synopsys.com> <20161025220122.0ae10f17@free-electrons.com> Message-ID: <1477478099.2593.35.camel@synopsys.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: buildroot@busybox.net Hi Thomas, On Tue, 2016-10-25 at 22:01 +0200, Thomas Petazzoni wrote: > Hello, > > On Fri, 5 Aug 2016 12:24:54 +0000, Alexey Brodkin wrote: > > > > > > > > > > > > > > nSIM OSCI emulates more functional boards having not only serial port > > > > but as well: > > > > ?* LCD screen > > > > ? * PS/2 keyboard > > > > ???* Ethernet controller (which communicates with the host) > > > > > > > > Signed-off-by: Vlad Zakharov ?? > > > > > > Thanks for this contribution. It would be useful to add a readme.txt > > > file that explains how to start the result of the Buildroot build with > > > those simulation platforms, especially the one that is freely available. > > Would it be possible to resubmit with a readme.txt file (not one per > configuration, but a global one that covers all those configurations). > > > > > > > > > > > > > > ?configs/snps_nsim_700_defconfig??????????| 18 ++++++++++++++++++ > > > > ?configs/snps_nsim_hs38_defconfig?????????| 19 +++++++++++++++++++ > > > > ?configs/snps_nsim_hs38_smp_defconfig?????| 19 +++++++++++++++++++ > > > > ?configs/snps_nsimosci_700_defconfig??????| 19 +++++++++++++++++++ > > > > ?configs/snps_nsimosci_hs38_defconfig?????| 20 ++++++++++++++++++++ > > > > ?configs/snps_nsimosci_hs38_smp_defconfig | 20 ++++++++++++++++++++?? > > > > > > Do we really need all those combinations? I do understand the 700 vs. > > > HS38 difference, but couldn't the rest be handled with just different > > > DTBs ? It seems annoying to have both one kernel defconfig for each of > > > those, and then one Buildroot configuration for each of those as well.?? > > > > Unfortunately it's not only different .dtb. > > On ARC for SMP we need to modify kernel configuration as well. > > > > I hope some day we'll be able to run the same one kernel binary on any > > flavor of ARCv2 CPU (be it UP or SMP) but now we cannot. > > OK. > > Then is there a way to where the nsim and its corresponding nsimosci > variant? What is the difference between the two? > > Looking at the kernel defconfigs, it seems that the only important > difference is the built-in DTB: > > -CONFIG_ARC_BUILTIN_DTB_NAME="nsim_hs" > +CONFIG_ARC_BUILTIN_DTB_NAME="nsimosci_hs" > > Can't you use an external DTB instead ? Hm... that's what happens when new pair of eyes starts to look at things :) Right I think we may unify those configs, for that we'll need to merge options for drivers and then the only real difference will be DTB. A small problem here is both simulators as of now are supposed to load just one Elf with everything built-in. And so we build 2 different binaries which at high-level differ in built-in .dtb. Indeed it's much more elegant to keep only 1 defconfig for all different flavors of boards like what IMX has now with http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/arch/arm/configs/imx_v6_v7_defconfig But I think it worth a separate discussion on linux-snps mailing list. As for this discussion I'd prefer to keep using in-tree kernel defconfigs compared to option to have combined defconfig in BR. That IMHO simplifies maintenance, i.e. if defconfigs get changed in upstream we won't miss thanges on the next kernel version bump in BR. > > In general I agree so many configs in BR look weird and what's worse requires > > much more work on maintenance. So we may try to select more useful configs but > > for me it's hard to make this decision because from one point of view in case of > > ARCv2 we're more interested in SMP configs (that's one of the main differences compared > > to older ARCompact CPUs) but in simulation SMP setups work significantly slower > > and IMHO UP versions are much more usable if SMP itself is not a requirement. > > Is nSIM even available for free ? Can I download it and test those > defconfigs ? Sure you may download a copy of Free nSIM here after a simple registration: https://www.synopsys.com/cgi-bin/dwarcnsim/req1.cgi Since that's a completely new tool for you I may expect you face some issues in that case please don't hesitate to contact me. Another important note: as of today Free version of nSIM only supports UP (single-core) configurations, which means you won't be able to try SMP configs in run-time. -Alexey