Hi Jon, > -----Original Message----- > From: Jon Nettleton [mailto:j...@solid-run.com] > Sent: 2018年9月7日 14:03 > To: Fabio Estevam <feste...@gmail.com> > Cc: Peng Fan <peng....@nxp.com>; Diego Dorta <diegohdo...@gmail.com>; > Fabio Estevam <fabio.este...@nxp.com>; U-Boot Mailing List > <u-boot@lists.denx.de> > Subject: Re: [U-Boot] [PATCH V5 00/31] imx: add i.MX8M support and i.MX8MQ > EVK > > On Thu, Sep 6, 2018 at 3:37 PM Fabio Estevam <feste...@gmail.com> wrote: > > > > Hi Peng, > > > > On Fri, Aug 31, 2018 at 3:27 AM, Jon Nettleton <j...@solid-run.com> wrote: > > > > > Peng, > > > > > > Will you be releasing the script you are using to convert that code > > > generated by the ddr training tool into this format? I would like > > > to test your patchset on our new SOMs and that would be very useful. > > > > Any comments about the DDR init script? Looks like we are stuck in > > this issue for a long time. > > > > I have been chewing on this over the past week, and I am starting to think > that > u-boot is not necessarily the right place for the DDR initialization for this > platform. We require ATF in order to load the > lpddr4 firmware regardless, so why not do the DDR training and configuration > there? This would also abstract out the proprietary memory bits so the code > doesn't need to be re-duplicated in the various bootloaders. In theory it > should > also simplify the MX8 image scripts since SPL would not be a requirement.
On i.MX8M, SPL is used to initialize DDR and load ATF/TEE/U-Boot to memory and kick ATF. Two option: 1. Move DDR code out of SPL and make it a standalone image, then pad it the SPL. The drawback is need a new repo holding ddr code. 2. Restructure the DDR code the make it generic enough. One engineer in NXP is looking into restructure code, but priority is low. Regards, Peng. > > Thoughts? _______________________________________________ U-Boot mailing list U-Boot@lists.denx.de https://lists.denx.de/listinfo/u-boot
