Hi Shawn, Am Freitag, 3. Juni 2016, 11:35:32 schrieb Shawn Lin: > ? 2016/6/3 9:25, Xing Zheng ??: > > On 2016?06?03? 08:54, Shawn Lin wrote: > >> I check all the Socs including RK2928/3000/3066/3028X/316X/312X/ > >> 3190/3188/3228/3368/3399/3036, and find all of them use high 16-bit > >> as write mask. Obviously we don't need ROCKCHIP_SOFTRST_HIWORD_MASK > >> any more(actually I don't know why we need it before). This patch > >> removes it to simplify the code and save a little cpu cycle when calling > >> assert or deassert callback. > > > > In my opinion, this flag can be used for compatibility, we can not > > ensure that our SoCs will not use the 32bit SOFTRST_CONs in future. > > > > Thanks. > > Thanks for sharing your thought. > > I'm not 100% sure, but I'm 99% sure about we won't let > it happened. You have to consider the backward compatibility > rather than the future ones. If you got a chip with 10 bit, or > 8bit for SOFTRST_CONX, so how do you wanna deal with it? > Should we now add ROCKCHIP_SOFTRST_X_BIT_MASK? :) older SoCs like the rk2818 (ARM9-based) do actually use 32bit softrst registers. And if I ever get my hands on one of those, I'd actually try to support it :-) . So I'd really like to keep the flag. Heiko
