On Thu, Sep 6, 2018 at 12:54 AM, Paul Burton <paul.burton@xxxxxxxx> wrote: > Hi Huacai, > > On Wed, Sep 05, 2018 at 05:33:01PM +0800, Huacai Chen wrote: >> New Loongson-3 (Loongson-3A R2, Loongson-3A R3, and newer) has SFB >> (Store Fill Buffer) which can improve the performance of memory access. >> Now, SFB enablement is controlled by CONFIG_LOONGSON3_ENHANCEMENT, and >> the generic kernel has no benefit from SFB (even it is running on a new >> Loongson-3 machine). With this patch, we can enable SFB at runtime by >> detecting the CPU type (the expense is war_io_reorder_wmb() will always >> be a 'sync', which will hurt the performance of old Loongson-3). > > This looks unchanged since v3, and I didn't see a response to my email > here: > > https://marc.info/?l=linux-mips&m=153248530725061&w=2 > > I still haven't seen any explanation for why you can't just do this as a > one-liner in C, the same way we enable tons of other CPU features during > cpu_probe(). > > I'm not saying I'll never accept the assembly version, but I want an > explanation for why it's necessary first. If that's not something you > can give, at least describe in the commit message what goes wrong when > you try to do it in C as justification for not doing it that way. In practise, I found that sometimes there are boot failures if I enable SFB/LPA in cpu_probe(). I don't know why because processorr designers also haven't give me an explaination, but I think this may have some relationships to speculative execution. Huacai > > Thanks, > Paul
