Re: [RFC PATCH 0/4] Speed booting by sorting ORC unwind tables at build time

[Peter Zijlstra]

On Fri, Nov 08, 2019 at 10:21:36AM +0100, Peter Zijlstra wrote:
> On Fri, Nov 08, 2019 at 09:42:55AM +0800, Shile Zhang wrote:
> 
> > > Can sort{ex,orc}table() be ran concurrently? Do they want to be the same
> > > (threaded) tool?
> 
> > I think it is possible to do those sort work concurrently, likes deferred
> > memory init which is big boot time speed up.
> > But I don't know if the exception table and ORC unwind tables can be
> > deferred, due to those tables might be used in early boot time, for early
> > exception handling and early debugging. I'm not familiar with that.
> 
> I meant at link time, run both sorts concurrently such that we only have
> to wait for the longest, instead of the sum of them.
> 
> They're not changing the same part of the ELF file, so it should be
> possible to have one tool have multiple threads, each sorting a
> different table.
> 
> Aside from the .ex_table and ORC there's also .jump_table that wants
> sorting (see jump_label_sort_entries()).

Oh, and I'll be adding .static_call_sites soon, see:

  https://lkml.kernel.org/r/20191007082708.013939311@xxxxxxxxxxxxx

(I should repost that)

That gives us 4 tables to sort which we can do concurrently in 4
threads.

> I agree that doing it at link time makes sense, I just hate to do all
> this sorting in sequence and blowing up the link time. I don't build for
> customers, I build for single use boot and linking _SUCKS_.