Transparent on-demand memory setup initialization embedded in the (GFP) buddy allocator
Quoted 10 lines · Andrew Morton
Quoted 2 lines · On Wed, 26 Jun 2013 11:22:48 +0200 Ingo Molnar <[email protected]> wrote:
except that on 32 TB
systems we don't spend ~2 hours initializing 8,589,934,592 page heads.That's about a million a second which is crazy slow - even my prehistoric desktop is 100x faster than that.
Where's all this time actually being spent?
The complexity of a directory-lookup architecture to make the (intrinsically unscalable) cache-coherency protocol scalable gives you a ~1us roundtrip to remote NUMA nodes.
Probably a lot of time is spent in some memsets, and RMW cycles which are setting page bits, which are intrinsically synchronous, so the initialising core can't get to 12 or so outstanding memory transactions.
Since EFI memory ranges have a flag to state if they are zerod (which may be a fair assumption for memory on non-bootstrap processor NUMA nodes), we can probably collapse the RMWs to just writes.
A normal write will require a coherency cycle, then a fetch and a writeback when it's evicted from the cache. For this purpose, non-temporal writes would eliminate the cache line fetch and give a massive increase in bandwidth. We wouldn't even need a store-fence as the initialising core is the only one online.
Daniel
Signature · 2 lines
Daniel J Blueman Principal Software Engineer, Numascale Asia
Quoted 17 lines · Daniel J Blueman
Quoted 2 lines · Andrew Morton
On Wed, 26 Jun 2013 11:22:48 +0200 Ingo Molnar <[email protected]>
Quoted 3 lines · wrote:
Quoted 3 lines
except that on 32 TB
systems we don't spend ~2 hours initializing 8,589,934,592 pageheads.
Quoted 2 lines
That's about a million a second which is crazy slow - even my
prehistoric desktop
Quoted 3 lines
is 100x faster than that.
Where's all this time actually being spent?
The complexity of a directory-lookup architecture to make the (intrinsically unscalable) cache-coherency protocol scalable gives you a ~1us roundtrip to remote NUMA nodes.
Quoted 7 lines
Probably a lot of time is spent in some memsets, and RMW cycles which are setting page bits, which are intrinsically synchronous, so the initialising core can't get to 12 or so outstanding memory transactions.
Since EFI memory ranges have a flag to state if they are zerod (which may be a fair assumption for memory on non-bootstrap processor NUMA nodes), we can probably collapse the RMWs to just writes.
A normal write will require a coherency cycle, then a fetch and a writeback when it's evicted from the cache. For this purpose, non-temporal writes would eliminate the cache line fetch and give a massive increase in bandwidth. We wouldn't even need a store-fence as the initialising core is the only one online.
Daniel
Could you elaborate a bit more? or suggest a specific area to look at?
After some experiments with trying to just set some fields in the struct page directly I haven't been able to produce any improvements. Of course there is lots about the area which I don't have much experience with.
Nate
Quoted 35 lines · Nathan Zimmer
Quoted 28 lines · Daniel J Blueman
Quoted 2 lines · Andrew Morton
On Wed, 26 Jun 2013 11:22:48 +0200 Ingo Molnar
Quoted 3 lines · <[email protected]> wrote:
Quoted 3 lines
except that on 32 TB
systems we don't spend ~2 hours initializing 8,589,934,592page heads.
Quoted 2 lines
That's about a million a second which is crazy slow - even my
prehistoric desktop
Quoted 3 lines
is 100x faster than that.
Where's all this time actually being spent?
The complexity of a directory-lookup architecture to make the (intrinsically unscalable) cache-coherency protocol scalable gives you a ~1us roundtrip to remote NUMA nodes.
Probably a lot of time is spent in some memsets, and RMW cycles which are setting page bits, which are intrinsically synchronous, so the initialising core can't get to 12 or so outstanding memory transactions.
Since EFI memory ranges have a flag to state if they are zerod (which may be a fair assumption for memory on non-bootstrap processor NUMA nodes), we can probably collapse the RMWs to just writes.
Quoted 5 lines
A normal write will require a coherency cycle, then a fetch and a writeback when it's evicted from the cache. For this purpose, non-temporal writes would eliminate the cache line fetch and give a massive increase in bandwidth. We wouldn't even need a store-fence as the initialising core is the only one online.
Quoted 6 lines
Could you elaborate a bit more? or suggest a specific area to look at?
After some experiments with trying to just set some fields in the struct page directly I haven't been able to produce any improvements. Of course there is lots about the area which I don't have much experience with.
Any such improvement will at most be in the 10-20% range.
I'd suggest first concentrating on the 1000-fold boot time initialization speedup that the buddy allocator delayed initialization can offer, and speeding up whatever remains after that stage - in a much more development-friendly environment. (You'll be able to run 'perf record ./calloc-1TB' after bootup and get meaningful results, etc.)
Thanks,
Ingo
Quoted 53 lines · Ingo Molnar
Quoted 35 lines · Nathan Zimmer
Quoted 28 lines · Daniel J Blueman
Quoted 2 lines · Andrew Morton
On Wed, 26 Jun 2013 11:22:48 +0200 Ingo Molnar
Quoted 3 lines · <[email protected]> wrote:
Quoted 3 lines
except that on 32 TB
systems we don't spend ~2 hours initializing 8,589,934,592page heads.
Quoted 2 lines
That's about a million a second which is crazy slow - even my
prehistoric desktop
Quoted 3 lines
is 100x faster than that.
Where's all this time actually being spent?
The complexity of a directory-lookup architecture to make the (intrinsically unscalable) cache-coherency protocol scalable gives you a ~1us roundtrip to remote NUMA nodes.
Probably a lot of time is spent in some memsets, and RMW cycles which are setting page bits, which are intrinsically synchronous, so the initialising core can't get to 12 or so outstanding memory transactions.
Since EFI memory ranges have a flag to state if they are zerod (which may be a fair assumption for memory on non-bootstrap processor NUMA nodes), we can probably collapse the RMWs to just writes.
Quoted 5 lines
A normal write will require a coherency cycle, then a fetch and a writeback when it's evicted from the cache. For this purpose, non-temporal writes would eliminate the cache line fetch and give a massive increase in bandwidth. We wouldn't even need a store-fence as the initialising core is the only one online.
Quoted 6 lines
Could you elaborate a bit more? or suggest a specific area to look at?
After some experiments with trying to just set some fields in the struct page directly I haven't been able to produce any improvements. Of course there is lots about the area which I don't have much experience with.
Any such improvement will at most be in the 10-20% range.
I'd suggest first concentrating on the 1000-fold boot time initialization speedup that the buddy allocator delayed initialization can offer, and speeding up whatever remains after that stage - in a much more development-friendly environment. (You'll be able to run 'perf record ./calloc-1TB' after bootup and get meaningful results, etc.)
Quoted 3 lines
Thanks,
Ingo
I had been focusing on the bigger gains but my attention had been diverted by hope of an easy, alibiet smaller, win.
I have been experimenting with the patch proper, I am just doing 2MB pages for the moment. The improvement is vast, I'll worry about proper numbers once I think I have a fully working patch.
Some progress is being made on the real patch. I think the memory is being set up correctly, On aligned pages setting the up the page as normal plus setting new PG_ flag.
Right now I am trying to sort out free_pages_prepare and free_pages_check.
Thanks,
Nate