On Wed, 21 Mar 2018, Christopher Lameter wrote: > One other thought: If you want to improve the behavior for large scale > objects allocated through kmalloc/kmemcache then we would certainly be > glad to entertain those ideas. > > F.e. you could optimize the allcations > 2x PAGE_SIZE so that they do not > allocate powers of two pages. It would be relatively easy to make > kmalloc_large round the allocation to the next page size and then allocate > N consecutive pages via alloc_pages_exact() and free the remainder unused > pages or some such thing. It may be possible, but we'd need to improve the horrible complexity of alloc_pages_exact(). This is a trade-of between performance and waste. A power-of-two allocation can be done quicky, but it wastes a lot of space. alloc_pages_exact() wastes less space, but it is slow. The question is - how many of these large-kmalloc allocations are short-lived and how many are long-lived? I don't know, I haven't measured it. Mikulas
