[tip:x86/mm] x86-64: Combine SRAT regions when possible
in [Kernel]
|
Prev: [tip:x86/cleanups] x86-64: Don't export init_level4_pgt
Next: Very bad regression from 2.6.33 as of 1600f9def
From: tip-bot for Jan Beulich on 28 Apr 2010 21:40 Commit-ID: 2e61878698781d6a9a8bfbaa4ea9c5ddb5a178c3 Gitweb: http://git.kernel.org/tip/2e61878698781d6a9a8bfbaa4ea9c5ddb5a178c3 Author: Jan Beulich <JBeulich(a)novell.com> AuthorDate: Wed, 21 Apr 2010 16:13:20 +0100 Committer: H. Peter Anvin <hpa(a)zytor.com> CommitDate: Wed, 28 Apr 2010 17:14:11 -0700 x86-64: Combine SRAT regions when possible .... i.e. when the hole between two regions isn't occupied by memory on another node. This reduces the memory->node table size, thus reducing cache footprint of lookups, which got increased significantly some time ago, and things go back to how they were before that change on the systems I looked at. Signed-off-by: Jan Beulich <jbeulich(a)novell.com> LKML-Reference: <4BCF3230020000780003B3CA(a)vpn.id2.novell.com> Signed-off-by: H. Peter Anvin <hpa(a)zytor.com> --- arch/x86/mm/srat_64.c | 48 ++++++++++++++++++++++++++++++++++++++++++++++++ 1 files changed, 48 insertions(+), 0 deletions(-) diff --git a/arch/x86/mm/srat_64.c b/arch/x86/mm/srat_64.c index 28c6876..3ebe651 100644 --- a/arch/x86/mm/srat_64.c +++ b/arch/x86/mm/srat_64.c @@ -363,6 +363,54 @@ int __init acpi_scan_nodes(unsigned long start, unsigned long end) for (i = 0; i < MAX_NUMNODES; i++) cutoff_node(i, start, end); + /* + * Join together blocks on the same node, holes between + * which don't overlap with memory on other nodes. + */ + for (i = 0; i < num_node_memblks; ++i) { + int j, k; + + for (j = i + 1; j < num_node_memblks; ++j) { + unsigned long start, end; + + if (memblk_nodeid[i] != memblk_nodeid[j]) + continue; + start = min(node_memblk_range[i].end, + node_memblk_range[j].end); + end = max(node_memblk_range[i].start, + node_memblk_range[j].start); + for (k = 0; k < num_node_memblks; ++k) { + if (memblk_nodeid[i] == memblk_nodeid[k]) + continue; + if (start < node_memblk_range[k].end && + end > node_memblk_range[k].start) + break; + } + if (k < num_node_memblks) + continue; + start = min(node_memblk_range[i].start, + node_memblk_range[j].start); + end = max(node_memblk_range[i].end, + node_memblk_range[j].end); + printk(KERN_INFO "SRAT: Node %d " + "[%Lx,%Lx) + [%Lx,%Lx) -> [%lx,%lx)\n", + memblk_nodeid[i], + node_memblk_range[i].start, + node_memblk_range[i].end, + node_memblk_range[j].start, + node_memblk_range[j].end, + start, end); + node_memblk_range[i].start = start; + node_memblk_range[i].end = end; + k = --num_node_memblks - j; + memmove(memblk_nodeid + j, memblk_nodeid + j+1, + k * sizeof(*memblk_nodeid)); + memmove(node_memblk_range + j, node_memblk_range + j+1, + k * sizeof(*node_memblk_range)); + --j; + } + } + memnode_shift = compute_hash_shift(node_memblk_range, num_node_memblks, memblk_nodeid); if (memnode_shift < 0) { -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |