106 lines · plain
1============================2NUMA resource associativity3============================4 5Associativity represents the groupings of the various platform resources into6domains of substantially similar mean performance relative to resources outside7of that domain. Resources subsets of a given domain that exhibit better8performance relative to each other than relative to other resources subsets9are represented as being members of a sub-grouping domain. This performance10characteristic is presented in terms of NUMA node distance within the Linux kernel.11From the platform view, these groups are also referred to as domains.12 13PAPR interface currently supports different ways of communicating these resource14grouping details to the OS. These are referred to as Form 0, Form 1 and Form215associativity grouping. Form 0 is the oldest format and is now considered deprecated.16 17Hypervisor indicates the type/form of associativity used via "ibm,architecture-vec-5 property".18Bit 0 of byte 5 in the "ibm,architecture-vec-5" property indicates usage of Form 0 or Form 1.19A value of 1 indicates the usage of Form 1 associativity. For Form 2 associativity20bit 2 of byte 5 in the "ibm,architecture-vec-5" property is used.21 22Form 023------24Form 0 associativity supports only two NUMA distances (LOCAL and REMOTE).25 26Form 127------28With Form 1 a combination of ibm,associativity-reference-points, and ibm,associativity29device tree properties are used to determine the NUMA distance between resource groups/domains.30 31The “ibm,associativity” property contains a list of one or more numbers (domainID)32representing the resource’s platform grouping domains.33 34The “ibm,associativity-reference-points” property contains a list of one or more numbers35(domainID index) that represents the 1 based ordinal in the associativity lists.36The list of domainID indexes represents an increasing hierarchy of resource grouping.37 38ex:39{ primary domainID index, secondary domainID index, tertiary domainID index.. }40 41Linux kernel uses the domainID at the primary domainID index as the NUMA node id.42Linux kernel computes NUMA distance between two domains by recursively comparing43if they belong to the same higher-level domains. For mismatch at every higher44level of the resource group, the kernel doubles the NUMA distance between the45comparing domains.46 47Form 248-------49Form 2 associativity format adds separate device tree properties representing NUMA node distance50thereby making the node distance computation flexible. Form 2 also allows flexible primary51domain numbering. With numa distance computation now detached from the index value in52"ibm,associativity-reference-points" property, Form 2 allows a large number of primary domain53ids at the same domainID index representing resource groups of different performance/latency54characteristics.55 56Hypervisor indicates the usage of FORM2 associativity using bit 2 of byte 5 in the57"ibm,architecture-vec-5" property.58 59"ibm,numa-lookup-index-table" property contains a list of one or more numbers representing60the domainIDs present in the system. The offset of the domainID in this property is61used as an index while computing numa distance information via "ibm,numa-distance-table".62 63prop-encoded-array: The number N of the domainIDs encoded as with encode-int, followed by64N domainID encoded as with encode-int65 66For ex:67"ibm,numa-lookup-index-table" = {4, 0, 8, 250, 252}. The offset of domainID 8 (2) is used when68computing the distance of domain 8 from other domains present in the system. For the rest of69this document, this offset will be referred to as domain distance offset.70 71"ibm,numa-distance-table" property contains a list of one or more numbers representing the NUMA72distance between resource groups/domains present in the system.73 74prop-encoded-array: The number N of the distance values encoded as with encode-int, followed by75N distance values encoded as with encode-bytes. The max distance value we could encode is 255.76The number N must be equal to the square of m where m is the number of domainIDs in the77numa-lookup-index-table.78 79For ex:80ibm,numa-lookup-index-table = <3 0 8 40>;81ibm,numa-distace-table = <9>, /bits/ 8 < 10 20 80 20 10 160 80 160 10>;82 83::84 85 | 0 8 4086 --|------------87 |88 0 | 10 20 8089 |90 8 | 20 10 16091 |92 40| 80 160 1093 94A possible "ibm,associativity" property for resources in node 0, 8 and 4095 96{ 3, 6, 7, 0 }97{ 3, 6, 9, 8 }98{ 3, 6, 7, 40}99 100With "ibm,associativity-reference-points" { 0x3 }101 102"ibm,lookup-index-table" helps in having a compact representation of distance matrix.103Since domainID can be sparse, the matrix of distances can also be effectively sparse.104With "ibm,lookup-index-table" we can achieve a compact representation of105distance information.106