Opened 15 years ago
Last modified 14 years ago
#40 new task
Add a topology description API
Reported by: | Nicolas Pouillon | Owned by: | Nicolas Pouillon |
---|---|---|---|
Priority: | major | Milestone: | Topology handling |
Component: | drivers | Keywords: | |
Cc: |
Description (last modified by )
Overview
Goals
- Abstract the low-level description APIs
- FlattenedDeviceTrees
- ACPI
- OpenFrimware?
- UEFI ?
- Enumerate the platform
- Component connection topology
- IRQ routing
- Memory configuration (Cacheability, latency, …)
- Deduce the correct OS mapping
- For memory allocators (region API)
- Scheduler lists repartition
- Shortest-path IRQ routing
- SRL tasks/resources mapping
APIs
- Explore the topology
- Extract journey costs
- Query the topology: Get nodes by a selector
- Proximity (
get all rams that are at less than 3 hops than cpu XX
) - Properties (
get all cpus of arch XX
) - Properties (
get all devices of type XX
) -- maybe redundant with hexo's device_s tree
- Proximity (
Optional features discussion
- Do we need to handle loads accounting (NoC load, CPU, memory, …) here ?
(i.e. something like "Ok, for now on I use an avg 1 MBps of NoC path from X to Y, then report to others)
- NoC is tightly coupled to topology, this could be interesting
- CPU is somewhat highly dynamic, even if we can sometimes predict, is it obvious here ?
- Memory has its own allocators, do not duplicate
- If we do not, we probably need a separate accounting library
- Maybe a-la mem_alloc(), per resource
- So we must duplicate topology for NoC accounting
- We can also put this is model-specific calls
libTopology data model
Conceptual
It is a graph of nodes. Nodes are associated to an element of the architecture. Some node examples:
- A routing element (DSPIN NoC (as a whole)), a local interconnect
- A CPU+cache
- A memory bank
- An ICU
- A device
Each node can be associated to a device -- There is at most 1 device per node, some nodes have no associated device (e.g. NoC has no dev)
Structures
struct topo_model_s; // forward decl. see below struct topo_node_s { const struct topo_model_s *model; /// model handling this node void *priv; /// model's private data struct device_s *dev; /// associated device, if any }; /** The only creteria telling two ports are different is private must be different. For models not using private, an useless but different value must be set. */ struct topo_port_s { const struct topo_node_s *node; // node owning the port void *private; // port's private. };
Needed APIs
Device to node bijection
struct device_s *topo_node_to_device(const struct topo_node_s *node); struct topo_node_s *topo_device_to_node(const struct device_s *dev);
Topology exploration
There must be a high-level call to explore the path from a node to another. We will assume:
- There is one such path
- We do not cross address spaces during the exploration (route a data packet, cant transform it in an IRQ)
- There is one address type for destination
High level call which does the routing and accumulates round-trip metrics:
error_t topo_journey_metrics( const struct topo_node_s *start, const struct topo_addr_s *dest, struct topo_metrics_s *result);
(round-trip is sufficient as there is roughly the same energy involved in reads and writes, and also gives the model the choice of what a round-trip involves (totally different for a noc and a bus))
An address is a destination valid for a type of routing exploration, most of the time it will be a bus address, but may be an IRQ number, an USB device no, …
enum topo_addr_type_e { TOPO_ADDR_BUS; TOPO_ADDR_IRQ; }; struct topo_addr_s { enum topo_addr_type_e type; union { struct { uintptr_t address; } bus; struct { size_t line_no; } irq; }; };
Accumulated metrics describe the features of the walked path:
struct topo_metrics_s { size_t hop_count; uint32_t latency; /// cycles ? µs ? uint32_t max_byte_per_sec; /// capped max on path uint32_t power_per_byte; /// overall consumption estimation };
Selection function
enum topo_crit_val_e { TOPO_CRIT_DROP, /// Node is not interesting, stop recursion here in the graph TOPO_CRIT_RECURS, /// Node is not selected, but we can recurs through it TOPO_CRIT_SELECT, /// Node is selected, and may be recursed down }; enum topo_cmp_e { TOPO_CMP_LEFT = -1, /// left better than right TOPO_CMP_EQ = 0, /// left equals right TOPO_CMP_RIGHT = 1, /// right better than left }; struct topo_select_item_s { const struct topo_node_s *node; struct topo_metrics_s metrics; // not a pointer ! }; /** Selector function @param item Node/metrics to filter @param priv Selector private data @returns whether the node is interesting */ typedef enum topo_crit_val_e topo_crit_func_t( const struct topo_select_item_s *item, const void *priv); /** Sort function @param left Node/metrics to compare @param right Node/metrics to compare @param priv Sort private data @returns comparaison result */ typedef enum topo_cmp_e topo_sort_func_t( const struct topo_select_item_s *left, const struct topo_select_item_s *right, const void *priv); error_t topo_select( const struct topo_node_s *start, /// Start node topo_crit_func_t *crit, const void *crit_priv, /// Criterium function, telling whether a node is eligible topo_sort_func_t *sort, const void *sort_priv, /// Sorting function, telling order of nodes size_t max_items, /// Maximal number of nodes to return struct topo_select_item_s *items); /// Output nodes+metrics. Table must be allocated by caller and be able to contain @tt max_items
Needed calls in model-specific functions
/** Initialize a topology node. */ typedef void topo_model_init_func_t(struct topo_node_s *node, const void *param); /** Destroy a topology node. Reclaim all associated memory */ typedef void topo_model_cleanup_func_t(struct topo_node_s *node); /** Compute routing and cost from this node to the next hop, routing towards a given destination @param node Node of our model where we are coming from @param input Input port in node. May be NULL on start node @param dest Destination address for the current address space @param metrics Metrics to update on the go @param peer_input Input port of the next hop */ typedef error_t topo_model_journey_next_func_t( const struct topo_node_s *node, const struct topo_port_s *input, const struct topo_addr_s *dest, struct topo_metrics_s *metrics, struct topo_port_s *peer_input); /** Explore the topology graph. Each call to this function must return next valid port. @param node Current node @param input Port where we came from, exploration should not "go back" to this edge. May be NULL on recursion start @param metrics Metrics up to this node @param item Returned item for next recursion @param state Pointer to a void* where select_next can store its internal state. @tt *state is @tt NULL on first call for each node. Successive calls to select_next for a given node in a given exploration use the same @tt state value. @note As there is no warning for end of iteration, there is no way of calling @tt free(), so @tt *state should be a pointer to preallocated internal data. */ typedef error_t topo_model_select_next_func_t( const struct topo_node_s *node, const struct topo_port_s *input, const struct topo_metrics_s *input_metrics, struct topo_select_item_s *item, void **state);
topo_model_s struct
struct topo_model_s { topo_model_init_func_t *init; topo_model_cleanup_func_t *cleanup; topo_model_journey_next_func_t *journey_next; topo_model_select_next_func_t *select_next; };
Change History (5)
comment:1 Changed 14 years ago by
Description: | modified (diff) |
---|
comment:2 Changed 14 years ago by
Description: | modified (diff) |
---|
comment:3 Changed 14 years ago by
Description: | modified (diff) |
---|
- Fixing the recursion API, we needed the input port when walking
- Adding a selection API. Should be ala SQL select
- a filter
- an order
- a limit
- Selection API still has to be modified
- We still have to tell whether a node is better than another (a sorting function)
comment:5 Changed 14 years ago by
Description: | modified (diff) |
---|
The node/model separation is exactly like the device/driver separation.
We should make topology a driver class. device <-> topo node bijection is then a subdevice handling.
We should merge this with device API rework ticket
Bringing the discussion here