Shared Persistent Heap Data Environment Manual
1.1.0
|
Shared Persistent Heap, single producer single consumer queue direct API. More...
#include <stdint.h>
#include "sastype.h"
#include "sasatom.h"
#include "sphlfentry.h"
#include "sphsinglepcqueue.h"
Go to the source code of this file.
Macros | |
#define | __C__ |
ignore this macro behind the curtain | |
Typedefs | |
typedef void * | SPHLFEntryDirect_t |
Instance of a Lock Free event direct data Handle. More... | |
Functions | |
static int | SPHLFEntryDirectComplete (SPHLFEntryDirect_t directHandle, sphLFEntryID_t entry_template, int catcode, int subcode) |
Marks the entry specified by the entry handle as complete. Also executes any memory barriers required by the platform to ensure that all previous stores by this thread to this entry are complete. More... | |
static void * | SPHLFEntryDirectGetFreePtr (SPHLFEntryDirect_t directHandle) |
Return the first free byte address for the direct entry specified by the direct entry handle. This is normally the byte after the sphLFEntry_t. More... | |
static void * | SPHLFEntryDirectGetPtrAligned (SPHLFEntryDirect_t directHandle, size_t alignval) |
Return the first free byte address, with required alignment, within the direct entry specified by the direct entry handle. This is normally the address after the sphLFEntry_t plus alignment padding. More... | |
static void * | SPHLFEntryDirectIncAndAlign (void *directptr, size_t incval, size_t alignval) |
Return the next free byte address within direct entry specified by a current address within that direct entry. More... | |
static int | SPHLFEntryDirectIsComplete (SPHLFEntryDirect_t directHandle) |
Return the status of the entry specified by the direct entry handle. More... | |
static int | SPHLFEntryDirectIsTimestamped (SPHLFEntryDirect_t directHandle) |
Return the status of the entry specified by the direct entry handle. More... | |
static int | SPHLFEntryDirectCategory (SPHLFEntryDirect_t directHandle) |
Return the entry category for the entry specified by the direct entry handle. More... | |
static int | SPHLFEntryDirectSubcat (SPHLFEntryDirect_t directHandle) |
Return the entry sub-category for the entry specified by the direct entry handle. More... | |
__C__ sphLFEntryID_t | SPHSinglePCQueueGetEntryTemplate (SPHSinglePCQueue_t queue) |
Return the entry template for an existing Lock Free Single Producer Single Consumer Queue. This template is used later to mark an allocated entry complete. More... | |
__C__ SPHLFEntryDirect_t | SPHSinglePCQueueAllocStrideDirect (SPHSinglePCQueue_t queue) |
Allows the producer thread to allocate and initialize the header of a queue entry for access. The allocation is from the specified Single Producer Single Consumer Queue. More... | |
__C__ SPHLFEntryDirect_t | SPHSinglePCQueueAllocStrideDirectSpin (SPHSinglePCQueue_t queue) |
Allows the producer thread to allocate and initialize the header of a queue entry for access. The allocation is from the specified Single Producer Single Consumer Queue. If space is not Immediately available, spin until it is. More... | |
__C__ SPHLFEntryDirect_t | SPHSinglePCQueueAllocStrideDirectSpinPause (SPHSinglePCQueue_t queue) |
Allows the producer thread to allocate and initialize the header of a queue entry for access. The allocation is from the specified Single Producer Single Consumer Queue. If space is not Immediately available, spin until it is. While spinning use appropriate arch specific instructions to free up core resources for other threads. More... | |
__C__ SPHLFEntryDirect_t | SPHSinglePCQueueGetNextCompleteDirectSpin (SPHSinglePCQueue_t queue) |
Allows the consumer to get the next completed queue entry from the specified single producer single consumer queue. More... | |
__C__ SPHLFEntryDirect_t | SPHSinglePCQueueGetNextCompleteDirectSpinPause (SPHSinglePCQueue_t queue) |
Allows the consumer to get the next completed queue entry from the specified single producer single consumer queue. More... | |
__C__ SPHLFEntryDirect_t | SPHSinglePCQueueGetNextCompleteDirect (SPHSinglePCQueue_t queue) |
Allows the consumer to get the next completed queue entry from the specified single producer single consumer queue. More... | |
__C__ int | SPHSinglePCQueueFreeNextEntryDirect (SPHSinglePCQueue_t queue, SPHLFEntryDirect_t next_entry) |
Allows the consumer to free the queue entry it just processed (using SPHSinglePCQueueGetNextComplete), from the specified single producer single consumer queue. More... | |
__C__ SPHLFEntryDirect_t | SPHSinglePCQueueGetNextEntryDirect (SPHSinglePCQueue_t queue) |
Allows the consumer to get the next allocated queue entry from the specified single producer single consumer queue. More... | |
__C__ int | SPHSinglePCQueueEntryIsCompleteDirect (SPHLFEntryDirect_t directHandle) |
Return the status of the entry specified by the direct entry handle. More... | |
Shared Persistent Heap, single producer single consumer queue direct API.
For shared memory multi-thread/multi-core applications. This implementation is based on the Lock Free Producer/Consumer Queue (SPHSinglePCQueue_t) but simplifies access to the Entry for lower latency.
This API supports atomic allocation of storage for queue entries for zero copy persistence and sharing. Zero copy queues divides the process of producing a queue entry in to three steps:
The consumer can access queue entries once they are marked complete. The consumer:
In this implementation the allocation of the entry is minimally serialized based on the assumption that only one (producer) thread will be allocating queue entries. Likewise the assumption is that there is only one consumer thread per SPHSinglePCQueue_t instance. This allows independent producer/consumer thread pairs to interact with a queue instance with minimum synchronization and overhead.
As an option the queue entry allocator will fill in a 4 byte entry header with:
Any additional storage allocated to the entry (after the header) is available for application specific data. This API also provides a direct pointer mechanism to store application data. The API provides a completion function (SPHSinglePCQueueEntryComplete) which provides any memory barriers required by the platform and marks the entry complete.
The API support simple circular queues and requires a constant entry stride. A stride that matches or is multiple of the cache line size can improve performance by avoiding "false sharing" of cache lines containing multiple queue entries across cores/sockets.
typedef void* SPHLFEntryDirect_t |
Instance of a Lock Free event direct data Handle.
Contains fields required to: locate the entry, record the total space allocated to the entry, and manage the next location within the entry and remaining storage.
Entry Handles should be allocated in private (local stack) storage to allow concurrent access to independent entries from multiple threads.
|
inlinestatic |
Return the entry category for the entry specified by the direct entry handle.
directHandle | Entry Handle for an allocated entry. |
|
inlinestatic |
Marks the entry specified by the entry handle as complete. Also executes any memory barriers required by the platform to ensure that all previous stores by this thread to this entry are complete.
directHandle | Entry Handle for an allocated entry. |
entry_template | from SPHSinglePCQueueGetEntryTemplate(). |
catcode | Category code to the completed entry. |
subcode | Subcategory code to the completed entry. |
|
inlinestatic |
Return the first free byte address for the direct entry specified by the direct entry handle. This is normally the byte after the sphLFEntry_t.
directHandle | Entry Handle for an allocated entry. |
|
inlinestatic |
Return the first free byte address, with required alignment, within the direct entry specified by the direct entry handle. This is normally the address after the sphLFEntry_t plus alignment padding.
directHandle | Entry Handle for an allocated entry. |
alignval | required alignment of the next value to be added. |
|
inlinestatic |
Return the next free byte address within direct entry specified by a current address within that direct entry.
directptr | current data address within Entry. |
incval | size of last entry address to stepped over. |
alignval | required alignment of the next value to be added. |
|
inlinestatic |
Return the status of the entry specified by the direct entry handle.
directHandle | Entry Handle for an allocated entry. |
|
inlinestatic |
Return the status of the entry specified by the direct entry handle.
directHandle | Entry Handle for an allocated entry. |
|
inlinestatic |
Return the entry sub-category for the entry specified by the direct entry handle.
directHandle | Entry Handle for an allocated entry. |
__C__ SPHLFEntryDirect_t SPHSinglePCQueueAllocStrideDirect | ( | SPHSinglePCQueue_t | queue | ) |
Allows the producer thread to allocate and initialize the header of a queue entry for access. The allocation is from the specified Single Producer Single Consumer Queue.
The allocation size is the stride set when the PC queue was initialized/created. The Entry status and length are stored in the header of the new entry. Returns an dire ctentry handle which allows the application to insert application specific data into the entry via the sphlfentry.h API. If the specified queue is full the allocation may fail.
queue | Handle of a producer consumer queue. |
__C__ SPHLFEntryDirect_t SPHSinglePCQueueAllocStrideDirectSpin | ( | SPHSinglePCQueue_t | queue | ) |
Allows the producer thread to allocate and initialize the header of a queue entry for access. The allocation is from the specified Single Producer Single Consumer Queue. If space is not Immediately available, spin until it is.
The allocation size is the stride set when the PC queue was initialized/created. The Entry status and length are stored in the header of the new entry. Returns an dire ctentry handle which allows the application to insert application specific data into the entry via the sphlfentry.h API. If the specified queue is full the allocation may fail.
queue | Handle of a producer consumer queue. |
__C__ SPHLFEntryDirect_t SPHSinglePCQueueAllocStrideDirectSpinPause | ( | SPHSinglePCQueue_t | queue | ) |
Allows the producer thread to allocate and initialize the header of a queue entry for access. The allocation is from the specified Single Producer Single Consumer Queue. If space is not Immediately available, spin until it is. While spinning use appropriate arch specific instructions to free up core resources for other threads.
The allocation size is the stride set when the PC queue was initialized/created. The Entry status and length are stored in the header of the new entry. Returns an dire ctentry handle which allows the application to insert application specific data into the entry via the sphlfentry.h API. If the specified queue is full the allocation may fail.
queue | Handle of a producer consumer queue. |
__C__ int SPHSinglePCQueueEntryIsCompleteDirect | ( | SPHLFEntryDirect_t | directHandle | ) |
Return the status of the entry specified by the direct entry handle.
directHandle | entry Handle for an allocated entry. |
__C__ int SPHSinglePCQueueFreeNextEntryDirect | ( | SPHSinglePCQueue_t | queue, |
SPHLFEntryDirect_t | next_entry | ||
) |
Allows the consumer to free the queue entry it just processed (using SPHSinglePCQueueGetNextComplete), from the specified single producer single consumer queue.
Mark the current queue tail entry as free (unallocated and invalid) and bump the queue tail pointer to the next entry. If the specified queue is empty or the current tail entry is not yet completed the Free may fail.
queue | Handle of a producer consumer queue, |
next_entry | Direct handle of the queue entry to free. |
__C__ sphLFEntryID_t SPHSinglePCQueueGetEntryTemplate | ( | SPHSinglePCQueue_t | queue | ) |
Return the entry template for an existing Lock Free Single Producer Single Consumer Queue. This template is used later to mark an allocated entry complete.
queue | Handle of a producer consumer queue. |
__C__ SPHLFEntryDirect_t SPHSinglePCQueueGetNextCompleteDirect | ( | SPHSinglePCQueue_t | queue | ) |
Allows the consumer to get the next completed queue entry from the specified single producer single consumer queue.
Returns an direct entry handle which allows the application to access the application specific data inserted by the produced thread. If the specified queue is empty or the next queue is not yet completed the get may fail.
queue | Handle of a producer consumer queue. |
__C__ SPHLFEntryDirect_t SPHSinglePCQueueGetNextCompleteDirectSpin | ( | SPHSinglePCQueue_t | queue | ) |
Allows the consumer to get the next completed queue entry from the specified single producer single consumer queue.
Returns an direct entry handle which allows the application to access the application specific data inserted by the produced thread. If the specified queue is empty or the next queue is not yet completed, spin until data is ready.
queue | Handle of a producer consumer queue. |
__C__ SPHLFEntryDirect_t SPHSinglePCQueueGetNextCompleteDirectSpinPause | ( | SPHSinglePCQueue_t | queue | ) |
Allows the consumer to get the next completed queue entry from the specified single producer single consumer queue.
Returns an direct entry handle which allows the application to access the application specific data inserted by the produced thread. If the specified queue is empty or the next queue is not yet completed, spin until data is ready. While spinning use appropriate arch specific instructions to free up core resources for other threads.
queue | Handle of a producer consumer queue. |
__C__ SPHLFEntryDirect_t SPHSinglePCQueueGetNextEntryDirect | ( | SPHSinglePCQueue_t | queue | ) |
Allows the consumer to get the next allocated queue entry from the specified single producer single consumer queue.
Returns an direct entry handle which allows the application to access the entry allocated by the produced thread. If the specified queue is empty or the next queue is not yet allocated the get may fail. Returning a entry does not mean the the producer has completed the entry and the consumer wait/spin (SPHLFEntryDirectIsComplete) for the entry to become complete.
queue | Handle of a producer consumer queue. |